**Figure 9.** (**A**) Immunohistochemical staining of VANGL1 in colorectal adenocarcinoma with an intensity score of 3+ with more than 70% of the stained cells. (**B**) Immunohistochemical staining of VANGL1 in normal sample.

(**A**) (**B**)

β

**Figure 10.** (**A**) Immunohistochemical staining of EGFR in colorectal adenocarcinoma with an intensity score of 2+ with more than 50% of the stained cells. (**B**) Immunohistochemical staining of LEF1 in colorectal adenocarcinoma with a positive expression of more than 20% of the stained cells.

**Figure 11.** (**A**) Immunohistochemical staining of SMAD4 in colorectal adenocarcinoma with an intensity score of 2+ with more than 50% of the stained cells. (**B**) Immunohistochemical staining of SMAD4 in normal sample.

*β*

#### **4. Discussion**

The prediction of CRC progression risk and the identification of novel biomarkers predictive of this risk could represent a relevant advancement [6]. In this study, we evaluated the expression levels of several genes that are involved in LNM and malignant transitions in CRC tissues via RT-qPCR and IHC methods.

Previous investigations showed a high tumor expression level of the *VANGL-1* gene in CRC patients compared with normal tissues [21–24]. Additionally, *VANGL1* gene expression levels have been suggested to play a critical role in CRC progression and to be notably related to tumor stages and LNM [21–24]. These findings are in substantial agreement with our results. Lee and et al. showed that *VANGL1* gene knockdown can decrease the mRNA expression level of *CYKLIND1*, *COX2*, *MMP3*, and *ERK1/2* and reduce tumor growth and invasion. In addition, they found that a high expression level of the *VANGL1* gene was associated with the overexpression of *AP-1* target genes, which have an important role in MAPK signaling in CRC [22]. Oh et al. indicated that *VANGL1* silencing reduced vascular endothelial growth factor A (VEGF-A) and hypoxia-inducible factor 1-alpha (HIF1A). They suggested that the *VANGL1* gene can increase angiogenesis and CRC malignancy [21]. Additionally, our past investigations showed that angiogenesis and the angiogenic factors *VEGF-A* and *HIFA* play an important role in CRC initiation and progression [25–27]. Thus, it seems that the *VANGL1* gene may interact with VEGF-A and HIF1A signaling and enhance tumor malignancy.

In this study, we showed that the *VANGL1* gene was an independent prognostic biomarker for CRC patients. Taken together, these results indicate that the *VANGL1* gene may have a key role in the regulation of several genes, including those involved in angiogenesis. Thus, *VANGL1* could be suggested as a potential biomarker for the prediction of tumor malignancy and targeted therapy in CRC.

*TGF-β* has been suggested to be a tumor suppressor gene able to stop the cell cycle at early stages of tumor, and SMAD proteins, being transcriptional mediators of TGF-β signaling, play a critical role in it [28]. In particular, *SMAD2* is located at 18q21 and plays a role as a tumor suppressor gene [29–31]. As a result of the loss of heterozygosity (LOH) in the 18q21 region, *SMAD2* gene expression is reduced in several cancers and increases cancer progression [31]. However, our findings showed a significant downregulation of *SMAD2* in well-differentiated tumors. Although this finding is substantial in contrast with most of the available data, the complexity of the mutational profile of *SMAD2* [32], its relationships with the other SMAD proteins [33], and the potential role of specific miRNAs on the regulation of *SMAD2* [34] could have contributed to this result.

*SMAD4* that forms a heterotrimer with *SMAD2* and *SMAD3* to exert transcriptional activity plays a crucial role in carcinogenesis [35], and loss of the *SMAD4* gene occurs in about 30% of CRC [36]. It was reported that loss of *SMAD4* was significantly related to CRC progression and metastasis and occurred in late stages [29,37–40]. Previous studies revealed that in colon cancer, the activation of TGF-β signaling induced ERK and P38 signaling and stimulated angiogenesis by *VEGF* upregulation when *SMAD4* was knocked down [36]. Our findings are in agreement with previous investigations [41,42] and indicate that the expression level of the *SMAD4* gene in stage III-IV CRC was lower than that in stage I-II, although this difference did not reach a statistically significant level.

We observed an upregulation of the *TGF-β* gene in stage III-IV CRC samples, although these results were not statistically significant. According to these data, loss of the *SMAD4* gene and upregulation of *TGF-β* may be associated with a more advanced tumor stage and can promote cancer initiation, progression, and metastasis [35].

The NOTCH signaling pathway activity has been reported in several cancers, such as CRC and hepatocarcinoma [43–46]. NOTCH signaling consists of several receptors (NOTCH1–4) and targets genes including the transcription factor *HES1* (HES family basic helix-loop-helix (bHLH) transcription factor 1) [43,47,48]. In addition, *HES1* has several functions, such as intestinal cell stability and apoptosis control [43,49]. Our study showed a downregulation of *NOTCH1* in stage III-IV CRC and a significant correlation between a high expression level of *NOTCH1* and longer patient survival. This is in contrast with a previous study that reported an upregulation of *NOTCH1* in advanced or metastatic CRC patients and a significant association between the upregulation of *NOTCH1* and poor survival [43]. Moreover, we observed a significant overexpression of *NOTCH1* levels in female patients.

In agreement with our findings, several studies demonstrated a significant overexpression of the *HES1* gene in CRC samples compared with a normal tissue [50–52]. Additionally, we found that the overexpression of the *HES1* gene in stage III-IV CRC was more marked than in stage I-II. This result is in agreement with past investigations [43,50,52]. Overall, our data, together with those of others, confirm that NOTCH signaling, especially *NOTCH1* and *HES1*, play a critical role in metastasis and invasion as well as in the activation of several other signaling pathways. The activation of *NOTCH1* is, in fact, able to induce *HES1* and to start cancer progression [53–56].

*IL2RA* and *IL2RB* bind interleukin 2, which is necessary for the stimulation of T-cell immune response, and act as signal transduction factors. A high level of *IL2*, *IL2RA*, and *IL2RB* gene expression and their relationships with tumor progression and malignancy have been previously reported [57]. However, in agreement with our results, Marshall et al. also found no significant association between *IL2RB* gene expression and cancer progression [58].In the present investigation, we observed that a downregulation of *IL2RA* was significantly associated with CRC patients younger than 50 years.

*ANXA3* plays a relevant role in tumor metastasis, invasion, and drug treatment resistance [59–62]. In a previous investigation, a blood-based biomarker panel, also including the *ANXA3* gene, able to stratify subjects according to their relative CRC risk in comparison with an average-risk population, was developed [58]. Similar to this study, a significant upregulation of *ANXA3* has been identified in CRC tissues compared with normal mucosa as well as in several other cancers, such as pancreas, breast, and lung cancers [63,64].

Several findings have shown that the suppression of *ANXA3* upregulation could inhibit cell proliferation and metastasis in CRC [65,66]. Thus, *ANXA3* could be considered a new potential prognostic biomarker and therapeutic target for CRC treatment [66,67]. Upregulation of the *ANXA3* gene and its correlation with gastric tumor size, stage, and LNMs were detected by Wang et al. [67] Moreover, these authors suggested that the overexpression of ANXA3 has a huge effect on gastric cancer malignancy, and it can be used as a novel prognostic biomarker and a suitable target for treatment [67]. Zhou et al. reported that high expression levels of ANXA3 were significantly correlated with breast tumor LNMs and tumor grade, suggesting ANXA3 as a biomarker for breast cancer prognosis [68]. In our study, a significant overexpression of the *ANXA3* gene in well-differentiated tumors compared with poorly differentiated tumors was instead observed. However, we also observed higher levels of *ANXA3* in tumors from patients with LNMs compared with tumors from patients without LNMs, although this difference did not reach statistical significance. Overall, we were only able to partially confirm the observations of other authors who suggested that the *ANXA3* gene may act as an oncogene and play a role in the transformation of a normal tissue into tumor, CRC invasion, and malignancy progression.

*BUB1* acts as a checkpoint factor during cell mitosis and proliferation, and *PCSK7* plays a role in cellular multiplication, mortality, and adhesion. These two genes are involved in cancer metastasis and invasion [69–72]. In our study, we observed a downregulation of the *BUB1* gene in CRC samples compared with normal mucosa, but this difference was not statistically significant. Additionally, previous studies showed a downregulation of the *BUB1* gene in gastric cancer and in CRC [70,73]. Furthermore, in agreement with previous studies, we found an upregulation of *PCSK7* in CRC. On the other hand, Jaaks et al. realized a significant upregulation of the *PCSK7* gene in colon cancer and considered it a potential biomarker [72]. Taken together, it could be suggested that the low expression level of *BUB1* and the upregulation of *PCSK7* have a critical role in malignant transition through colorectal carcinogenesis.

EGFR is a transmembrane receptor that binds to EGF and stimulates cell growth in tissues. Overexpression of the *EGFR* gene has been observed in several cancers, including colorectal, lung, breast, and bladder [74]. It has been reported that the *EGFR* gene may play a role in CRC development [75] since its expression increases with malignant transformation from normal colonic mucosa to metastatic CRC [75,76]. Previous studies, in fact, showed that *EGFR* gene overexpression was significantly related to tumor stages, metastasis, and well-differentiated tumors [75,77]. Although we found a higher expression of *EGFR* in tumor tissues compared with normal tissues, as well as in the right colon compared with the left colon, we did not find correlations between tumor *EGFR* gene expression and clinicopathological features. However, our observations substantially confirm the role of EGFR in CRC progression.

#### **5. Conclusions**

In conclusion, the main results of this study highlight that the expression of the tumor *VANGL1* gene is an independent prognostic biomarker and could be considered a potential predictor for detecting malignancy risk in CRC patients. Additionally, LNMs were highly predicted by the 20-gene panels. However, validation studies including a higher number of patients are required.

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2075-442 6/11/2/126/s1, Table S1: Logistic Regression Test of the 20 Genes (Analysis–Metastasis); Table S2: Logistic Regression Test of the 20 Genes (Analysis–Stage).

**Author Contributions:** Conceptualization, N.P. and E.N.M.; data curation, S.N., E.M., and E.N.M.; formal analysis, M.O.; funding acquisition, H.A.A.; investigation, E.N.M.; methodology, N.P., A.M., K.B., F.A., and E.N.M.; project administration, E.N.M.; resources, H.A.A. and M.R.Z.; software, N.P.; supervision, E.N.M.; validation, Z.P.; visualization, K.N.; writing—original draft, N.P., Z.P., and E.N.M.; writing—review and editing, N.P., S.N., Z.P., E.M., and E.N.M. All authors have read and agreed to the published version of the manuscript.

**Funding:** This project was completely supported and funded by the Gastroenterology and Liver Diseases Research Center, Research Institute for Gastroenterology and Liver Diseases, Shahid Beheshti University of Medical Sciences, with Grant Nos. 858 and 988, and the Medical Ethical Committee of the RCGLD, with Ethics No. IR.SBMU.RIGLD.REC.1396.947.

**Institutional Review Board Statement:** This study was approved by the ethical committee (Ethics No. IR.SBMU.RIGLD.REC.1396.947) of the Research Institute for Gastroenterology and Liver Disease, Shahid Beheshti University of Medical Sciences, Tehran, Iran.

**Informed Consent Statement:** Informed consent was obtained from all subjects involved in the study.

**Data Availability Statement:** The data presented in this study are available on request from the corresponding author.

**Acknowledgments:** The authors would like to thank all the staff of the Cancer Department in the Research Institute for Gastroenterology and Liver Diseases, Shahid Beheshti University of Medical Sciences, Tehran, Iran.

**Conflicts of Interest:** The authors declare no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

#### **References**


## *Article* **Precision Medicine for the Management of Therapy Refractory Colorectal Cancer**

**Hossein Taghizadeh 1,2 , Robert M. Mader 1,2 , Leonhard Müllauer 2,3 , Friedrich Erhart 4 , Alexandra Kautzky-Willer <sup>5</sup> and Gerald W. Prager 1,2, \***


Received: 14 October 2020; Accepted: 9 December 2020; Published: 11 December 2020

**Abstract:** In this analysis, we examined the efficacy, feasibility, and limitations of molecular-based targeted therapies in heavily pretreated metastatic colorectal cancer (mCRC) patients after failure of all standard treatments. In this single-center, real-world retrospective analysis of our platform for precision medicine, we mapped the molecular profiles of 60 mCRC patients. Tumor samples of the patients were analyzed using next-generation sequencing panels of mutation hotspots, microsatellite instability testing, and immunohistochemistry. All profiles were reviewed by a multidisciplinary team to provide a targeted treatment recommendation after consensus discussion. In total, we detected 166 mutations in 53 patients. The five most frequently found mutations were *TP53, KRAS, APC, PIK3CA,* and *PTEN*. In 28 cases (47% of all patients), a molecularly targeted therapy could be recommended. Eventually, 12 patients (20%) received the recommended therapy. Six patients (10%) had a clinical benefit. The median time to treatment failure was 3.1 months. Our study demonstrates the feasibility and applicability of using targeted therapies in daily clinical practice for heavily pretreated mCRC patients. This could be used as a targeted treatment option in half of the patients.

**Keywords:** molecular oncology; precision medicine; colorectal cancer; targeted therapy; molecular profiling

#### **1. Introduction**

Colorectal cancer (CRC) is one of the most frequent cancer types and is a major cause of mortality and morbidity. According to GLOBOCAN 2018, CRC is the fourth most common cancer disease throughout the world, equally affecting both men and women, with over 2 million new cases in 2018 [1]. It accounts for approximately 1 million deaths annually and 11% of all cancer deaths, ranking as the third most common cause of cancer death [1]. CRC particularly affects developed countries where inhabitants follow a western lifestyle that bears important risk factors for the carcinogenesis of CRC including alcohol intake, tobacco use, immoderate red and processed meat consumption, low intake of fiber, obesity, and a sedentary lifestyle [2].

In recent years, considerable effort has been made to explore the complex tumor biology of CRC and to expand and enrich the therapeutic armamentarium with new therapeutic agents. The therapeutic landscape for the management of metastatic CRC (mCRC) is rapidly evolving.

The development and application of monoclonal antibodies and tyrosine kinase inhibitors in addition to systemic cytotoxic (poly)chemotherapy have significantly improved the prognosis, median overall survival, and quality of life of mCRC patients [3,4].

Despite diagnostic and therapeutic advances in the management of mCRC, the 5-year survival rate for mCRC is approximately 14%, with a median overall survival of 30 months [5,6]. Moreover, after failure of standard therapy lines, therapeutic options are limited.

One way to offer treatment concepts for therapy refractory mCRC would be to analyze the molecular profile of tumors to identify actionable pathologic molecular alterations to develop an individually coordinated therapy plan. This individually tailored, tissue-agnostic molecular-based treatment approach is referred to as precision medicine in oncology or simply precision oncology [7,8].

In the last few years, more and more targeted therapy agents have been introduced for the management of several cancer diseases, such as trastuzumab in human epidermal growth factor receptor 2 (HER2 positive) breast cancer or gastric cancer [9,10], imatinib in in KIT+ gastrointestinal stromal tumor (GIST) [11], and B-rapidly accelerated fibrosarcoma (BRAF)-directed therapy with vemurafenib or dabrafenib/trametinib in melanoma [12].

Thus, exploring the molecular profile of mCRC may aid in the development of molecular targeted therapies and allow their efficacy to be tested.

In this study, we conducted a retrospective subgroup analysis of all 60 patients with advanced therapy refractory mCRC that had been enrolled and profiled via our special platform for precision medicine of the Comprehensive Cancer Centre of the Medical University of Vienna (CCC-MUV).

We sought to map the molecular profiles of mCRC to identify and target specific molecular alterations. We discuss the challenges, limitations, and the time to treatment failure (TTF) of precision medicine approaches in this patient group.

#### **2. Methods**

#### *2.1. Patients and Design of the Precision Medicine Platform*

All patients with heavily pretreated advanced metastatic CRC who had progressed to all standard treatment options and had undergone molecular profiling from June 2013 to June 2020 were included in this retrospective single-center study. This study was conducted at the Clinical Division of Oncology of the tertiary care university hospital Medical University of Vienna. Cancer patients refractory to all standard therapies were eligible for inclusion in our precision medicine platform, provided that tissue samples for molecular profiling were available. The specimens were either obtained by fresh tumor biopsy performed by physicians at the Department of Interventional Radiology or were provided by the archives of the Department of Pathology when tumor biopsy was not feasible. Patients had to have an Eastern Cooperative Oncology Group (ECOG) performance status of ≤1. All patients in this analysis had to be at least 18 years at the time of molecular analysis and had to provide informed consent before inclusion in our platform. Our precision medicine platform is not a clinical trial but intends to provide targeted therapy recommendations to patients where no standard antitumoral treatment is available. This analysis was approved by the Institutional Ethics Committee of the Medical University of Vienna (Nr. 1039/2017). The General Hospital of Vienna directly covered all costs for molecular profiling and targeted therapy provided the cancer patients had no further standard treatment options.

#### *2.2. Evaluation of Outcome and Follow-Up*

All patients with heavily pretreated advanced metastatic CRC who had progressed to all standard treatment options were confirmed by the response evaluation criteria in solid tumors 1.1 (RECIST 1.1) criteria [13]. These international criteria provide a basis for standardized and objective assessment of the change in tumor burden during treatment. The criteria distinguish four types of change:


Follow-up was done every 8 to 12 weeks for outcome evaluation by radiological assessment depending on the respective therapy. If the patient did not appear on the follow-up date, we searched our electronic data processing system that is linked to the national death register to check and ascertain the death of the patient in the meantime.

#### *2.3. Tissue Samples*

Formalin-fixed, paraffin-embedded tissue samples from patients with metastatic CRC who had progressed through all standard therapy regimens were obtained from the archives of the Department of Pathology, Medical University of Vienna, Austria.

#### *2.4. Cancer Gene Panel Sequencing*

DNA was extracted from paraffin-embedded tissue blocks with a QIAamp Tissue KitTM (Qiagen, Hilden, Germany). From each tissue sample, 10 ng of DNA was provided for sequencing. The DNA library was created by multiplex polymerase chain reaction with the Ion AmpliSeq Cancer Hotspot Panel v2 (Thermo Fisher Scientific, Waltham, MA, USA), which covers mutation hotspots of 50 genes. The panel includes driver mutations, oncogenes, and tumor suppressor genes. In mid-2018, the gene panel was expanded using the 161-gene next-generation sequencing panel of Oncomine Comprehensive Assay v3 (Thermo Fisher Scientific, Waltham, MA, USA), which covers genetic alterations, gene amplifications, and gene fusions. The Ampliseq cancer hotspot panel was sequenced with an Ion PGM (Thermo Fisher) and the Oncomine Comprehensive Assay v3 on an Ion S5 sequencer (Thermo Fisher Scientific, Waltham, MA, USA). The generated sequencing data were analyzed afterwards with the help of Ion Reporter Software (Thermo Scientific Fisher). We referred to the BRCA Exchange, ClinVar, COSMIC, dbSNP, OMIM, and 1000 genomes for variant calling and classification. The variants were classified according to a five-tier system comprising pathogenic, likely pathogenic, uncertain significance, likely benign, or benign modifiers. This classification was based on the standards and guidelines for the interpretation of sequence variants of the American College of Medical Genetics and Genomics [14]. The pathogenic and likely pathogenic variants were taken into consideration for the recommendation of targeted therapy.

#### *2.5. Immunohistochemistry*

Immunohistochemistry (IHC) was performed using 2-µm-thin tissue sections that were read by a Ventana Benchmark Ultra stainer (Ventana Medical Systems, Tucson, AZ, USA). The following antibodies were applied: anaplastic lymphoma kinase (ALK) (clone 1A4; Zytomed, Berlin, Germany); CD20 (clone L26; Dako); CD30 (clone BerH2; Agilent Technologies, Vienna, Austria); DNA mismatch repair (MMR) proteins including MLH1 (clone M1, Ventana Medical Systems), PMS2 (clone EPR3947, Cell Marque, Rocklin, CA, USA), MSH2 (clone G219-1129, Cell Marque), and MSH6 (clone 44, Cell Marque); epidermal growth factor receptor (EGFR) (clone 3C6; Ventana); estrogen receptor (clone SP1; Ventana Medical Systems); human epidermal growth factor receptor 2 (HER2) (clone 4B5; Ventana Medical Systems); HER3 (clone SP71; Abcam, Cambridge, UK); C-kit receptor (KIT) (clone 9.7; Ventana Medical Systems); MET (clone SP44; Ventana); NTRK (clone EPR17341, Abcam); phosphorylated mammalian target of rapamycin (p-mTOR) (clone 49F9; Cell Signaling Technology, Danvers, MA, USA); platelet-derived growth factor alpha (PDGFRA) (rabbit polyclonal; Thermo Fisher Scientific); PDGFRB (clone 28E1, Cell Signaling Technology); programmed death-ligand 1 (PD-L1) (clone E1L3N; Cell Signaling Technology till mid-2018, as of mid-2018 the clone BSR90 from Nordic

Biosite, Stockholm, Sweden is used); progesterone receptor (clone 1E2; Ventana); phosphatase and tensin homolog (PTEN) (clone Y184; Abcam); and ROS1 (clone D4D6; Cell Signaling Technology).

To assess theimmunostainingintensity for the antigens EGFR, p-mTOR, PDGFRA, PDGFRB, and PTEN, a combinative semiquantitative score for immunohistochemistry was used. The immunostaining intensity was graded from 0 to 3 (0 = negative, 1 = weak, 2 = moderate, and 3 = strong). To calculate the score, the intensity grade was multiplied by the percentage of corresponding positive cells: (maximum 300) = (% negative × 0) + (% weak × 1) + (% moderate × 2) + (% strong × 3).

The immunohistochemical staining intensity for HER2 was scored from 0 to 3+ (0 = negative, 1+ = negative, 2+ = positive, and 3+ = positive) pursuant to the scoring guidelines of the Dako HercepTestR from the company Agilent Technologies (Agilent Technologies, Vienna, Austria). In case of HER2 2+, a further test with HER2 in situ hybridization was performed to verify HER2 gene amplification.

Estrogen receptor and progesterone receptor staining were graded according to the Allred scoring system from 0 to 8. MET staining was scored from 0 to 3 (0 = negative, 1 = weak, 2 = moderate, and 3 = strong) based on a paper by Koeppen and coworkers. [15].

For PD-L1 protein expression, the tumor proportion score, which is the percentage of viable malignant cells showing membrane staining, was calculated. In addition, since 2019, the expression has also been determined by a combined positive score.

ALK, CD30, CD20, and ROS1 staining were classified as positive or negative based on the percentage of reactive tumor cells, however without graduation of the staining intensity. In ALK- or ROS1-positive cases, the presence of possible gene translocation was evaluated by fluorescence in situ hybridization (FISH).

All antibodies used in this study were validated and approved at the clinical institute of pathology of the Medical University of Vienna and are used in routine IHC staining for clinical purposes. The antibodies have been validated by proper positive and negative tissue controls and by non-IHC methods, such as immunoblotting and flow cytometry, to detect the respective epitopes of the antigens. For control, the use of the antibodies was optimized in terms of intensity, concentration, signal/noise ratio, incubation times, and blocking. The negative control involved omitting the primary antibody and substituting an isotype-specific antibody and serum at the exact same dilution and laboratory conditions as the primary antibody to preclude unspecific binding.

For positive control, the antibodies were shown not to cross-react with closely related molecules of the target epitope.

The status of microsatellite instability-high (MSI-H) was analyzed by the MSI Analysis System, version 1.1 (Promega Corporation, Madison, WI, USA).

#### *2.6. Fluorescence in situ Hybridization (FISH)*

FISH was only applied in selected cases to verify PTEN loss. FISH was performed with 4-µm-thick formalin-fixed, paraffin-embedded tissue sections. The following FISH probes were utilized: PTEN (10q23.31)/centromere 10 (ZytoVision, Bremerhaven, Germany). Two hundred cell nuclei were evaluated per tumor. The PTEN FISH was considered positive for PTEN gene loss with ≥30% of cells with only one or no PTEN signal. A chromosome 10 centromere FISH probe served as a control for the ploidy of chromosome 10.

#### *2.7. Multidisciplinary Team for Precision Medicine*

After thorough examination of the molecular profile of each tumor sample by a qualified molecular pathologist, the results were reviewed by a multidisciplinary team (MDT) that met every other week.

Members of the MDT included molecular pathologists, radiologists, clinical oncologists, surgical oncologists, and basic scientists. The MDT recommended targeted therapy based on the specific molecular profile of each patient, which included established pathological parameters. The targeted therapies included tyrosine kinase inhibitors, checkpoint inhibitors (e.g., anti-PD-L1 monoclonal

antibodies), and growth factor receptor antibodies with or without endocrine therapy. The treatment recommendations by the MDT were prioritized using the level of evidence from high to low according to phase III to phase I trials.

In cases where more than one druggable molecular aberration was identified, the MDT recommended a therapy regimen to target as many molecular aberrations as possible, with special consideration given to the toxicity profile of each antitumoral agent and its potential interactions. Since all patients were given all available standard treatment options for their cancer disease prior to inclusion in our precision medicine platform, nearly all targeted agents suggested had off-label use. If the tumor profile and clinical characteristics of a patient met the requirements to be enrolled in a recruiting clinical trial for targeted therapies at our cancer center, patients were asked whether they wanted to participate in the respective trial, and trials adhered to ethical and regulatory guidelines.

#### *2.8. Study Design and Statistics*

The Fisher's exact test was employed to explore potential gender-specific differences regarding the therapy recommendation rate and the molecular profile. Student's *t*-test was applied to test differences in the outcome between tumor samples that were obtained during tumor biopsy versus specimens that were obtained during surgical resection. A *p*-value of less than 0.05 was considered to be statistically significant. For statistical analysis, the software package IBM SPSS Statistics Version 26 was used.

#### **3. Results**

#### *3.1. Patient Characteristics*

From the initiation of our platform for precision medicine in June 2013 until June 2020, we identified 60 patients with therapy refractory mCRC with no further standard treatment option available. These patients were all included in this subgroup analysis of our platform. All 60 patients were Caucasian, including 42 men (70%) and 18 women (30%). The cohort of mCRC patients comprised 47 patients (78%) with left-sided CRC and 13 patients (22%) with right-sided CRC (see Figure 1 for the patients' flow and Table 1 for the patient characteristics).

The median age at first diagnosis was 54.9 years (range: 16.9 to 81.2 years), and the median age at the time of molecular profiling was 57.9 years (range: 19.3 to 84.3 years). Tumor tissue used for molecular profiling was obtained by biopsy in 25 patients (42%) or during surgical treatment in the other 35 patients (58%). Biopsy was performed after failure of all standard treatment options. The median time interval between resection and molecular analysis of the tumor tissue was 11.9 months (range: 1–46 months). The median time interval between biopsy and completion of the molecular analysis of the tumor tissue was 31 days (range: 15–56 days). The median turnaround time between the initiation of molecular profiling and discussion by the MDT and molecular-based therapy initiation for the 12 patients who received the targeted therapy was 30 and 42 days, respectively. The median time interval between the initiation of molecular profiling and discussion of the MDT and molecular-based therapy initiation for the other patients (*n* = 48) was 29 days.

Twenty-one patients experienced a disease relapse. All of the patients had metastases, mainly in the liver, lungs, and bones. Seventeen patients had additional intraperitoneal dissemination of the CRC, causing peritoneal carcinomatosis. The patients received a median of three lines of prior palliative systemic chemotherapy, ranging from two to six lines. The palliative therapy regimens included FOLFOX + cetuximab, FOLFOX + bevacizumab, FOLFIRI + cetuximab, FOLFOX + panitumumab, FOLFIRI + panitumumab, FOLFOXIRI + bevacizumab, FOLFIRI + bevacizumab, FOLFIRI + aflibercept, FOLFIRI + ramucirumab, regorafenib, trifluridine/ tipiracil, and raltitrexed + oxaliplatin. Twenty-six patients (43%) had received at least four lines of palliative chemotherapy prior to molecular profiling.

**Figure 1.** Flow chart of the 60 metastatic colorectal cancer (mCRC) patients.

#### *3.2. Molecular Profile*

In total, we detected 166 mutations in 53 patients (88%). The five most frequent mutations were *TP53* (*n* = 36; 60.0%), *KRAS* (*n* = 29; 48.3%), *APC* (*n* = 15; 25.0%), *PIK3CA* (*n* = 9; 15%), and *PTEN* (*n* = 8; 13.3%), which accounted for more than half of all mutations (58.4%). No mutations were detected in seven (12%) patients with our sequencing panel (Table 2). Seven gene fusions were identified in five patients: *FGFR3-TACC3* (*n* = 2), *FNDC3B-PIK3CA*, *SND1-BRAF*, *EIF3E-RSPO2*, *PTPRK-RSPO3*, and *WHSC1L1-FGFR1*. Moreover, we detected eight gene amplifications in six different tumor specimens, including *CCND2 (n* = *3), FLT3 (n* = *3), FGFR1,* and *MYC.*

Further, IHC revealed common expressions of phosphorylated mTOR and EGFR in 49 (82%) and 45 (75%) patients, respectively. The median IHC scores of mTOR and EGFR were 100 and 90, respectively. Ten patients (17%) had high levels of phosphorylated mTOR expression with an mTOR score between 200 and 300. EGFR expression was between 200 and 300 in nine patients (15%). In our cohort, two patients (3%) were HER2-positive and seven patients (12%) were HER3-positive. IHC identified six patients (10%) with a loss of PTEN, which was subsequently verified and characterized by FISH as heterozygous PTEN deletions. High expressions levels were also observed for MET (*n* = 28) and PDGFRA (*n* = 15). Four patients (7%) were given a status of MSI-H. In four patients, the PD-L1 combined positive score was ≥1. Three patients displayed a weak KIT expression. The expression of other markers was not observed. IHC and FISH could not be performed for one male patient due to insufficient tumor material. See Figure 2a–n.


**Table 1.** Patient characteristics (*n* = 60).

#### *3.3. Therapy Recommendations and Outcome*

In 28 cases (47% of all patients), a molecularly targeted therapy was recommended. The other 32 patients (53%) did not qualify for targeted therapy due to the lack of actionable molecular targets. In over two-thirds of all recommendations (*n* = 20/28, 71.0%), the molecular-driven treatment approach was mainly derived from the molecular characteristics determined by immunohistochemistry. The 28 recommended targeted treatments included everolimus, pembrolizumab, nintedanib, cetuximab, vismodegib, vemurafenib, afatinib and trastuzumab combined with lapatinib, crizotinib, erlotinib, and sunitinib plus capecitabine. Table 3 describes the rationale for the recommended targeted therapy approaches. Eventually, 12 patients (20%) received the recommended targeted therapy. One patient with the KIT mutation was included in the clinical phase II trial SUNCAP and was treated with sunitinib and capecitabine. Another patient died prior to restaging. Finally, ten patients underwent radiological assessment (see Table 4). Four patients (7%) experienced progressive disease. Four patients with MSI-H status were given pembrolizumab and achieved a partial response (*n* = 2), complete response, and stable disease, respectively. Two patients treated with other targeted agents achieved a stable disease. Thus, the disease control rate was 10%. The three patients who achieved therapy response and one of the three patients who had stable disease experienced also an improvement in their quality of life due to an improvement in tumor pain intensity. The molecularly targeted therapies applied were pembrolizumab (*n* = 4), trastuzumab (*n* = 2), everolimus plus bevacizumab (*n* = 2), afatinib, everolimus, sunitinib plus capecitabine, and everolimus plus raltitrexed (see Tables 3 and 4 for further information). The median time to treatment failure (TTF) in patients who received the targeted therapy was 3.1 months (range: 0.3–30.6 months; see Figure 3 and Table 4). The median overall survival (mOS)

of these 12 patients after initial diagnosis of mCRC was 50.1 months. The mOS after initiation of targeted therapy was 10.9 months (see Figure 4a,b). Three patients were lost to follow-up after the suggestion of molecular-driven targeted therapy. Thirteen patients (22%) did not receive the offered targeted therapy. Reasons for not applying the recommended targeted agent included the following: rapid deterioration of performance status (*n* = 10), death of patient (*n* = 1), the treating oncologist favoring another treatment regimen due to the clinical overall situation of the patients, or patients' refusal of any further treatment including targeted therapy options (*n* = 2).


**Table 2.** Genomic profile of the therapy refractory CRC patients (*n* = 60).

(**g**) p-mTOR low (**h**) p-mTOR high

**Figure 2.** *Cont.*

(**m**) PTEN negative (**n**) PTEN high

**Figure 2.** (**a**––**n**) These original images of immunohistochemistry show the differences between low and high expressions of various markers (Images by kind courtesy of Professor Dr. Müllauer).




#### **Table 3.** *Cont.*

ABL1, Abelson murine leukemia viral oncogene homolog 1; AML, acute myeloid leukemia; ALL, acute lymphatic leukemia; BCR, breakpoint cluster region; CML, chronic myeloid leukemia; CRC, colorectal cancer; EGFR epidermal growth factor receptor; EMA, European Medicines Agency; FDA, Food and Drug Administration; FLT3, fms like tyrosine kinase 3; GIST, gastrointestinal stromal tumor; GNRHR, gonadotropin-releasing hormone receptor; HER2, human epidermal growth factor receptor 2; HL, Hodgkin lymphoma; HNSCC, head and neck squamous cell carcinoma; MCL, mantle cell lymphoma; MDS, myelodysplastic syndrome; MPD, myeloproliferative disorder; NSCLC, non-small-cell lung carcinoma; PD, progressive disease; PD-1, programmed cell death protein 1; PDAC, pancreatic ductal adenocarcinoma; PDGFR, platelet-derived growth factor receptor; Ph+: Philadelphia chromosome positive; p-mTOR, phosphorylated mammalian target of rapamycin; RCC, renal cell carcinoma; RET, rearranged during transfection; SD, stable disease; VEGFR, vascular endothelial growth factor.

phase II SUNCAP trial.


**Table 4.**Characteristics of the CRC patients receiving the molecular-based targeted therapy recommendation (*n*=12).

**Table 4.** *Cont.*


n.a., not applicable; AR, androgen receptor; CPS, combined prognostic score; ECOG PS, Eastern Cooperative Oncology Group performance status; EGFR epidermal growth factor receptor; MSI-H, microsatellite instability-high; PD, progressive disease; PDL1, programmed death ligand 1; PDGFRA, platelet-derived growth factor receptor alpha; p-mTOR, phosphorylated mammalian target of rapamycin; SD, stable disease, PTEN, phosphatase and tensin homolog; TPS, tumor positive score.

> 0 5 10 15 20 25 30 35 1 2 3 4 5 6 7 8 9 10 11 Time in months Patients Time to treatment failure

**Figure 3.** Time to treatment failure (TTF) in 11 CRC patients who received the recommended targeted therapy: the median time to treatment failure was 3.1 months.

**Figure 4.** (**a**) Kaplan–Meier survival curve showing overall survival after initiation of targeted therapy in twelve patients receiving the targeted therapy and (**b**) Kaplan–Meier survival curve showing overall survival after initial diagnosis of mCRC in twelve patients receiving the targeted therapy.

Three tumor specimens from the ten patients who underwent radiological assessment were obtained during a conventional tumor biopsy. Two of these three patients experienced progressive disease, and one patient achieved stable disease. The tumor sample of the remaining seven patients was yielded during surgical resection of the primary tumor. We found no significant difference between tumor samples obtained during tumor biopsy versus surgical resection in terms of TTF (*p* = 0.319) and mOS (*p* = 0.396).

According to the Fisher's exact test, the gender-specific differences regarding the 28 targeted therapy recommendations were not statistically significant (*p* = 0.516).

#### **4. Discussions**

In our study, we recommended 28 molecularly targeted therapies based on the respective individual molecular profile of heavily pretreated mCRC patients. Thus, precision medicine approaches were found to be feasible and implementable in daily clinical routine in approximately half of the patients who had no further standard treatment option. In over two-thirds of all recommendations, the molecular-based targeted treatment approach was mainly derived from the molecular characteristics determined by immunohistochemistry. This fact underlines the major clinical relevance of immunohistochemistry in precision medicine as immunohistochemistry and next-generation sequencing complement each other. This could be very useful as sequencing panels are updated and enlarged routinely, adding valuable information for treatment decision making.

However, one important limitation of this study was that other parts of the molecular portrait were not analyzed. The molecular profile of a tumor is intricate and complex and goes beyond these two techniques. Comprehensive mapping of the molecular profile is multilayered and multi-faceted and includes many other aspects, including genomics, epigenomics, transcriptomics, proteomics including phospho- and glycoproteomics, metabolomics, epigenetics, and microbiomics [16]. The processing and integration of these extremely large quantities of data and their translation into targeted therapy recommendations is a grand challenge that scientists and clinicians are confronted with. There are close links among all involved disciplines to achieve common objectives. Further, CRC is characterized by highly dynamic and complex molecular intratumoral and intertumoral heterogeneity that changes both temporally and spatially [16–21]. The tumor tissue used for molecular profiling was obtained by biopsy in 25 patients (42%) or during surgical treatment in the other 35 patients (58%). Biopsy was performed after failure of all standard treatment options. Whenever possible, we used metastatic tissue for molecular profiling, which was particularly suited when fresh biopsies were obtained. If this approach was not feasible, e.g., if the anatomic site was not suitable for biopsy, we used the information obtained from the primary tumor site. Despite potential spatial heterogeneity, we assumed that most of the genetic aberrations in the primary cancer were also present at the metastatic site. Studies have shown that there is a high biomarker concordance between primary colorectal cancer and its metastases [22]. We found no significant difference between tumor samples obtained during tumor biopsy versus surgical resection in terms of TTF. However, the number of patients (n = 10) who underwent radiological response assessment after application of the targeted therapy was too limited to examine the influence of biomarker concordance on the outcome. Thus, further clinical trials and studies are required to examine the degree of biomarker concordance between primary cancer and metastases.

In our cohort, the five most frequent mutations, *TP53, KRAS, APC, PIK3CA,* and *PTEN*, together accounted for more than 50% percent of all detected mutations. Except for PIK3CA mutations, there are still no molecularly targeted therapies that directly target the mutations in *TP53*, *KRAS*, *APC*, and *PTEN*. Thus, there is an unmet clinical need for the inhibition of these genetic aberrations. The rest of the detected mutations were of low frequency (below 10%) and reflect the well-known molecular heterogeneity and diversity of CRC. The detected genetic aberrations are in line with the results of previous studies [23–25].

Moreover, a growing body of evidence shows that the antitumoral therapy itself may affect, influence, and drive tumor molecular evolution [26–28]. A prime example of this phenomenon is the recommendation of cetuximab for two patients who were initially KRAS-mutated and were therefore not treated with an anti-EGFR therapy; however, at the time of molecular profiling, they had developed the RAS wildtype [29].

One way to monitor the dynamic molecular landscape of cancer disease would be the utilization of real-time liquid biopsy to adapt antitumoral therapy according to the current molecular portrait [30]. To this end, the multicenter clinical phase II trial MoliMor (EudraCT number: 2019–003714–14) is evaluating the efficacy and safety of intermittent addition of cetuximab to a FOLFIRI-based first-line therapy to patients with RAS-mutant mCRC at diagnosis who convert to the RAS wildtype using monitoring of the RAS mutation status by liquid biopsy. Liquid biopsy may be also suitable for therapy response and for the detection of early signs of therapy resistance. Furthermore, the application of liquid biopsy may also help to reduce the long turnaround time of one month from biopsy of the tumor tissue to completion of the molecular profile. One of the main limitations of this study was the relatively long turnaround time between biopsy and completion of molecular analysis and between initiation of molecular profiling and discussion of the MDT and molecular-based therapy initiations with 30 and 42 days, respectively.

Time is a highly critical factor in the therapeutic management of mCRC, and a turnaround time of over one month without administration of effective therapy means that the mCRC progresses further. The growing metastases, particularly liver metastases, may lead to liver failure, increasing bilirubin values, and rapid health deterioration, making administration of the recommended targeted therapy impossible. From 28 molecularly targeted therapy recommendations, less than half of the patients eventually received the therapy. The TTF of 3.1 months and the disease control rate of 10% are modest outcomes. One reason for these modest outcomes may be that, due to the long turnaround time, there was not enough time for the targeted therapy to display its full potential. Other reasons may be the aforementioned tumor heterogeneity and the non-consideration of other aspects of the molecular profile as an important limitation. Interestingly, the three patients who achieved therapy response and one of the three patients who had stable disease experienced also improvements in their quality of life due to improvements in tumor pain intensity.

Furthermore, the design of this study is retrospective and it may be associated—in contrast to a prospective randomized controlled trial—with several limitations, including selection bias, insufficient documentation of clinical data, inadequate consideration of potential confounders, and the lack of randomization.

An additional limiting factor is that this retrospective study was conducted at a single center with a relatively small number of patients. It is difficult to demonstrate treatment efficacy, treatment difference, or certain findings in a small sample. In addition, single-center studies lack external validity to support or confirm the findings. Further research and clinical trials are warranted to evaluate the role and value of precision medicine for the management of mCRC patients.

In our study, we paid close attention to potential gender-specific differences. We did not find any gender-specific differences regarding the 28 targeted therapy recommendations.

Our study emphasizes the relevance and efficacy of pembrolizumab, even in metastatic patients with heavily pretreated therapy refractory mCRC who were classified as MSI-H.

It was shown that MSI-H status is a favorable prognostic factor at early local stages, particularly in stage II [31]. However, in stage IV CRC, MSI-H confers an inferior prognosis when compared to microsatellite stable metastatic CRC [32–34].

In our limited subset of patients with MSI-H, we administered pembrolizumab to four patients as no other druggable target could be derived from the molecular profile. Although MSI-H is not a favorable predictive marker in stage IV CRC, we achieved an impressive response with one complete remission, two partial remissions, and one stabile disease.

Pembrolizumab was the first tissue-agnostic treatment that was granted approval by the Food and Drug Administration (FDA) by mid-2017 for patients with unresectable or metastatic MSI-H or with mismatch repair deficient (dMMR) solid tumors that progress following prior treatment and who have no satisfactory alternative treatment options or those with MSI-H or dMMR colorectal cancer that progresses following treatment with a fluoropyrimidine, oxaliplatin, and irinotecan [35]. However, its use for MSI-H patients has not been still approved by the European Medicines Agency (EMA), and it was applied in off-label form in this study.

Taken together, the management of mCRC patients poses several major challenges, including the long turnaround time and the complex molecular heterogeneity of CRC. Our study underscores the relevance of immunohistochemistry and underlines the importance of time as a highly critical factor in precision medicine. Based on our study, molecular-based treatment approaches can be of clinical benefit in select heavily pretreated mCRC patients. In this study, the overall benefit for precision medicine approaches was limited and the TTF was relatively modest. However, precision medicine is a rapidly evolving field. In the next few years, technical advances will allow us to employ larger gene panels to cover and identify more mutations, amplifications, deletions, and gene fusions in a shorter period of time. The development of new and potent molecularly targeted therapies together with technical progresses in molecular profiling allow us to hope that, in the future, we may be able to yield deep and durable responses in heavily pretreated mCRC patients.

**Author Contributions:** Conceptualization, H.T., R.M.M. and G.W.P.; methodology, H.T., R.M.M., and G.W.P.; formal analysis, H.T., R.M.M., L.M., F.E., A.K.-W. and G.W.P.; investigation, H.T., R.M.M., G.W.P.; resources, H.T., R.M.M., L.M., F.E., A.K.-W. and G.W.P.; data curation, H.T., R.M.M., L.M., F.E. and G.W.P.; writing—original draft preparation, H.T.; writing—review and editing, all authors.; visualization, H.T.; supervision, R.M.M. and G.W.P.; project administration, H.T., R.M.M., L.M. and G.W.P. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

*Article*

## **Prognostic Impact of the Neutrophil-to-Lymphocyte and Lymphocyte-to-Monocyte Ratio, in Patients with Rectal Cancer: A Retrospective Study of 1052 Patients**

**Zsolt Zoltán Fülöp 1,**† **, Réka Linda Fülöp 1,**† **, Simona Gurzu 2,3, \* , Tivadar Bara, Jr. 1 , József Tímár 4 , Em ˝oke Drágus <sup>5</sup> and Ioan Jung 2**


Received: 27 August 2020; Accepted: 14 October 2020; Published: 16 October 2020

**Abstract:** Despite the description of several new prognostic markers, colorectal cancer still represents the third most frequent cause of cancer-related death. As immunotherapy is considered a therapeutic alternative in such patients, neutrophil-to-lymphocyte (NLR) and lymphocyte-to-monocyte ratio (LMR) are hypothesized to provide reliable prognostic information. A retrospective study was conducted on 1052 patients operated on during 2013–2019 in two clinical hospitals from Hungary and Romania. Inclusion criteria targeted patients over 18 years old, diagnosed with rectal cancer, with preoperatively defined NLR and LMR. The overall survival rate, along with clinical and histopathological data, was evaluated. Overall survival was significantly associated with increased NLR (*p* = 0.03) and decreased LMR (*p* = 0.04), with cut-off values of 3.11 and 3.39, respectively. The two parameters were inversely correlated (*p* < 0.0001). There was no statistically significant association between tumor stage and NLR or LMR (*p* = 0.30, *p* = 0.06, respectively). The total mesorectal excision was especially obtained in cases with low NLR (*p* = 0.0005) and high LMR (*p* = 0.0009) values. A significant association was also seen between preoperative chemoradiotherapy and high NLR (*p* = 0.0001) and low LMR (*p* = 0.0001). In patients with rectal cancer, the preoperative values of NLR and LMR can be used as independent prognostic parameters. An NLR value of ≥3.11 can be used to indicate the response to preoperative chemoradiotherapy, but a low chance of sphincter preservation or obtaining a complete TME. Higher values of NLR and lower values of LMR require a more attentive preoperative evaluation of the mesorectum.

**Keywords:** neutrophil-to-lymphocyte ratio; lymphocyte-to-monocyte ratio; prognosis; rectal cancer; mesorectum; sphincter preserving

#### **1. Introduction**

Colorectal cancer (CRC) is responsible for about 10% of all diagnosed malignant tumors and cancer-related deaths worldwide [1]. Regarding its diagnosis between genders, it is the third most common cancer in men and the second most frequent in women [1]. Approximately one-third of CRC cases are diagnosed within the rectum [2].

In 1863, Rudolf Virchow first described a possible association between malignant tumors and inflammation, highlighting the role of the density of white blood cells in carcinoma behavior [3–6]. Recently, multiple studies have investigated the role of the systemic inflammatory response (SIR) in carcinogenesis, progression, and prognosis of different cancer types [6–8], but the results are controversial. The SIR is defined by several parameters, including the neutrophil-to-lymphocyte ratio (NLR) and lymphocyte-to-monocyte ratio (LMR). It is thought that NLR can predict prognosis, due to its close relationship with the cancer stage [3,7,9,10]. Increased preoperative NLR is caused by neutrophilia and/or lymphopenia, the two conditions of a pro-tumor inflammatory process. This value is, however, questioned because an elevated number of neutrophils indicates an acute SIR, whereas, cancers cause chronic SIR [5,11].

Chronic SIR is thought to be estimated by the LMR value [5]. In rectal cancer, in contrast to previously reported findings with NLR, it was observed that the preoperative LMR values were lower in patients with large tumors diagnosed in late stages [12]. Preoperative serum parameters are important for establishing treatment strategies [5]. Large cohorts are needed to establish the reliability of these cheap and easily quantified markers.

The present study aimed to validate the possible prognostic or predictive impact of preoperative NLR and LMR in a large cohort of patients with rectal cancer. The included patients underwent surgery in two university surgical departments, one from Romania and one from Hungary.

#### **2. Materials and Methods**

#### *2.1. Patient Selection*

The approval of the three Ethical Committees (Ethical Committee of the Clinical County Emergency Hospital and the Ethical Committee of the George Emil Palade University of Medicine, Pharmacy, Sciences and Technology, Targu Mures, Romania; Institutional Research Ethics Committee from Budapest, Hungary) was obtained for this study.

We performed a retrospective observational study that included all consecutive patients with rectal cancer who underwent surgery, between January 2013 and August 2019, in two university hospitals: The National Institute of Oncology of Budapest, Hungary, and the Emergency Clinical County Hospital of Targu Mures, Romania.

Besides preoperative serum values of NLR and LMR ratio, the following clinicopathological parameters were examined: Patient's gender and age, the type of surgical procedure (with or without sphincter preservation), presence or absence of preoperative chemoradiotherapy (CRT), and tumor location (low vs. mid/upper rectum), along with the pTNM stage and overall survival rate (OS) rate. Patient follow-up ranged from 1 to 76 months.

Most of the patients received capecitabine/long-course radiation therapy or 5-fluorouracil (5-FU)/long-course radiation therapy (50 Gy in 28 fractions). Blood analyses were done one day before the surgical intervention or early in the morning on the day of surgery. To allow patient recuperation, an interval of 6–8 weeks passed between CRT and surgery.

The surgical approaches were abdominal-perineal excision of the rectum (APER), anterior resection (Dixon), and Hartmann's resection. These interventions were performed by two highly experienced surgical teams in the two clinical centers.

The pathologists evaluated the macroscopical quality of total mesorectal excision (TME), which was scored as complete, partially complete, and incomplete. They also performed the microscopic evaluation of the depth of infiltration (pT stage), the quality of resection margins, the lymph node status (pN stage), and lymph node ratio (LNR) and established the pTNM stage [4].

Our study included patients over 18 years old, diagnosed with rectal cancer, with or without CRT, with preoperatively defined NLR and LMR, who underwent laparoscopic or open surgery. Exclusion criteria included biopsies, cases where death occurred less than one month postoperatively, patients with associated sepsis, autoimmune or hematologic diseases, and cases with incomplete available information.

#### *2.2. Statistical Analysis*

Data analysis was performed using Graph Pad Prism 7 and SPSS software. Nominal variables were characterized using frequencies. Quantitative variables were tested for normality of distribution using the Kolmogorov-Smirnov test and were characterized by median and percentiles (25th–75th) or by the mean and standard deviation (SD), as appropriate.

We used the Chi-squared test, Student's t-test, Mann Whitney test, and Spearman correlation test. We used the cut-off value of 3.11 for NLR and 3.39 for LMR, respectively. The cut-off values were defined according to TME quality (1 was considered complete, and 0 was considered partially complete or incomplete). A receiver-operating characteristic (ROC) curve analysis was used to test the predictive power and to determine cut-off values for NLR and LMR. We estimated the OS using the Kaplan-Meier curves; log-rank tests were applied for pair-wise comparison of survival. To distinguish non-significant cofactors from significant independent predictors of OS, multivariate Cox proportional hazards regression analyses and backward stepwise elimination were used. The Cox model was adjusted for age, gender, pTNM stage, lymph node ratio (LNR), and distal resection margin quality. All tests were two-tailed tests, and a *p*-value < 0.05 was considered statistically significant.

#### **3. Results**

#### *3.1. Clinicopathological Parameters*

We retrospectively evaluated a database of 1052 patients with rectal cancer diagnosed over six years. Included patients had a mean age of 64.29 ± 11.32 (range 21–94) years. There was a male predominance, with 61.9% males and 38.1% females (M:F ratio was 1.62:1).

In three-quarters of the patients (74.8%), the tumor involved the middle/upper rectum (which encompassed an area 5–15 cm). The ratio between middle/upper and lower (<5 cm) rectum involvement was 2.96:1. As a consequence, sphincter-preserving surgery was done in 70.5% of patients; with a ratio of 2.39:1 (preserving vs. non-preserving sphincter). The ratio between the node-negative and node-positive cases was 1.39:1. A ratio of 2.20:1 was observed between locally advanced stages (pT3-4) and cases with a low level of infiltration (pT1-2) (Table 1).

**Table 1.** Correlation between clinicopathological factors and serum indicators of systemic inflammatory response.


(NLR = neutrophil-to-lymphocyte ratio; LMR = lymphocyte-to-monocyte ratio; \* Chi square test).

The average time of the surgical interventions was 145 min. The laparoscopic surgeries required a significantly longer time (*p* < 0.0001), compared to the open interventions (162 vs. 128 min), without improving OS (*p* = 0.44) (Figure 1). However, significantly longer hospitalization was needed for patients who underwent open surgical intervention. Patients spent an average of two additional days in the hospital, compared with patients who underwent a laparoscopic intervention (*p* = 0.004). A significant difference was observed between open and laparoscopic approaches in the timing of the patient's first stool in the postoperative period (*p* < 0.0001). The first stool was present one day earlier after laparoscopic interventions than after open surgery.

**Figure 1.** Patients with rectal cancer who underwent laparoscopic surgery did not show a longer overall survival than those who were treated by open surgery (*p* = 0.44).

#### *3.2. Particularities of the Systemic Inflammatory Response*

An inverse correlation between LMR and NLR values was observed (*p* < 0.0001). This correlation was also highlighted when we compared the LMR and NLR values in males vs. females. Comparing with females, male patients showed slightly higher NLR and lower LMR values, independently by the patient's age, especially for the cases of lower rectum who received preoperative CRT (Table 1).

As regarding NLR, independently from the depth of tumor infiltration, the NLR values were slightly lower in patients with tumors of the mid/upper rectum who did not show lymph node metastases and were treated with sphincter preserving, before receiving CRT (Table 1). Furthermore, there was no significant association between the clinical TNM tumor stage and NLR or LMR (*p* = 0.30, *p* = 0.06, respectively). LNR was not influenced by NLR (*p* = 0.13), but was associated with LMR value (*p* = 0.03).

A significant association was observed between the tumor distance from the anal verge and NLR (*p* < 0.0001) or LMR value (*p* < 0.0001), indicating that a longer distance from the anal margin results in a lower NLR and a higher LMR. The distance from the anal verge was directly associated with the tumor stage (*p* = 0.006). The lower rectum tumors presented mainly lower stages, while tumor perforation was more common when the tumor was localized at the upper level (*p* = 0.01). When CRT was initiated, the incidence of tumor perforation decreased considerably (*p* = 0.01). However, independent of tumor location (low vs. mid/upper rectum), a considerable number of cases (83.2%) from the low NLR group underwent sphincter-preserving surgery, in contrast with the high NLR group, where this rate was only 66.6% (Table 1).

LMR values were lower in patients with cancers of the lower rectum who responded to preoperative CRT and did not show lymph node metastases or deep infiltration (pT1-2N0 cases). In these patients, sphincter preservation was not frequently the therapy of choice (Table 1).

The integrity of the mesorectum following TME was significantly associated with the LMR (*p* = 0.0009), and the NLR value (*p* = 0.0005), same as with the total number of harvested lymph nodes (*p* = 0.01) and LNR (*p* = 0.04). In cases with higher NLR values, the integrity of the mesorectum was only partially complete or incomplete, reflecting an abundant tissue inflammation. In these cases, the value of the NLR did not significantly correlate with the duration of the surgery (*p* = 0.18, *r* = 0.06), although the surgery time was associated with the TME quality.

In those cases that required a shorter time of surgery, the TME was more frequently complete (*p* = 0.01), in contrast with difficult cases, which demanded a longer time. Obviously, TME quality was significantly associated with the tumor invasion of the circumferential resection margin invasion (*p* < 0.0001).

The above-mentioned correlations and associations showed that an NLR value of ≥3.11 can be used to indicate the response to preoperative CRT, but a low chance of sphincter preservation or obtaining a complete TME. Based on the same algorithm, an LMR value of ≥3.39 might indicate deep invasion or absence of preoperative CRT (Table 1).

#### *3.3. Overall Survival*

The Kaplan-Meier analysis showed a significant influence of neoadjuvant treatment on patients' OS (*p* = 0.0001) (Figure 2). OS was found to be significantly influenced by SIR, defined by the cut-off values for NLR (*p* = 0.03) (Figure 3) and LMR (*p* = 0.04) (Figure 4).

**Figure 2.** Neoadjuvant treatment significantly influences patients' survival rate, compared to those who did not receive preoperative oncotherapy (*p* = 0.0001).

**Figure 3.** The cut-off value of 3.11 for the NLR can be used as an independent prognostic parameter for patients with rectal cancer (*p* = 0.03).

**Figure 4.** The cut-off value of 3.39 for the LMR can be used as an independent prognostic parameter for patients with rectal cancer (*p* = 0.04).

≥

Other parameters which significantly influenced patients' survival rate were age group (*p* = 0.04) and gender (*p* = 0.007, male patients live longer), TNM stage (*p* < 0.0001), LNR (*p* < 0.0001), positivity of the distal resection margin (*p* < 0.0001), and tumor perforation (*p* < 0.0001). There was no statistically significant correlation between OS and tumor distance from the anal margin (*p* = 0.52).

#### **4. Discussion**

The immune response and SIR influence the rate of tumor growth and the risk of metastasis [3]. Strong tumor infiltration by inflammatory cells (including neutrophils) may contribute to intensified proliferation and tumor angiogenesis [4,9,12–14]. In these cases, the response to CRT may be altered [4].

As defined, the NLR is the ratio between the absolute number of neutrophils and the absolute number of lymphocytes [4,15]. Furthermore, LMR represents the absolute number of lymphocytes divided by the absolute number of monocytes [11]. Regarding NLR, an antitumor immunity suppressing role is attributed to neutrophils [15]. This suppression contributes to cancer progression, which is augmented by lymphocytopenia [3,14]. In contrast, high levels of tumor-infiltrating lymphocytes (TIL) are correlated with a longer OS [3,12]. This is attributed to cytotoxic activity and anti-angiogenetic cytokine production [4,11].

Although the investigated parameters (NLR, LMR) might be associated with the tumor stage, the literature data are controversial, and the underlying mechanisms of these results are not fully elucidated. For this reason, these ratios cannot yet consider independent prognostic factors [11,16]. They are influenced by several factors, which should be considered. Moreover, due to the behavior, anatomical topography, and therapeutic management of rectal cancer, the rate of inflammatory markers could have different relevance [13,17]. For example, an increased LMR might be induced by autoimmune or hematologic diseases, but also by infections [9], aspects that can explain the higher rate of incomplete TME in the cases with LMR ≥3.39.

The NLR and LMR values might also be influenced by preoperative CRT. Abe et al. demonstrated lower values for LMR in patients with tumors diagnosed in pT1-2N0 stages, in patients with rectal cancer who did not receive preoperative CRT, without correlation of LMR with tumor stage, after CRT [5]. Other authors, such as Caputo et al., have demonstrated no correlation with tumor stage [4], with a negative impact on OS [8,18] or, contrary, no impact of LMR on OS or disease-free survival [19]. Our apparently contradictory results, which proved at the limit of statistical significance (*p* = 0.05), lower LMR values for pT1-2N0 cases, are in line with data reported by authors, such as Mallapa et al. [16] or Abe et al. [5]. We did not find an association of LMR with LNR, which was defined as the number of positive lymph nodes divided by the total number of lymph nodes harvested [20,21], and not reported to the rectal lymph nodes only.

The controversial literature reports might be explained based on the number of examined cases (usually below 300), the used protocols therapeutically and the homogeneity of the cohort. As most of the authors perform statistical examinations in consecutive cases and establish "in-house" managed ratios, using receiver operating curve analysis [8], such in our material, controversial data might be influenced by these above-mentioned factors. It is necessary to mention, for example, that 68.8% of patients included in the present study were diagnosed in patients with pT3/4 staged tumors, with or without lymph node metastases, and correlations were done based on the cut-off value of 3.39. In contrast with other authors [7,16], we have included in the examined databases only rectal carcinomas, without tumors located in the colon or in the anal canal. On the other hand, some authors, such as Xiao et al. included only patients in the pT3N0 stage [18].

In patients with CRC, poorer OS was predicted when pre-treatment NLR showed elevated values [22]. In our material, the applied cut-off value for NLR was 3.11. An NLR of >3, before CRT/before surgery, is usually considered an indicator of poor prognosis, high recurrence rate, and low 5-year OS [4,15,23]. As regarding tumor stage, Caputo et al. reported, similar to our data, no significant association between the pTNM stage and NLR [4]. A higher clinical stage III rate was observed by Cha et al. in their cases within high NLR [11]. Other authors reported increased NLR in cases with positive

nodal status on MRI [13,14]—this latter aspect being also proved by our data. Based on an in-house evaluation of large cohorts, a cut-off value should be established for prognostic evaluation [24,25]. In our material, NLR ≥3.11 and LMR <3.39 proved to be indicators of the response to preoperative CRT and lower risk for lymph node metastases.

Regarding TME, its quality was more frequent incomplete or only partially complete in cases with high NLR values. This aspect can be related to the surgical intervention quality or can reflect a peritumoral extensive inflammatory profile, which does not allow a proper and complete mesorectal excision. Scarce data can be found in the literature about the possible relationship between NLR and TME quality. Authors, such as Sung et al. [26] or Portal et al. [10], agree with the idea that pre-CRT and post-CRT NLR value might be used as a blood biomarker with a potential prognostic role in the case of patients who underwent curative TME intervention [10,26], but we did not find other supplementary explanation for this relationship. However, a high NLR and a low LMR value will lead in the majority of the cases to an incomplete TME quality, which will negatively influence patient's OS, disease-free survival, and recurrence survival. In these circumstances, surgeons can identify preoperatively patients with a high risk of developing postoperative complications [4,27,28].

The strength of the present study consists of the investigation of one of the largest sample sizes so far, including only rectal cancer cases, from two medical centers. The limitations consist of its retrospective manner, and of including only two surgical centers, with predominantly locally advanced cases. These limitations were mitigated by using the largest databases reported in the literature, establishing an in-house designed value for both NLR and LMR. To elucidate the possible geographic-related differences, also reported for other cancers [29], a further multicenter study, in a larger cohort, should be performed.

#### **5. Conclusions**

The investigated parameters, NLR and LMR, are useful independent prognostic parameters. An NLR value of ≥3.11 can be used to indicate the response to preoperative CRT, but a low chance of sphincter preservation or obtaining a complete TME. An LMR value of ≥3.39 might indicate deep tumor invasion. The fact that NLR and LMR seem to influence the TME integrity is of great importance, which should be considered by rectal surgeons.

**Author Contributions:** Conceptualization, S.G. and Z.Z.F.; data curation, Z.Z.F. and J.T.; formal analysis, Z.Z.F. and R.L.F.; investigation, R.L.F., T.B.J. and E.D.; methodology, R.L.F., T.B.J. and I.J.; project administration, S.G.; resources, J.T.; software, E.D.; supervision, I.J.; validation, E.D.; visualization, E.D.; Writing—review & editing, J.T. and I.J.; Z.Z.F. and R.L.F. have equal contribution to the paper. All authors have read and agreed to the published version of the manuscript.

**Funding:** This research was funded by the Collegium Talentum 2019 Program of Hungary (research bursary for first author) and a grant of the Romanian National Authority for Scientific Research, CNCS—UEFISCDI, project number 20 PCCF/2018, code: PN-III-P4-ID-PCCF-2016-0006. The APC was funded by GE Palade University of Medicine, Pharmacy, Sciences and Technology, Targu-Mures, Romania.

**Acknowledgments:** The authors would like to express their sincere gratitude for the Hungarian team from the National Institute of Oncology, namely Mersich Tamás, Head of the Visceral Surgery Department, Gödény Mária, Head of the Radiology Department and Strausz Tamás from the Pathology Department. We also thank to our colleagues from Clinical County Emergency Hospital of Targu-Mues, Romania, for their help in treating patients and performing their follow-up and to Septimiu Voidăzan for statistical assessment.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

© 2020 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## *Review* **Predictive Biomarkers of Oxaliplatin-Induced Peripheral Neurotoxicity**

**Roser Velasco 1,2, \* , Montserrat Alemany 1 , Macarena Villagrán <sup>1</sup> and Andreas A. Argyriou 3**


**Abstract:** Oxaliplatin (OXA) is a platinum compound primarily used in the treatment of gastrointestinal cancer. OXA-induced peripheral neurotoxicity (OXAIPN) is the major non-hematological dose-limiting toxicity of OXA-based chemotherapy and includes acute transient neurotoxic effects that appear soon after OXA infusion, and chronic non-length dependent sensory neuronopathy symmetrically affecting both upper and lower limbs in a stocking-and-glove distribution. No effective strategy has been established to reverse or treat OXAIPN. Thus, it is necessary to early predict the occurrence of OXAIPN during treatment and possibly modify the OXA-based regimen in patients at high risk as an early diagnosis and intervention may slow down neuropathy progression. However, identifying which patients are more likely to develop OXAIPN is clinically challenging. Several objective and measurable early biomarkers for OXAIPN prediction have been described in recent years, becoming useful for informing clinical decisions about treatment. The purpose of this review is to critically review data on currently available or promising predictors of OXAIPN. Neurological monitoring, according to predictive factors for increased risk of OXAIPN, would allow clinicians to personalize treatment, by monitoring at-risk patients more closely and guide clinicians towards better counseling of patients about neurotoxicity effects of OXA.

**Keywords:** neurotoxicity; oxaliplatin; chemotherapy-induced peripheral neuropathy; biomarker; genomics; neuropathy; FOLFOX; FOLFIRINOX; XELOX; gastrointestinal cancer

#### **1. Introduction**

Oxaliplatin (OXA) is widely used for the treatment of gastrointestinal cancers including colorectal (CRC), gastric, and pancreatic cancer, both in the adjuvant and metastatic setting [1,2]. OXA-induced peripheral neurotoxicity (OXAIPN) is the major non-hematological cause for dose-reduction as also discontinuation of OXA-based chemotherapy and it is manifested with two clinically distinct forms. The acute, neuromyotonia-like syndrome, as a result of hyperexcited sensory and motor nerves, appears soon after OXA, is transient and usually completely reversible within hours or days [3]. Patients may also develop chronic sensory symmetrical symptoms, including tingling, numbness and pain in a 'stocking/glove' distribution developing during treatment, while up to 20% of patients can be severely affected to develop sensory ataxia and increased susceptibility to falls [4–6]. Five years after finishing chemotherapy, 25–30% of patients suffer from clinically significant chronic OXAIPN [3,7], without modification in this rate over 3–8 years [8–11]. Persistent OXAIPN is associated with psychological distress, depression and impaired quality of life in long-term gastrointestinal cancer survivors [8].

**Citation:** Velasco, R.; Alemany, M.; Villagrán, M.; Argyriou, A.A. Predictive Biomarkers of Oxaliplatin-Induced Peripheral Neurotoxicity. *J. Pers. Med.* **2021**, *11*, 669. https://doi.org/10.3390/ jpm11070669

Academic Editors: Enrico Mini and Stefania Nobili

Received: 7 June 2021 Accepted: 13 July 2021 Published: 16 July 2021

**Publisher's Note:** MDPI stays neutral with regard to jurisdictional claims in published maps and institutional affiliations.

**Copyright:** © 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https:// creativecommons.org/licenses/by/ 4.0/).

Given the lack of effective symptomatic or preventive treatment strategies against both acute and chronic OXAIPN [9], in daily practice, neurological symptoms referred by patients are usually taken into account to adapt OXA dosing in order to prevent severe neuropathy (Figure 1). Treating physician will indicate dose adjustment of OXA administering significantly less cumulative than planned doses of anti-cancer treatment what may compromise patient survival [10], therefore becoming a critical decision balancing the maximization of therapeutic benefit and the minimization of this significant OXA toxicity. Importantly, the observed worsening of OXAIPN weeks after cessation of treatment complicates clinical decisions regarding the duration and total dose of OXA in individual patients based on symptoms of sensory neuropathy alone.

**Figure 1.** Personalized assessment of neuropathy by present and future strategies to early detecting and monitoring oxaliplatin-induced peripheral neuropathy in gastrointestinal cancer patients.

≤ Cumulative dose of OXA is the main predictor of OXAIPN [12–14], with increasing rates from 42.5% to 76.7% with median cumulative dose of ≤780 mg/m<sup>2</sup> and >780 mg/m<sup>2</sup> , respectively [11], also when re-challenging patients with OXA in further lines [15,16]. However, the relationship between dose and neurotoxicity might not be linear before reaching a cumulative dose level beyond which the toxicity becomes dose-dependent [17]. OXAIPN development cannot be accurately predicted in a gastrointestinal cancer patient before OXA treatment initiation and the common inter-individual variability in severity of OXAIPN in the setting of a uniform insult is a major challenge in clinical practice. Early prediction of development and progression of OXAIPN and a timely decision to decrease the OXA dosing in patients at high risk is clinically important. In recent years, several clinical and neurophysiological predictive biomarkers that can be easily obtained before or early during treatment to estimate which patients are at higher risk for OXAIPN have been described. From the personalized medicine perspective, having non-invasive, sensitive and specific biomarkers, will allow patients more liable to OXAIPN to prevent the occurrence of long-term toxicity or permanent damage. These objective markers may aid in the prediction of the development, severity and duration OXAIPN, and in adjusting OXA dose more precisely to balance the risk of neurotoxicity against antineoplastic efficacy. This review focuses on the currently available biomarkers of early OXAIPN detection that may allow clinicians to closely monitor at-risk patients and personalize treatment according to neurotoxicity risk of OXA.

#### **2. Clinical Factors Associated with OXAIPN**

#### *2.1. Pre-Treatment Factors*

A variety of pretreatment patient- or environmental-related risks have been described in the literature with conflicting results (Table 1). Methodological issues, such as the relatively small size, the retrospective design and the lack of statistical approach with multivariate regression analysis in many of them, likely underly limitations in the generalization of available results. To date, no demographic factors or preexisting comorbidities including age, diabetes mellitus or preexisting neuropathy have been consistently identified across multiple studies helpful in the prediction of the development, severity or duration of OXAIPN [18] that should be considered for upfront screening in a priori patients' risk classification.

**Table 1.** Factors investigated that are or not associated with the incidence and severity of oxaliplatin induced peripheral neurotoxicity.


\* Studies including multivariate analysis. # Shorter time to develop OXAIPN & No differences in the incidence of OXAIPN.

#### *2.2. Acute Neurotoxicity: The Main Clinical Predictor of OXAIPN*

Shortly after every OXA infusion, many patients can experience cold-triggered painful paresthesia or neuromyotonia syndrome related with a transient axonal hyperexcitability of the peripheral nerve secondary to oxalate, generated after OXA biotransformation into its active form [3]. These acute symptoms are experienced by most of patients (90%) at some point of time during treatment [18,20,22], being usually reversible within hours or days and typically is triggered or exacerbated by cold [18,21,23,47] (Table 2). Similar to persistent OXAIPN, no factors beyond dose or infusion time seem related with the risk of developing acute OXAIPN [10,21,23–28].

**Table 2.** Acute symptoms or signs of OXA induced neurotoxicity.

Acute neurotoxicity is a well-established risk factor for developing chronic OXAIPN at the end of chemotherapy. Clinically, several studies show that the onset and severity of acute OXAIPN is associated with the occurrence of persistent neurotoxicity [4,10,20,24,29,30]. Attal et al. showed that the duration of cold-evoked pain and intensity of neuropathic symptoms experienced during the first three cycles predicted the extent of chronic pain experienced one year later. Cold-evoked symptoms lasting four days or more after 3rd OXA cycle predicted chronic OXAIPN (OR: 22; 95% CI: 1.54–314.74; *p* = 0.02) in a comprehensive comparative prospective study including 28 cancer patients receiving OXA, mostly CRC [20]. Our group identified that the burden of acute symptoms measured when patients have completed half of the planned OXA-based treatment (mid-treatment) was independently associated with nearly double risk (OR 1.9; CI 95% 1.2 to 3.2; *p* = 0.012) of developing severe chronic OXAIPN in a prospective multicenter study including 200 CRC patients [4,24]. In this line, presence of any acute neuropathy during cycles 1–3 was associated with persistent OXAIPN (HR: 3.65 (CI 95%; 1.40–9.56) *p* = 0.008) in a retrospective analysis of 50 CRC receiving FOLFOX schedule [12].

Early onset of acute OXAIPN seems particularly predictive of long-term neurotoxicity. Pachman et al. identified that patients who experienced severe acute OXAIPN within 6 days after first OXA infusion experienced more severe neuropathy in the remaining cycles and increased incidence of chronic OXAIPN (*p* < 0.001) [48]. Particularly, hyperacute neuropathy on the first day of the first OXA cycle was found to be a hallmark of risk of OXAIPN. Up to one third (27.7%) of patients developed hyperacute neuropathy in a retrospective study including 47 CRC patients receiving OXA-based chemotherapy. Of 13 patients who experienced hyperacute neuropathy, 12 (92.3%) eventually developed persistent OXAIPN. Multivariate analysis including the total dose of OXA and the presence of hyperacute neuropathy demonstrated that these two variables independently predicted OXAIPN [49]. The role of a such early clinical symptoms as further predictors require further validation in a in large multicenter study.

Diverse strategies have been employed for assessing acute OXA neurotoxicity. The Common Terminology Criteria for Adverse Events from the National Cancer Institute (NCI.CTC) [10,44,50], the oxaliplatin-specific neurotoxicity scale [51], or a score based on recording the frequency of symptoms with an OXA-Neuropathy Questionnaire (yes/no response format) [17,52,53], are among the most common systems for recording their presence in the daily practice. The severity of acute neurotoxicity syndrome has been defined according the burden of symptoms [54], or according to a visual analogical scale 0 (no problem) to 10 (major problem) numerical rating scale for any of the four acute neuropathy symptoms [48,55,56]. More sophisticated techniques to objectively assess acute syndrome are described below in detail.

Importantly, gastrointestinal cancer patients receiving OXA must be specifically interviewed about the presence of typical and atypical neuropathic symptoms, either these are transient or persistent. For example, recognition of acute atypical forms of OXAIPN requires a prolonged OXA infusion rate from 2 to 4 or 6 h in order to reduce risk of persistent

OXAIPN [3,57]. The implementation of a simple standardized assessment tool to monitor acute neurotoxicity in daily clinical practice should be considered due to large amount of evidence supporting the predictive role of these early manifestations in predicting persistent OXAIPN.

#### **3. Neurophysiological and Device-Dependent Predictors**

#### *3.1. Nerve Conduction Studies (NCS) and Electromyography*

Several longitudinal studies, including NCS during OXA therapy, have showed a significant progressive decrease in sensory nerve action potentials (SNAPs) and preservation of motor action compound (CMAPs), in keeping with the presence of an axonal sensory neuropathy, and consistent with the clinical symptoms and signs worsening during the treatment [17,41,58–64]. NCS are capable of objectively assessing the extent of peripheral nerve damage and may also facilitate the identification of patients that manifest subclinical peripheral neuropathy prior to the onset of clinically significant neurotoxicity. One cross-sectional study including 17 patients that had received a median of seven [8,45,65–68] treatment cycles and 850 mg/m<sup>2</sup> at the time of testing, disclosed that almost half of patients had evidence of sensory neuropathy. After sensory examination, reductions in upper and lower limb SNAPs of patients were the most sensitive early markers of neuropathy observed in 40% [69]. In this line, in one prospective study including 60 gastrointestinal cancer patients, sural nerve velocity and SNAP revealed a significant decrease after 50% and 100% of the planned dose, respectively [60].

NCS have also shown being useful in early predicting the neurological outcome at OXA completion. Early changes in the NCS results obtained during treatment were able to predict the development of severe OXAIPN in several prospective studies. A large multicenter study, including 200 CRC patients under treatment with FOLFOX-4, 6, and XELOX, identified at mid-treatment compared to baseline values a >30 % decrease in radial and dorsal sural SNAPs, while these abnormalities yielded a sensitivity and specificity of 96.3% and 79.1%, respectively, with positive and negative predictive values of 53% and 98.9%, for predicting severe OXAIPN at treatment completion. In the multivariate analysis, the three factors obtained at mid-treatment to independently and significantly be associated with an increased risk of severe neuropathy were: (1) having shown more symptoms of acute neurotoxicity (2) having a drop in the amplitude of the SNAP of the dorsal sural and radial nerves greater than 30%. The combination of these three factors allowed the patient with a high negative predictive value close to 99% to be classified a priori, so that in those patients with optimal NCS in the middle of treatment and who have not developed many symptoms of acute neuropathy, we could ensure with a high probability that he/she will not develop severe OXAIPN [17]. The predictiveness of dorsal sural nerve in risk stratification for OXAIPN was further evaluated in a secondary analysis of 100 CRC patients. An algorithm based on the dorsal sural nerve recordings showed that mid-treatment NCS could assign each patient to a 'neurophysiological risk class' for OXAIPN at the end of treatment [70]. In this line, reductions of the SNAPs of >11.5% in the median nerve between baseline and four cycles of OXA (odds ratio = 5.603, *p* = 0.031) and of >22.5% in the sural nerve between four and eight cycles of chemotherapy (odds ratio = 5.603, *p* = 0.031) were independently associated with the risk of developing severe OXAIPN [63]. However, very recently, negative results were obtained in another study evaluating the role of the sural nerve after administering 25% or 50% of the planned OXA dose in predicting the occurrence of clinically significant OXAIPN in 55 CRC patients [59]. The assessment of sural nerve instead dorsal sural (Figure 2), and the size of the study could underlie these negative findings. Long-term longitudinal neurophysiological assessments of OXA-treated patients have revealed a significant recovery of the SNAPs in sensory nerves in some studies [58,63,71] but not in other studies [59,61]. The length of follow-up observation may explain these differences. Unfortunately, to date, the correlation between SNAPs impairment and degree of neurotoxicity recovery remains unknown. In summary,

growing evidence supports that NCS in distal sensory nerve segments offers clinicians a practical means of identifying patients more prone to severe chronic OXAIPN.

**Figure 2.** Nerve conduction study of the dorsal sural nerve.

Muscle sampling with needle electromyography (nEMG) can show repetitive myokymic discharges and neuromyotonic runs within 1–4 days after the first OXA administration [72–74]. However, the invasive nature of nEMG hampers its feasibility as screening tool to monitor gastrointestinal cancer patients undergoing OXA. Very recently, a simple painless objective tool to detect nerve hyperexcitability acute syndrome by a surface electromyography (sEMG) has been tested in a small study including CRC patients after the second (*n* = 14) and fourth (*n* = 8) OXA infusions revealing that sEMG is more sensitive (82%) than neurological examination (55%) to detect objective signs of acute neurotoxicity [55]. As such sEMG might be a promising test to evaluate acute oxaliplatin-induced motor nerve hyperexcitability and warrants to be further investigated in future studies.

#### *3.2. Quantitative Sensory Tests (QST)*

Quantitative sensory testing (QST) examines subjective sensory function by measuring the abnormal detection and pain thresholds to several sensory modalities. The usefulness of these non-invasive psychophysical measurements in the clinic setting for detecting subclinical neurologic changes early on to identify patients that will experience OXAIPN has been largely explored [12,20,20,42,51,53,54,56,75–82].

Vibration sensation testing can be performed by a computer controlled vibrometer or, more easily, with a tuning fork placed on a bony prominence, such as the hallux or malleollus. In both, the subject reports when they can no longer detect vibration. Impairment of the vibration detection threshold (VDT) is generally seen over the treatment course [13,20,58,61] in correlation with the progressive loss of large myelinated fibers. However, conflicting results regarding the predictive role of early changes in VDT are seen in the literature. Whereas VDT in 30 patients with gastrointestinal malignancies receiving OXA evaluated at baseline and during infusion cycles 1, 2, 4, and 6, showed no clear relationship with OXAIPN development [80], two studies including 17 [69] and 60 patients [60] showed being the earliest marker of neuropathy, present at low cumulative doses of OXA. Significant change in VDT were present after the 25% of the planed dose, being the earliest among other measures [60]. Very recently, Kroigard et al. identified VDT measure obtained before treatment correlated as a predictor of clinically significant OXAIPN six months post-treatment. However, sensitivity of a baseline VDT < 5 (maximum 8) for the prediction of clinically significant neuropathy six months after treatment was modest (76.0% (95% CI 54.9% to 90.6%)) and specificity was low (53.3% (95% CI 34.3% to 71.7%)) [59].

Mechanical detection thresholds (MDT) measures have also shown significant deterioration with increasing OXA doses in some studies [42,60] but not in others [20,82]. Touch threshold changes became statistically significant in the fingertips at middle and 8 chemotherapy doses [42] whereas these findings only occurred after treatment completion in Kroigard's study [60]. Very recently, electronic von Frey anaesthesiometer, which evaluates hyperalgesia based upon mechanical pain thresholds (MDTs), was tested prospectively in 46 CRC patients treated with OXA, and showed a decrease of 40 g in the MDT of both hands and feet as cut-off for diagnosing grade 2 or 3 OXAIPN with a total accuracy of

84.2% and 81.6% in hands and feet, respectively [78]. Besides diagnostic utility, the role MDT changes for predicting OXAIPN requires further research.

Thermal detection thresholds have been particularly investigated due to the coldinduced nature of early acute OXAIPN [79]. Attal et al. was able to detect sustained signs of neurotoxicity at an early stage when the clinical manifestations appeared to revert between OXA cycles in a comparative prospective study including 48, mainly CRC patients with who were evaluated before OXA (*n* = 28) or cisplatin (*n* = 20) and after cycles 3, 6 and 9 and after completion. Enhanced pain in response to cold (20 ◦C stimulus on the hand) predicted severe neuropathy (OR: 39; 95% CI: 1.8–817.8 *p* = 0.02) [20]. Early changes in cold (CDT) and heat detection thresholds (HDT), as predictors of clinically significant neuropathic pain six months after treatment, has been recently identified by Kroigard et al. in a prospective study including 55 patients, 14 out of them with neuropathic pain. Reduced CDT after 25% of the planned OXA dose and reduced HPT after 50% of the planned dose measured at the dorsum of the right foot was correlated with neuropathic pain intensities. Change of −0.05 ◦C in CDT had a sensitivity and specificity of 92.3% (95% CI 64.0% to 99.8%) 64.9% (95% CI 47.5% to 79.8%), respectively, in predicting neuropathic pain six months after finishing OXAIPN. For change in HPT and the prediction of neuropathic pain, −0.85 ◦C had a sensitivity of 64.3% (95% CI 35.1–87.2%) and a specificity of 70.0% (95% CI 53.5–83.4%) [59]. Conversely, no association in cold and warm thresholds in 35 cancer patients treated with OXA-based regimen and OXAIPN was identified [82].

The role of QST to early identify OXAIPN remains vaguely defined. Among limitations, technically challenging methods of QST are not widely available, are time-consuming, and need standardized assessment algorithms and normative data which are not universally defined [83,84]. Besides, QST requires patient's collaboration because results are based on a subjective response of the patient, compromising the objectivity desirable in a biomarker, and make QST not applicable in a subset of patients with impaired cognition and attention. Among QST parameters, VPT has the advantage of being quickly performed. Additionally, the equipment required is very portable and requires only basic training to operate. Accordingly, our and other authors experience [69] would favor VDT as the simplest and best routine marker among QST for detecting early OXAIPN in the clinic setting.

Other devices to quantify tactile sensation (Bumps Detection test) [42,81] or small fiber (Sudoscan) [85] have been investigated for early diagnosis of large or small fiber impairments in subjects suspected of having OXAIPN, and for monitoring change over time, with promising albeit very preliminary results. Baseline deficits in sensory functioning measured using the Bumps Detection test were predictive of increasing numbness/tingling during the first 26 weeks of OXA-based chemotherapy [81]. Very recently, a multicenter study including 101 patients evaluated the usefulness of the CLIP test for early prediction of the risk of progression ≥grade 2 neuropathy in patients receiving chemotherapy with OXA. By testing the difficulty of patients in picking up and moving five gem clips one by one two squares and assessing patients experience in performing the test, authors identified that a positive result on the CLIP test (by asking patient to pick up and move a gem clip and whether there was some wrongness in doing it) was associated with an 8.3-fold higher risk of progression to ≥grade 2 OXAIPN. Noteworthy, a positive conversion of the CLIP test occurred before the progression to ≥grade 2 OXAIPN in 14 of the 21 (67%) patients [86]. The usefulness of this simple, cheap, and widely available assessment tool should be further validated in larger, multicenter prospective comparative studies.

#### *3.3. Axonal Excitability and Skin Biopsy*

OXA produces acute changes in peripheral nerve excitability by modulating axonal voltage-gated Na<sup>+</sup> channel activity [74,87]. Nerve excitability studies evaluate axonal excitability to provide information of the properties of the nerve membrane and of the ion channels expressed on these axons [88]. Acute symptoms after OXA infusion correlate with nerve excitability findings [56]. This method can assess acute OXA-induced abnormalities in sensory or motor nerve function [56], and its cold-triggered aggravation [89].

Measurement of excitability parameters have been consistently shown to be a sensitive early biomarker of ongoing OXAIPN, even preceding the reduction in the SNAP and development of symptoms [76,90,91]. Patients who demonstrated changes in excitability in early treatment, shortly after infusion, were subsequently more likely to develop moderate to severe neurotoxicity. Park et al. reported that an increase in the superexcitability of more than 15% prior to 5 of 12 chemotherapy cycles was able to identify 80% of patients with moderate or severe chronic OXAIPN. Acute changes in axonal excitability parameters that developed in early treatment cycles anticipated development of later neurotoxicity in patients who completed seven or more treatment cycles. Patients who completed treatment with moderate to severe neurotoxicity showed greater changes in early treatment (cycles 1 or 2), particularly reductions in the associated hyperexcitability [76]. Despite OXA causes acute excitability changes in both motor and sensory axons, progressive cumulative changes were only found in sensory nerves, and motor nerve excitability studies did not reveal early cumulative changes following treatment with OXA [69,76].

Nerve excitability testing provides a sodium channel dysfunction index and an objective biomarker of acute OXA neurotoxicity useful to improve prediction and risk stratification for OXAIPN prior to the onset of chronic neuropathy [56]. However, their scarce availability in most of centers, time-consuming nature and the lack of standardization for the clinical testing [88], converts this technique in too complex for routine screening, and not applicable in daily clinical practice for early detection of OXAIPN.

Skin biopsy is a minimally invasive method to evaluate neuropathy, especially small fiber nerve damage. Five prospective studies have incorporated skin biopsy in assessing ongoing OXAIPN. Contradictory results on the change over time of intraepidermal nerve fiber (IEFN) are available [59,60,64,92,93]. No significant early reduction in IENF during OXA treatment has been demonstrated; evidence which could be related with the fact that loss of IENF, a marker of axonal degeneration, is usually a later event occurring in peripheral nerves [60,84]. Besides, ongoing regeneration of small nerve fibers during OXA could contribute to these discrepancies [59]. Therefore, skin biopsy should not be used for predicting OXAIPN.

#### **4. Pharmacogenomic Biomarkers**

Genetic factors may contribute to a patient's risk of experiencing OXAIPN. Over the last years, the development of pharmacogenetics, used to characterize human genetic variation, facilitated extensive efforts to understand the genetic basis of OXAIPN and to identify a specific genetic profile that can identify patients who are more liable to severe chronic neurotoxicity at the end of treatment. The majority of published studies assessed individual OXAIPN susceptibilities on single nucleotide polymorphisms (SNPs), which are mainly associated with gene variations in detoxification enzymes; DNA repair; drug transport; metabolism; neuronal receptors and ion channels (Figure 3). Furthermore, other genome wide analysis studies (GWAS) attempted to identify and validate SNPs mainly in genes encoding proteins implicated to neuronal function [94,95].

**Figure 3.** Cellular processes and main candidate genes implicated in oxaliplatin-induced neurotoxicity.

Some of these studies are of interest; tellingly, however, they have provided inconsistent findings and failed, in principle, to be replicated by other independent groups because of significant methodological limitations, including small sample sizes; retrospective study design; implementation of a post hoc analysis of oncology-based databases of different, not pre-planned size; lack of a pre-study hypothesis based on the known role of the investigated targets in the peripheral nervous system; inappropriate outcome measures for neurological impairment and differences related to DNA origin, extraction and genotyping [96].

#### *4.1. SNP Studies*

4.1.1. Glutathione-S-Transferase P1 (GSTP1), Glutathione-S-Transferase T1 (GSTT1) and Glutathione-S-Transferase M1 (GSTM1) Genotyping

Genetic variants for GSTP1 exon 5 (Ile105Val), GSTP1 exon 6 (Ala114Val), GSTM1 (homozygous deletion), and GSTT1 (homozygous deletion) were examined in a cohort of 64 OXA-treated CRC patients, among whom 15 had grade 3 chronic OXAIPN. Patients homozygous for the GSTP1 105Ile allele more frequently encountered grade 3 OXAIPN compared to patients homozygous or heterozygous for the GSTP1 105Val allele (OR: 5.75; 95% CI: 1.08–30.74; *p* = 0.02). GSTM1, GSTT1, or GSTP1 exon 6 genotypes have not been associated with severe chronic OXAIPN [97]. In another study enrolling 63 OXA-treated (mFOLFOX6) metastatic CRC patients, it was shown that GSTP1–105 (*p* = 0.03) and GSTM1 (*p* = 0.02) were associated with increased incidence of severe chronic OXAIPN [98]. Another two studies disclosed a significantly reduced risk of OXAIPN with the GSTP1 AA genotype (Ile/Ile) [99,100], while the same increased risk of OXAIPN manifestation was reported with the GG (Val/Val) genotype [101,102]. Noteworthy, there are several reports with

controversial results showing no association of GSTP1 Ile105Val with increased incidence of OXAIPN [103–105].

#### 4.1.2. ATP-Binding Cassette Transporter 2 (ABCG2) Genotyping

Custodio et al. performed genotyping in a cohort of 206 stage II-III OXA-treated CRC patients and a validation set of another 181 patients. Significant associations emerged for the CCNH rs2230641 C/C (OR: 5.03, 95% CI: 1.061–2.41, *p* = 0.042) and the ABCG2 rs3114018 A/A alleles (OR:2.67; 95% CI: 0.95–4.41; *p* = 0.059) with higher risk of grade 2–3 OXAIPN, while patients harboring the combination of these genotypes had significantly increased risk of severe OXAIPN, compared to patients carrying the CCNH any T and ABCG2 any C genotypes (37.73% vs. 19.42%; OR:2.46; 95% CI: 1.19–5.07; *p* = 0.014) [106].

However, replication of these results failed to achieve in a subsequent study enrolling 465 stage II or III CRC patients of Asian origin who were treated with the adjuvant-modified FOLFOX6 regimen. In the latter setting, comparison of low grade (0/1) OXAIPN with significant grade 2/3 OXAIPN did not showed any significant associations with any of the 12 examined SNP markers, including ABCG2 rs3114018 and CCNH rs2230641 [107]. In line with these negative results concerning the relevance of ABCG2 SNPs with OXAIPN, are the findings of a recently published study that tested germline DNA from 120 OXA-treated CRC patients together with a validation cohort of 80 patients and found no significant associations between ABCG2 (c.421 C > A/rs2231142) and increased incidence of OX-AIPN [108].

#### 4.1.3. Cyclin-H (CCNH) Genotyping

The association of CCNH rs2230641 C/C with an increased incidence of severe OX-AIPN has been demonstrated in the Custodio et al.'s study (2014), which was mentioned earlier [106]. Furthermore, the same effect of CCNH genotypes in acute OXAIPN was demonstrated in another study enrolling 228 OXA-treated digestive tract cancer patients. This study revealed that the CCNH-rs2230641 (AA vs. AG+GG; dominant model) and CCNH-rs3093816 (AA vs. AG+GG; dominant model) were both found significant for higher risk of more frequent and severe acute OXAIPN [109].

#### 4.1.4. X-ray Repair Cross-Complementing Protein 1 (XRCC) Genotyping

Genetic variants of the XRCC1 G/G polymorphism, which results in an Arg399Gln substitution has been tested for OXAIPN relevance in a study prospectively enrolling 292 Korean patients treated with FOLFOX for CRC and found that patients harboring XRCC1 23885GG experienced less grade 2–4 OXAIPN (adjusted OR:0.52, 95% CI: 0.27–0.99) [110].

Nonetheless, many other studies reported concordant evidence for lack of relevance between XRCC1 SNPs and OXAIPN, including Arg399Gln substitution [106,111]; Arg194Trp [112]; Arg280His [106,112]; rs3213239 [112]; rs12611088 and rs3213255 [106]. Subsequently published studies further provided evidence in support of the absence of any relevance of XRCC1 variants to OXAIPN features [105,107]. Tellingly, however, in the Kanai et al. study (2016), the proportions of patients developing grade 2–3 OXAIPN was quite higher compared to the Ruzzo et al.'s study (2014) (40.2% vs. 25.5).

#### 4.1.5. Voltage-Gated Sodium Channels (SCNA) Genotyping

The SCN2A R19K polymorphism failed to be associated with liability to OXAIPN in a study in which 62 advanced CRC patients were genotyped [113]. Similarly, no significant association emerged between SCN9A variant rs6746030 and OXAIPN in a subsequent study comprising 200 CRC patients [114], contrasting the results of a smaller study in which SCN9A rs6746030 was protective of severe OXAIPN in a heterogeneous population of 94 patients with various digestive tract cancers, and an increased incidence of coexisting diabetes (24%) in patients with grade 3–4 OXAIPN [115]. Tellingly however, a subsequently published study performed genotyping in 228 Indian OXA-treated digestive tract cancer patients and found increased susceptibility to chronic OXAIPN with the rs6746030 polymorphic variant of SCN9A (GA+AA vs. GG: OR: 1.8; 95% CI:1.04–3.4; *p* = 0.04; dominant model), while the SCN10A polymorphic variant was associated with severity of chronic OXAIPN (OR:2.0; 95% CI:1.2–3.3; *p* = 0.006) [11].

Finally, in the Argyriou et al. study (2013), it was disclosed a significant association between the SCN4A variant rs2302237 and increased risk of any grade chronic OXAIPN (OR:2.47; 95% CI: 1.04–5.85; *p* = 0.037) and more severe acute OXAIPN (OR: 2.50; 95% CI:1.35–4.63; *p* = 0.0029) [114].

#### 4.1.6. Voltage-Gated Potassium Channels (KCCN3) Genotyping

Basso et al. (2011) provided evidence for a significant association between 13–14 CAG repeat allele of KCNN3 (SK3) gene and development of acute OXAIPN (OR with >15 repeats; 0.381, 85% CI: 0.247–0.590, *p* = 0.001) in a small cohort of 40 CRC patients [116]. Another study, enrolling 86 CRC patients, did not show a significant association between chronic OXAIPN and KCNN3 repeat polymorphism [117]; concurring with the same negative results of a larger study studying 151 CRC patients and which provided evidence for lack of significant association between variations of the KCNN3 repeat polymorphism and development of either acute or chronic OXAIPN [113]. In any case, several mechanistic insights in OXAIPN pathogenesis are not supportive for a significant and direct involvement of K+channels in OXAIPN manifestation [118].

#### *4.2. GWAS Studies*

Won et al. (2012) performed GWAS in a discovery set of 96 and a validation cohort of 247 OXA-treated CRC patients of Asian origin in order to identify potential genetic markers for severe OXAIPN. This study identified and validated nine SNPs in eight genes [119], that failed to replicate in an independent validation in Caucasian patients [120]. The different genetic background of patients in the original and replication GWAS might hold the main response for conflicting results. No association was found with SNPs in ERCC1, GSTP1, XRCC1 and SNCA1 [119]. Finally, a very recently published study, enrolling over 1000 patients treated with paclitaxel, paclitaxel plus carboplatin, or oxaliplatin reported significant associations between rs56360211 near PDE6C (*p* = 7.92 × 10−<sup>8</sup> ) and rs113807868 near TMEM150C (*p* = 1.27 × 10−<sup>8</sup> ) with peripheral neurotoxicity. Tellingly, however, these results emerged from a polled analysis of treated patients and genetic associations were not tailored according to different chemotherapy compounds delivered [121].

There is increasing evidence pointing to the role of pharmacogenetics and pharmacogenomics in neurotoxicity susceptibility to OXA. Whereas pharmacogenetic results are currently being used in clinical decision making to inform treatment regimen choice with agents such as anthracycline [122] or fluoropyrimidine [123], larger-scale and validation studies are needed to further identify susceptibility markers of OXAIPN and to develop pharmacogenomics based-risk profiling to improve quality of life of gastrointestinal cancer patients.

#### **5. Imaging Biomarkers**

Data on imaging biomarkers in OXA-treated patients are still very limited, despite the benefit of their non-invasive nature. Nerve size, estimated by cross-sectional area (CSA), measured by nerve high-resolution ultrasound (HRUS), is an imaging modality that allows a quantitative structural analysis of the nerves. Existing clinical data on the application of HRUS in early assessment of OXAIPN are restricted to two small-samplesize studies. In 2013, Briani et al. conducted a cross-sectional study exploring the use of nerve HRUS in a cohort of 15 oxaliplatin-treated patients. The results showed an increased in CSA at common entrapment sites in 9 out of 15 patients prior to clinical symptoms and neurophysiological changes. At the end, 13 of the 15 patients developed a sensory axonal neuropathy [124]. More recently, Pitarokoili et al. conducted a prospective study on 13 oxaliplatin treated patients confirming an increase in CSA at upper limb entrapment sites and figuring out a CSA enlargement in tibial and fibular nerves. These findings also

appeared either prior to or simultaneously with clinical and neurophysiology OXAIPN detection. No correlation between nerve HRUS, clinical severity and neurophysiology abnormalities was detected [125]. Moreover, a cross-sectional study of magnetic resonance neurography (MRN) in OXAIPN reported a significant dorsal root ganglia (DRG) hypertrophy after evaluating 20 patients. This finding correlates with one of the major mechanisms described for the initiation of neurotoxicity induced by this agent: the accumulation of platinum-DNA adducts in DRG [3]. Further investigation is required to establish the role of MRN as predictor of OXAIPN. Corneal confocal microscopy (CCM) is an ophthalmic noninvasive imaging modality that provides information of small sensory fibers by direct observation and scanning of corneal innervation with high resolution and magnification. Two prospective studies in gastrointestinal cancer patients are available, with conflicting results. Campagnolo M et al. demonstrated a reduction in the number and density of the corneal fibers after chemotherapy completion in 15 oxaliplatin treated patients [126]. Conversely, another study using CCM in 13 patients with upper GI cancer, eight of them who received 3 cycles of OXA containing regimes, identified a significant increase in corneal nerve fibre length [127].

Besides neuroimaging, other imaging techniques have been explored as surrogate biomarkers of early neurotoxicity. Preliminary evidences of the analysis body composition by computed tomography (CT) in gastrointestinal cancer patients reveal the loss of lean body muscle [128] or sarcopenic obesity [29] were independently associated with the occurrence of OXAIPN, supporting its potential predicitive usefulness that deserves further investigation. Other radiologic measures including spleen volumetric analysis in CT scan [129] or muscle ultrasound [130] have been anecdotically explored in OXAIPN. To date, further larger research in imaging techniques is needed to provide further in-depth objective evidence in order to transfer them into daily clinical practice as predictive biomarkers of OXAIPN.

#### **6. Blood Biomarkers**

Several whole blood biomarkers have been investigated as predictive tools in the assessment of OXAIPN. Neurofilament light chain (NfL) is a neuron-specific cytoskeletal protein expressed in large-myelinated axons [131] released in blood when nerve damage occurs. Supported by a preclinical studies in vincristine-induced peripheral neuropathy [132], Kim et al. conducted a prospective study including 43 patients treated with OXA and measured serum NfLs during (at 3 months) and post-treatment (at 6 months). An increase in NfL levels in both periods was observed, being higher at 6 months of OXA-treatment. High serum NfL levels correlated with neuropathy severity, providing compelling evidence of NfLs as a potential predictive marker of OXAIPN. Interestingly, after 4–6 months of follow-up a meaningful reduction on NfL levels was observed, indicating that NfLs can also discriminate recovery patients from OXAIPN [133]. Despite promising, the predictive usefulness of NfL as a biomarker requires independent validation with further studies with larger sample sizes that allow researchers to establish universal reference values in order to maximize the correct interpretation of NfL in the management of OXAIPN [131].

Preclinical evidences showed a significant reduction in Nerve Growth Factor (NGF) levels during OXA administration in rat [134], which was correlated with the onset of peripheral neuropathy. Considering clinical studies, limited data is available. Velasco et al. conducted a prospective observational study including 60 cancer patients, of which 35 were OXA-treated with. The objective was to investigate the changes in circulating NGF levels and IENF in the presence or absence of neuropathic pain [93]. This research was based on the rationale that NGF receptor, TrKA, is located in the terminals of sensory neurons. Thus, the interaction between NGF and its receptor activates intracellular pathways affecting the sensitivity of nociceptors. The results of the study demonstrated an association between increased NGF levels and patients developing painful chemotherapy-induced peripheral neuropathy (CIPN), whereas NGF levels remained stable in patients with either painless or absent CIPN. Additionally, NGF level increases correlated with the severity of

neuropathic pain reported by patients. However, no association between NGF levels and IENF was detected. Despite these promising results, previous studies did not report this association, therefore the role of NGF as a biomarker for the severity of painful OXAIPN remains unclear.

Other parameters routinely available from whole blood have been reported potentially useful in predicting the development of OXAIPN (Table 1). Pretreatment low hemoglobin level or anemia has been identified as a risk prognostic factor for OXAIPN in several studies. Mizrahi et al. carried out a large cross-sectional study including 105 patients treated with OXA. The results showed a correlation between reduced pretreatment levels of hemoglobin, detected in 24.5% of the cohort, with a greater severity of neuropathy [14]. In addition, previous studies on patients receiving OXA concluded similar results [15,19,22,27]. At pathophysiological level, a plausible explanation relating low hemoglobin levels and chronic neurotoxicity development is still unknown.

Among pretreatment metabolic and nutritional blood-based biomarkers, there are conflicting results. Such as higher albumin level [14] as hypoalbuminemia [15,19] have been associated with the risk of developing OXAIPN. Despite higher rates of neurotoxicity were described in patients with low levels of magnesium, the lack of evidence in this line [14] would support the known inefficacy of calcium-magnesium supplementation during OXA treatment as a neuroprotective approach [9]. Very recently and for the first time, higher serum gamma-glutamyl transferase (GGT) and a lower level of vitamin D have been identified as independent predictors of grade 2–3 OXAIPN in the multivariate analysis of a retrospective study including 186 gastrointestinal cancer patients [19]. Despite their easy collection and measure, non-invasiveness and objective interpretation of the results make blood tests perfect candidates to monitor neurotoxicity, more research is needed to better understand the pathophysiologic mechanisms underlying its role as prognostic biomarkers.

#### **7. Conclusions**

Early detection of OXAIPN is essential in the prevention of irreversible nerve damage and should be prioritized when assessing and evaluating patients receiving OXA treatment so that adequate adjustment in scheduled treatment plan can be made. Early clinical and neurophysiological signs of OXAIPN can be observed after low doses of OXA. The assessment of acute neurotoxicity symptoms in the routine clinical evaluation is a reliable biomarker for predicting the occurrence and development of OXAIPN, and particularly, the first cycles can be very informative. Regarding neurological examination, conflicting evidence on the timing of the threshold impairments hamper their currently use to inform clinicians in the prophylaxis of neuropathy. The development of a clinical standardized prognostic neuropathy assessment tool in order to early detect neuropathy should be validated. Despite, currently not part of common oncology practice, neurological monitoring with NCS may provide valuable reliable metric data in accurately disclosing and following the course of OXAIPN over time, offering at the same time useful information for dose reduction or discontinuation of the treatment before the progression to severe OXAIPN. Additionally, screening methods incorporating pharmacogenetics, may help to predict OXAIPN on the basis of genetic susceptibility and, consequently, allow a better, more personalized treatment. The further identification and validation of the simplest, noninvasive reliable and valid blood biomarkers for the premature screening of OXAIPN to reduce the morbidity and impairment in the quality of life of patients with gastrointestinal cancers associated with chronic OXAIPN might be of particular interest in neuroprotection trials. In the next future, by combining clinical, neurophysiological, genetic and potentially serum-based risks, decision-making would be improved to optimize treatment and prevent potentially serious neurotoxicity. The best noninvasive and easy-to-perform objective method to early detect and follow OXAIPN progression in the daily clinical practice in the hospital setting warrants further investigation and validation in larger prospective studies. **Author Contributions:** Conceptualization, R.V. and A.A.A.; methodology, R.V., M.A., M.V. and A.A.A. writing—original draft preparation, R.V., M.A., M.V. and A.A.A.; writing—review and editing, R.V. and A.A.A.; supervision, R.V.; project administration, R.V.; funding acquisition, R.V. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was partially supported by a grant from Instituto de Salud Carlos III through the project PI20/00283 (Co-funded by European Regional Development Fund (ERDF)). We also thank CERCA Programme/Generalitat de Catalunya for institutional support.

**Institutional Review Board Statement:** Not applicable.

**Informed Consent Statement:** Not applicable.

**Data Availability Statement:** Data sharing not applicable.

**Conflicts of Interest:** The authors declare no conflict of interest. The funders had no role in the design of the study; in the collection, analyses, or interpretation of data; in the writing of the manuscript, or in the decision to publish the results.

#### **Abbreviations**


#### **References**


MDPI St. Alban-Anlage 66 4052 Basel Switzerland Tel. +41 61 683 77 34 Fax +41 61 302 89 18 www.mdpi.com

*Journal of Personalized Medicine* Editorial Office E-mail: jpm@mdpi.com www.mdpi.com/journal/jpm

MDPI St. Alban-Anlage 66 4052 Basel Switzerland

Tel: +41 61 683 77 34 Fax: +41 61 302 89 18

www.mdpi.com ISBN 978-3-0365-3807-5