**Role of Postoperative Complications in Overall Survival after Radical Resection for Gastric Cancer: A Retrospective Single-Center Analysis of 1107 Patients**

**Christian Galata 1, Susanne Blank 1, Christel Weiss 2, Ulrich Ronellenfitsch 3, Christoph Reissfelder <sup>1</sup> and Julia Hardt 1,\***


Received: 3 November 2019; Accepted: 25 November 2019; Published: 27 November 2019

**Abstract:** *Background:* The aim of this study was to investigate the impact of postoperative complications on overall survival (OS) after radical resection for gastric cancer. *Methods:* A retrospective analysis of our institutional database for surgical patients with gastroesophageal malignancies was performed. All consecutive patients who underwent R0 resection for M0 gastric cancer between October 1972 and February 2014 were included. The impact of postoperative complications on OS was evaluated in the entire cohort and in a subgroup after exclusion of 30 day and in-hospital mortality. *Results:* A total of 1107 patients were included. In the entire cohort, both overall complications (*p* < 0.001) and major surgical complications (*p* = 0.003) were significant risk factors for decreased OS in univariable analysis. In multivariable analysis, overall complications were an independent risk factor for decreased OS (*p* < 0.001). After exclusion of patients with complication-related 30 day and in-hospital mortality, neither major surgical (*p* = 0.832) nor overall complications (*p* = 0.198) were significantly associated with decreased OS. *Conclusion:* In this study, postoperative complications influenced OS due to complication-related early postoperative deaths. In patients successfully rescued from early postoperative complications, neither overall complications nor major surgical complications were risk factors for decreased survival.

**Keywords:** gastric cancer; gastrectomy; complications; outcome; survival

#### **1. Introduction**

Even today, surgery for gastric cancer remains challenging, and patients undergoing radical resection are reported to have high complication and failure-to-rescue rates [1,2]. Failure-to-rescue rates are reported to be even higher after surgery for gastric cancer than after esophageal resections [3]. Recently, several studies have reported adverse effects of postoperative complications on overall survival (OS) in these patients. Such studies have attracted particular interest as they suggest that postoperative complications have a negative impact on oncologic outcomes. However, to understand the importance of postoperative morbidity for oncologic outcomes, patients with complication-related early postoperative mortality must be critically considered. A recent systematic review and meta-analysis including 16 retrospective studies found that postoperative complications are correlated with poor prognosis after radical gastrectomy [4]. Thirteen of these studies reported effects of postoperative

complications on OS, but only eight excluded influences from in-hospital death in the survival analysis. This is of particular interest because the pooled hazard ratio in this meta-analysis was notably lower after the exclusion of in-hospital mortality (1.40 vs. 1.79). Moreover, of the six studies reporting correlations between postoperative infectious complications and OS, only four excluded in-hospital mortality. Furthermore, the authors found a lower pooled hazard ratio (1.47 vs. 1.86) depending on whether in-hospital mortality was excluded from their analysis. They also reported different 95% confidence intervals (CI) for these two scenarios (1.22–2.83 for included in-hospital mortality vs. 0.90–2.40 for excluded in-hospital mortality).

In general, studies investigating the effect of postoperative morbidity on OS after surgery for gastric cancer either exclude or include patients with complication-related early postoperative deaths. Data on how the inclusion or exclusion of in-hospital mortality affects the role of postoperative complications in decreased OS within the same study population are scarce. The aim of this study was to investigate risk factors for decreased OS in patients undergoing radical resection for gastric cancer with special regard to the effect of postoperative complications. Therefore, we investigated two different cohorts, one including and one excluding patients with complication-related postoperative deaths.

#### **2. Results**

#### *2.1. Patient Characteristics*

Data of 1107 consecutive patients who underwent R0 resection for M0 gastric cancer at our institution between October 1972 and February 2014 were included in the analysis. Patient characteristics are shown in Table 1. There were more males than females (54.9% vs. 45.1%) and the median age was 65 years old. Most tumors were proximally located (60.9%) and predominantly classified as non-diffuse type (59.1%) according to the Laurén classification. The proportion of signet-ring cell carcinomas was 27.6%. As a rigorous standard at our institution, all oncologic resections with curative intent were performed by senior surgeons specialized in upper gastrointestinal surgery. During the study period, the standard approach for radical resection of gastric cancer at our department was open total or subtotal (4/5) gastrectomy with D2 lymphadenectomy (LAD). Total gastrectomy was performed in 47.1% of the cases and subtotal gastrectomy in 52.9% of the cases. Types of reconstruction after total gastrectomy were Roux-en-Y (n = 313, 28.3%), Longmire's reconstruction (n = 143, 12.9%), Schloffer's reconstruction (n = 69, 5.7%) and esophago-duodenostomy (n = 2, 0.2%). For subtotal gastrectomies, the type of reconstruction was documented in 489 cases (83.5%). Most patients received Billroth II procedures (total: n = 447, 40.3%; retrocolic: n = 234, 21.1%; antecolic: n = 213, 19.2%), whereas Billroth I procedures were performed in 42 patients (3.8%). The remaining 97 patients (16.5%) underwent subtotal gastrectomy without documentation of the reconstruction method. Multivisceral resections were performed in 45.9% of the patients, with splenectomies (34.2%) and cholecystectomies (10.9%) being the most common procedures.

Locally advanced tumor stages (pT2–4) were observed in 75.9% of the patients, and positive lymph nodes at the final pathology workup (pN1–3) were present in 52.1%. While pN stages were documented from the inception of the database, for over more than four decades, the extent of LAD and the number of harvested lymph nodes were not documented until 1998. However, when the 214 cases where the number of harvested lymph nodes was available were analyzed, the median number of harvested lymph nodes was 21 (17–27), indicating an adequate extent of LAD. Of the 136 patients where D1–3 LAD was documented, the vast majority underwent D2 LAD (n = 120, 86.9%) while D1 LAD and D3 LAD was performed in 10 (7.3%) and 8 (5.8%) patients, respectively. Data on the American Society of Anesthesiologists (ASA) physical status classification system was not documented before 2008. When patients with data on ASA grading (n = 46) were evaluated, 13.0% were categorized as ASA I, 37.0% as ASA II, 45.7% as ASA III and 4.3% as ASA IV.


**Table 1.** Patient characteristics.

IQR: interquartile range; AJCC: American Joint Committee on Cancer; UICC: Union Internationale Contre le Cancer.

#### *2.2. Postoperative Outcomes*

Postoperative outcomes are reported in Table 2. The median length of hospital stay was 14 (13–19) days. An overall complication rate of 25.3% was observed. Major surgical complications (defined as anastomotic leak, postoperative abdominal abscess, fascial dehiscence, peritonitis, sepsis, secondary hemorrhage, and relaparotomy for any reason during the postoperative course) were observed in 10.6% of the patients. The number of overall (*p* = 0.126) and major (*p* = 0.238) postoperative complications between total and subtotal gastrectomies was not significantly different. Postoperative 30 day mortality rate and in-hospital mortality rate were 4.7% and 5.7%, respectively. The median follow-up time was 27 (10–70) months with an estimated 5 year survival rate of 53.7%. OS was significantly different across American Joint Committee on Cancer/Union for International Cancer Control (AJCC/UICC) stages (*p* < 0.001). Median OS of all patients was 61 (95% CI: 50.05–71.95) months. Figure 1 shows the corresponding Kaplan–Meier survival curves. Operating times and intraoperative blood loss were not documented before 2004 and 2005, respectively. However, when patients with data on operating time (n = 132) were analyzed, the median operating time was 248 (213–298) min. For patients with data on intraoperative blood loss (n = 82), a median blood loss of 300 (200–600) mL was observed.


**Table 2.** Patient outcomes.

**<sup>a</sup>** Patients with in-hospital mortality excluded **<sup>b</sup>** All patients.

**Figure 1.** Kaplan–Meier survival curves. OS was significantly different across UICC stages (*p* < 0.001). Median OS of all patients was 61 (95% CI: 50.05–71.95) months. OS: overall survival; UICC: Union Internationale Contre le Cancer.

To investigate the impact of postoperative complications on OS, patients were stratified into two cohorts: one cohort consisting of all patients in the study and one cohort comprising only patients without complication-related postoperative deaths (30 day mortality and in-hospital mortality). Univariable and multivariable Cox regression analyses were performed to determine the parameters that might influence OS.

#### *2.3. Risk Factors for Decreased Overall Survival in the Entire Cohort*

The impact of clinically relevant variables on OS of all patients in the study is shown in Table 3. Overall complications (*p* < 0.001), major surgical complications (*p* = 0.003) and anastomotic leak (*p* < 0.001) were significant risk factors for decreased OS in univariable analysis. Other significant risk factors in univariable analysis were higher pT (*p* < 0.001), higher pN (*p* < 0.001), and higher AJCC/UICC stages (*p* < 0.001), older patient age (*p* < 0.001), earlier year of surgery (*p* = 0.003), non-antropyloric

compared to antropyloric tumor location (*p* = 0.025), multivisceral resection (*p* = 0.041), splenectomy (*p* = 0.012), additional intestinal resections (*p* < 0.001), and additional pancreatic procedures (*p* = 0.010). When multivariable analysis was performed, the occurrence of overall postoperative complications was an independent risk factor for decreased OS (*p* < 0.001) together with advanced pT (*p* < 0.001) and pN (*p* < 0.001) stages, older patient age (*p* < 0.001), and earlier year of surgery (*p* < 0.001). For patients with data on operating time and intraoperative blood loss were available, neither parameter had a significant impact on OS in univariable analysis (operating time: *p* = 0.327, HR 0.997; blood loss: *p* = 0.147; HR 0.999).



AJCC: American Joint Committee on Cancer; OS: Overall survival; UICC: Union Internationale Contre le Cancer. *p* values in bold type indicate statistical significance in multivariable analysis.

#### *2.4. Risk Factors for Decreased Survival after Exclusion of Early Postoperative Mortality*

For this subgroup analysis (n = 1042), patients with complication-related early postoperative mortality (30 day mortality and in-hospital mortality) were excluded (Table 4). Overall complications (*p* = 0.198), major surgical complications (*p* = 0.832) and anastomotic leak (*p* = 0.396) did not reach statistical significance as risk factors for decreased OS in univariable analysis. Significant risk factors in univariable analysis were advanced pT (*p* < 0.001), pN (*p* < 0.001), and AJCC/UICC (*p* < 0.001) stages, older patient age (*p* < 0.001), proximal tumor location (*p* = 0.008), multivisceral resection (*p* = 0.007), splenectomy (*p* = 0.004), additional intestinal resections (*p* < 0.001), additional pancreatic procedures (*p* = 0.006), diffuse histologic type according to the Laurén classification (*p* = 0.046), and performance of a total vs. subtotal gastrectomy (*p* = 0.027). Earlier year of surgery did not reach statistical significance on the α = 0.050 level in univariable analysis but showed a trend towards shorter OS (*p* = 0.064) and was therefore included in the multivariable analysis. In a multivariable Cox regression analysis, independent risk factors associated with decreased OS were advanced pT (*p* < 0.001) and pN (*p* < 0.001) stage, older patient age (*p* < 0.001), earlier year of surgery (*p* = 0.010), proximal tumor location (*p* = 0.040), and diffuse histologic type according to the Laurén classification (p = 0.013). For patients with data on operating time and intraoperative blood loss available, neither parameter had a significant impact on OS in univariable analysis (operating time: *p* = 0.821, HR 0.999; blood loss: *p* = 0.290; HR 0.999).


**Table 4.** Factors associated with OS (in-hospital mortality excluded).

AJCC: American Joint Committee on Cancer; OS: Overall survival; UICC: Union Internationale Contre le Cancer. *p* values in bold type indicate statistical significance in multivariable analysis.

#### *2.5. Neoadjuvant and Adjuvant Treatment*

Administration of neoadjuvant and adjuvant chemotherapy was not documented in the database before the year 2005 and 2007, respectively. When patients treated with neoadjuvant (n = 106) and adjuvant (n = 75) therapy were investigated, neither treatment had a significant impact on OS in univariable analysis, neither before (neoadjuvant: *p* = 0.104, HR 0.466; adjuvant: *p* = 0.698, HR 1.229) nor after exclusion of in-hospital mortality (neoadjuvant: *p* = 0.214, HR 0.550; adjuvant: *p* = 0.501, HR 1.461).

#### *2.6. Subgroup Analysis of AJCC*/*UICC Stages after Exclusion of Early Postoperative Mortality*

As the impact of postoperative complications on OS might vary between patients with different tumor stages, subgroup analyses for AJCC/UICC stages I–IV were performed after exclusion of early postoperative mortality. For patients with AJCC/UICC stage I (n = 489) and IV (n = 57), neither overall complications (*p* = 0.452, HR 0.843; *p* = 0.669, HR 1.219), nor major postoperative complications (*p* = 0.274, HR 0.698; *p* = 0.521, HR 0.519) or anastomotic leak (*p* = 0.420, HR 0.624; *p* = 0.743, HR 0.713) reached statistical significance in univariable analysis. For patients with AJCC/UICC stage II (n = 249) no influence of overall complications (*p* = 0.100), anastomotic leak (*p* = 0.234) nor major surgical complications (*p* = 0.061) on OS was observed in univariable analysis. When multivariable analysis was performed (Appendix A Table A1), only the type of gastrectomy (*p* = 0.037) and the year of surgery (*p* = 0.001) remained in the model as factors with significant impact on OS. For patients with AJCC/UICC stage III (n = 246), no significant influence of overall complications (*p* = 0.288), major surgical complications (*p* = 0.705), or anastomotic leak (*p* = 0.097) on OS was observed in univariable analysis. When multivariable analysis was performed (Appendix A Table A2), older patient age (*p* = 0.012) was the only significant factor associated with OS.

#### **3. Discussion**

This study examined factors associated with OS after radical resection for gastric cancer over a time period of more than four decades in a cohort of 1107 consecutive patients at a European university surgical center. In our previous work, we investigated trends in postoperative morbidity, mortality, and failure to rescue in this study population (Christian Galata, Ulrich Ronellenfitsch, Susanne Blank, Christoph Reissfelder, Julia Hardt: Postoperative morbidity and failure to rescue in surgery for gastric cancer: A single center analysis of 1107 patients from 1972 to 2014; submitted and under review, November 2019). Here, we aimed to investigate risk factors for decreased OS with particular consideration for the role of postoperative overall complications and major surgical complications.

The main finding of this study was that postoperative complications had a significant impact on OS in the entire cohort in both univariable and multivariable analysis. However, when complication-related early postoperative deaths (30 day mortality and in-hospital mortality) were excluded, a statistically significant effect of postoperative complications on patient survival was no longer observed. This also holds true for all AJCC/UICC stages subgroups.

The clinicopathological features of the patients in our study are comparable to those of other long-term evaluations [5]. Multivariable analysis identified overall complications, advanced pT stage, advanced pN stage, older patient age, and earlier year of surgery as risk factors for decreased OS in the entire cohort. After exclusion of patients with early postoperative mortality, multivariable analysis rendered advanced pT stage, advanced pN stage, older patient age, earlier year of surgery, proximal tumor location, and diffuse histologic type according to the classification of Laurén as risk factors for decreased OS. The risk factors identified in our cohort are in line with those reported by other authors [6,7]. Possible reasons for the association of an earlier year of surgery with worse OS are improvements in surgical strategy, perioperative management, and oncologic therapy. Notably, when complication-related postoperative deaths were excluded neither major surgical complications nor anastomotic leak or overall complications were significant risk factors for decreased OS in univariable and multivariable analysis.

Recently, a number of studies have investigated the impact of postoperative complications on OS after surgery for gastric cancer, but the way in which the results have been reported is inconsistent. Some studies excluded patients with in-hospital mortality [6,8–11], while others did not exclude in-hospital mortality, thus potentially increasing the chance of detecting significant correlations between postoperative morbidity and patient survival [7,12–16]. A meta-analysis by Wang et al. found postoperative complications to correlate with poor prognosis after radical gastrectomy [4]. However, this effect was markedly weaker when the authors analyzed the subgroup of studies in which patients with in-hospital mortality were excluded.

In general, studies on this topic usually either include or exclude early postoperative mortality. Data on the effects of including or excluding surgery-related mortality are rarely reported for the same study population. In our analysis, we show that postoperative complications have a significant impact on OS if complication-related early postoperative deaths are not excluded. After exclusion of 30 day and in-hospital mortality, neither overall complications nor major surgical complications or anastomotic leak showed a significant effect on OS in the entire cohort. Likewise, no significant effect of these parameters on OS was observed in multivariable analysis when subgroup analyses of AJCC/UICC stages were performed. Thus, our data suggest that the identification of postoperative complications as a risk factor for inferior oncologic outcomes after radical resection for gastric cancer may have been overestimated depending on whether and how early postoperative mortality is excluded. These results are not in line with the study results by Jin et al. who report that postoperative complications remained an independent risk factor for decreased OS after curative resection for gastric cancer, even after exclusion of patients who died within 30 days postoperatively. Moreover, patients who experienced postoperative complications were 50% less likely to receive adjuvant therapy. The combination of postoperative complications and failure to receive adjuvant therapy increased the risk of death more than twice compared to patients without postoperative morbidity who successfully underwent adjuvant therapy [17]. To summarize, the currently available evidence on the impact of postoperative morbidity on long-term oncologic outcome is still heterogeneous.

Some limitations of this study must be mentioned. The study was retrospective and, thus, the inherent potential for misclassification may limit the validity of our data. In addition, the median OS exceeded the median follow-up which may limit the conclusions. There may be confounding variables that were not available for analysis, in particular preoperative ECOG (Eastern Cooperative Oncology Group) performance status, which might impact patient outcomes. However, since we investigated a cohort that underwent radical resection with curative intent, it may seem justified to assume that the vast majority of our patients were in a general condition that allows for extensive upper abdominal surgery. This is supported by the data of patients for whom ASA grading was available, which were categorized as ASA II and III in 82.7% of the cases. No continuous documentation of the extent of LAD, the number of harvested lymph nodes or the administration of perioperative chemotherapy was available over the long period covered in this study, but data of patients for whom the extent of LAD or the number of harvested lymph nodes were available indicate that D2 LAD with adequate extent was performed for the vast majority of patients in our cohort. When patients with data on neoadjuvant and adjuvant therapy were analyzed, neither of the treatments had a significant impact on OS in univariable analysis, neither before nor after exclusion of in-hospital mortality. These data must be interpreted cautiously, as the small sample size and lack of data on the completeness of chemotherapy and chemotherapy regimens used may have led to a bias. Neoadjuvant treatment, which was introduced at our institution in 2005, was not found to increase postoperative morbidity or mortality in several other studies [18,19], but in itself improves the prognosis of patients with locally advanced gastric cancer [20]. Concerning the extent of LAD, a more aggressive (D 1–3 or D 1–2 instead of D1 alone) surgical procedure might be associated with an increase in postoperative morbidity [21,22]. On the other hand, a more radical lymph node dissection (D 1–2 vs. D1 alone) has been shown to prolong OS [23,24].

#### **4. Materials and Methods**

#### *4.1. Ethics Approval*

Ethics board approval was obtained from the Medical Ethics Commission II of the Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany (2019–849R). All patient data used in this analysis were completely anonymized. The study was performed according to the Declaration of Helsinki.

#### *4.2. Patients*

A retrospective analysis of our institutional database for surgical patients with gastroesophageal malignancies was performed. Medical records of 2252 consecutive patients operated on between October 1972 and February 2014 were examined, and 1107 patients with M0 gastric cancer who underwent R0 resection were included in the analysis. Patients with Barrett carcinoma, gastric remnant cancer, atypical gastric resections, esophageal resections or pT0 stage on final histology workup were excluded. Tumors of the subcardial stomach (Siewert type III) were included, whereas esophagogastric junctional adenocarcinomas (Siewert type I and II) were excluded, as these are classified and staged according to the esophageal scheme in the current AJCC/UICC system [25]. A flow chart of the study population is shown in Figure 2.

**Figure 2.** Flow chart of the study population.

#### *4.3. AJCC*/*UICC Stages*

For gastric cancer patients operated on between 1972 and 2001, AJCC/UICC stages according to the 5th edition of the AJCC/UICC staging system were available. The 6th and 7th edition of the AJCC/UICC classification were used from 2002 until 2009 and from 2010 until 2014, respectively. Before analysis, all patients in this study were restaged according to the 6th edition of the AJCC/UICC staging system for gastric cancer.

#### *4.4. Postoperative Complications*

Data on postoperative complications were extracted from the database, where they had been documented based on medical records. Major surgical complications were defined as one of the following events during the postoperative course: anastomotic leak (including duodenal stump insufficiency), postoperative abdominal abscess, fascial dehiscence, peritonitis, sepsis, secondary hemorrhage, and relaparotomy for any reason. When multiple complications occurred, the most severe complication was used for classifying if a patient had major surgical complications. Complication-related postoperative mortality is presented as early postoperative (30 day) and in-hospital mortality.

#### *4.5. Follow-Up and Overall Survival*

Follow-up in the database was based on medical records and direct contact with the patient or with the treating physicians. OS time was defined as the interval from surgery to death or latest time point the patient was known to be alive.

#### *4.6. Statistical Analysis*

Mean and standard deviations were calculated for quantitative variables. Qualitative variables were quoted as absolute numbers and relative frequencies. Median and interquartile range (IQR) are presented for skewed or ordinal scaled parameters. All statistical tests for the comparison of two groups were two-tailed. In general, a test result was considered statistically significant if *p* < 0.050. For qualitative variables, a Fisher's exact test was used. Univariable and multivariable Cox regression analyses were performed to identify factors that might influence OS. Variables reaching a significance level of α = 0.100 in univariable Cox regression analyses were used as covariates in multivariable Cox regression analyses. In the multiple analyses, the backward stepwise selection based on the probability of the Wald statistic was used, and a significance level of α = 0.050 was chosen to detect several parameters that might influence the outcome. Hazard ratios in the multiple analyses are presented together with their 95% CI. The Kaplan–Meier method was used to present survival data and the log-rank test was used to compare survival distributions. Statistical analyses were performed using the SAS statistical analysis software (release 9.4, Cary, NC, USA).

#### **5. Conclusions**

In summary, our data support the evidence that postoperative complications are a significant risk factor for poor OS in patients undergoing radical resection for gastric cancer. However, our study shows that this was an effect caused by complication-related early postoperative mortality. Indeed, postoperative complications did not have an impact on OS in patients who were successfully rescued from postoperative overall or major surgical complications.

**Author Contributions:** Conceptualization, C.G., S.B., U.R., C.R. and J.H.; Formal analysis, C.G. and C.W.; Methodology, C.G., S.B., C.W., U.R., C.R. and J.H.; Software, C.G. and C.W.; Supervision, U.R. and C.R.; Writing—original draft, C.G., S.B., U.R., C.R. and J.H.; Writing—review and editing, C.G., S.B., U.R., C.R. and J.H.

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

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

#### **Appendix A**

**Table A1.** Factors associated with OS in patients with AJCC/UICC stage II (in-hospital mortality excluded).


AJCC: American Joint Committee on Cancer; OS: Overall survival; UICC: Union Internationale Contre le Cancer.

*p* values in bold type indicate statistical significance in multivariable analysis.


**Table A2.** Factors associated with OS in AJCC/UICC stage III (in-hospital mortality excluded).

AJCC: American Joint Committee on Cancer; OS: Overall survival; UICC: Union Internationale Contre le Cancer. *p* values in bold type indicate statistical significance in multivariable analysis.

#### **References**


© 2019 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*

### **Clinical Pathways for Oncological Gastrectomy: Are They a Suitable Instrument for Process Standardization to Improve Process and Outcome Quality for Patients Undergoing Gastrectomy? A Retrospective Cohort Study**

**Patrick Téoule 1, Emrullah Birgin 1, Christina Mertens 2, Matthias Schwarzbach 3, Stefan Post 1, Nuh N. Rahbari 1, Christoph Reißfelder <sup>1</sup> and Ulrich Ronellenfitsch 4,\***


Received: 7 January 2020; Accepted: 12 February 2020; Published: 13 February 2020

**Abstract:** (1) *Background*: Oncological gastrectomy requires complex multidisciplinary management. Clinical pathways (CPs) can potentially facilitate this task, but evidence related to their use in managing oncological gastrectomy is limited. This study evaluated the effect of a CP for oncological gastrectomy on process and outcome quality. (2)*Methods*: Consecutive patients undergoing oncological gastrectomy before (*n* = 64) or after (*n* = 62) the introduction of a CP were evaluated. Assessed parameters included catheter and drain management, postoperative mobilization, resumption of diet and length of stay. Morbidity, mortality, reoperation and readmission rates were used as indicators of outcome quality. (3) *Results*: Enteral nutrition was initiated significantly earlier after CP implementation (5.0 vs. 7.0 days, *p* < 0.0001). Readmission was more frequent before CP implementation (7.8% vs. 0.0%, *p* = 0.05). Incentive spirometer usage increased following CP implementation (100% vs. 90.6%, *p* = 0.11). Mortality, morbidity and reoperation rates remained unchanged. (4) *Conclusions*: After implementation of an oncological gastrectomy CP, process quality improved, while indicators of outcome quality such as mortality and reoperation rates remained unchanged. CPs are a promising tool to standardize perioperative care for oncological gastrectomy.

**Keywords:** clinical pathways; gastric surgery; oncological gastrectomy; quality of care; outcomes; standardization

#### **1. Introduction**

Gastric cancer is the fifth most common neoplasm and still ranks third among the world's leading causes of cancer deaths, affecting approximately 783,000 people annually [1]. Regardless of improvements in surgical technique and perioperative management, surgery for gastric cancer remains challenging and patients who undergo radical resection are reported to have high complication rates [2,3]. One reason is that more and more elderly and multimorbid patients are resected [4,5]. On the other hand, due to preoperative malnutrition of patients with gastric neoplasms and chronic comorbidities, perioperative mortality can reach up to 8.8% [6]. Therefore, multidisciplinary perioperative management is required to reduce the risk of possibly severe perioperative complications during and after oncological gastrectomy. The implementation of clinical pathways (CPs) can potentially improve the quality of perioperative management [7]. CPs are specific instruments developed to improve the quality of outcomes of care by standardizing treatment processes. They can be defined as a protocol stipulating all tasks that should be carried out during a defined treatment [8–10]. The designated goal of CPs is to transfer evidence to the bedside. They comprise all disciplines involved in patient care [11,12]. For several gastrointestinal operations, CPs have proven advantageous with regard to perioperative outcomes [13]. Several studies have reported the results of patients undergoing oncological gastrectomy and treated with CPs. These studies showed a reduction in the length of stay (LOS) and reported a non-significant decrease in total complications, mortality and reoperation [14]. However, all of these studies were conducted in Asian countries. In Europe only a few studies have assessed the influence of multimodal management after gastrectomy. They were focused on laparoscopic gastrectomy or a comparative pre-CP group was missing [15–18].

Given that the expected effects of CPs must be considered specific to health systems, we performed a study in a German tertiary care hospital to evaluate an oncological gastrectomy CP with respect to its effects on process and outcome quality.

#### **2. Results**

#### *2.1. Patient Characteristics*

A total of 126 patients underwent oncological gastrectomy during the study period. The pre-CP group comprised 64 patients and the CP group involved 62 patients. Patient characteristics are displayed in Table 1. The clinical and demographic characteristics of both groups were comparable. The proportion of total gastrectomies was non-significantly higher in the pre-CP group, and correspondingly, there were proportionally more tumors extending to the entire stomach in this group. The type of surgical reconstruction differed significantly between the two groups. While all patients received a Roux-en-Y reconstruction, the proportion of handsewn esophagojejunostomies was higher in the pre-CP group (23.4%) than in the CP group (8.1%; *p* = 0.01).


**Table 1.** Characteristics of the study groups.


**Table 1.** *Cont.*

ASA = American Society of Anesthesiology; X = missing data; Pre-CP group = Pre-Clinical pathway group; CP group = Clinical pathway group; dignity others Pre-CP-Group = in declining order: two neuroendocrine tumors, one leiomyosarcoma; dignity others CP-Group = in declining order: two leiomyosarcomas, one leiomyoma, one gastric metastasis of kidney cell carcinoma; IQR = interquartile range; # = multiple answers are possible; g/l = gram/ liter; \* = *p*-value ≤ 0.05.

#### *2.2. Process Quality*

Table 2 gives an overview of the comparison of the outcomes that reflect process quality. In the CP group, patients received liquid nutritional supplements significantly earlier (median 5.0 vs. 7.0 days in the pre-CP group; *p* < 0.0001). The usage of incentive spirometers increased following CP implementation, although the difference did not reach statistical significance (100% vs. 90.6% in the pre-CP group; *p* = 0.11). Foley and arterial catheters were removed significantly earlier in the pre-CP group (median of 1.0 vs. 4.0 and 2.0 vs. 5.0 days, respectively; *p* = 0.01).



Pre-CP Group = Pre-Clinical pathway group; CP group = Clinical pathway group; PDA = peridural anesthesia; IQR = interquartile range; EF = easy flow; X = missing data; \* = *p*-value ≤ 0.05.

#### *2.3. Outcome Quality*

Table 3 summarizes the results for outcome quality. There were two postoperative fatalities in the pre-CP group. Cause of death was respiratory failure following aspiration pneumonia in one case and multiorgan failure caused by sepsis following anastomotic leakage in the other. In the CP group, four patients died due to multiorgan failure caused by sepsis: one caused by duodenal stump leakage with severe peritonitis, one caused by aspiration pneumonia and myocardial infarction, one due to anastomotic leakage, and one due to bowel leakage with severe peritonitis.

Regarding outcome quality, groups differed significantly in three parameters. Median length of hospital stay (LOS) in the intermediate care and intensive care units was significantly shorter in the pre-CP group than the CP group (median stay 2.0 vs. 3.0, *p* = 0.0005; and 0.0 vs. 0.0, *p* = 0.01, respectively). The median of the highest measured visual-analogue-scale (VAS) pain score was significantly lower in the pre-CP group (5 compared to 6 in the CP group; *p* = 0.03). The readmission rate was higher in the pre-CP group (7.8% vs. 0; *p* = 0.05). No differences could be observed between groups with regard to postoperative morbidity and mortality. Additionally, groups did not differ

regarding the summary measures for specific complications. The discharge goal of the CP could not be obtained and LOS did not differ between groups.


**Table 3.** Parameters of outcome quality.

Pre-CP-Group = pre-clinical pathway group; CP-Group = clinical pathway group; VAS = visual analogue scale; IMC = intermediate care unit; ICU = intensive care unit; RBCC = red blood cell concentrate; IQR = interquartile range; \* = *p*-value ≤ 0.05.

#### **3. Discussion**

This study assessed the effects of an oncological gastrectomy CP with regard to parameters of perioperative process and outcome quality. Because gastric surgery and the associated perioperative care are complex, it should only be done in a specialized setting by dedicated and experienced surgeons. A reduction in perioperative mortality has been observed in recent years. However, procedure-associated morbidity remains high and this is a relevant issue for patients and treatment teams [19,20]. The fact that much older and severely co-morbid patients, as well as patients in advanced tumor stages and with compromised performance status are resected might partly explain this fact [3–5]. Nevertheless, high morbidity and mortality might also be associated with insufficient standardization of perioperative treatment, and in particular with so called "failure to rescue", a situation in which emerging complications are not detected and managed appropriately, resulting in the death of the patient [2,21–24]. Therefore, this study was designed to assess if implementing an oncological gastrectomy CP resulted in increased standardization of perioperative treatment and improved the process and outcome quality. Given that the relevant evidence is almost exclusively related to Asian countries [14,18,25], we conducted a study in a Germany tertiary care center.

In order to measure protocol adherence, process quality parameters were used as key performance indicators. Following CP implementation, we detected an improvement in some of these parameters, while others remained unaltered or even worsened.

A meta-analysis has shown that early enteral nutrition is associated with lower mortality and a shorter hospital stay after gastrectomy [26]. We observed a significantly earlier intake of liquid nutritional supplement, and a non-significantly earlier intake of soft and full diet after CP implementation. The incidence of postoperative pneumonia can be decreased by the use of incentive spirometers [27]. All patients used incentive spirometers after CP implementation, compared to only 90% in the pre-CP group. The fact that postoperative pneumonia did not decrease after CP implementation is therefore rather surprising. One potential explanation could be that more ASA III patients, who have a higher baseline risk for acquiring pneumonia, were operated on after CP implementation (56.1% vs. 46.7%). Given that ascending infections are related to indwelling catheters, early removal should be aimed for [28–30]. In our study, however, the median day when abdominal drains as well as peripheral and central venous catheters, epidural catheters and nasogastric tubes were removed remained unchanged after CP implementation. Drain fluid was checked for its amylase concentration on postoperative day 5 in all patients. Drains remained in situ in case of an elevated concentration. Therefore, a potential explanation for the delayed easy flow (EF) drain removal might be the higher proportion of pancreatic fistula in the CP group, with 11.3% vs. 6.3% for the pre-CP group, as well as duodenal stump leakage rate (3.2 vs. 0). In contrast to what was expected from CP implementation, two parameters showed an apparent decrease regarding their process quality. Foley and arterial catheters were removed on average one day later in the CP group. One hypothetical explanation for the delayed removal in patients treated with the CP could be that they stayed on average one day longer in intermediate care and intensive care units. A higher proportion of associated procedures and co-morbid patients could explain this fact.

Perioperative morbidity and mortality were not significantly different before and after CP implementation. While the 30-day mortality rate is frequently used, we employed the in-hospital mortality rate to account for prolonged treatment courses, which are common nowadays given advanced intensive care and interventional techniques. In-hospital mortality was 6.5% in the CP and 3.1% in the pre-CP group. This two-fold increase in mortality after CP implementation is worrisome. However, this observation is based on only two additional postoperative fatalities in the CP group, and the difference is not statistically significant. The result might therefore be spurious and must be interpreted with much caution. In comparison, the overall postoperative morbidity rate according to the Clavien-Dindo classification in our patients seems high. This can possibly be explained by the fact that this scheme counts every deviation from what is considered a normal postoperative course as a complication. Consequently, only 14 patients in the CP and 20 in the pre-CP group were classified as being without complications in our study.

The Enhanced Recovery After Surgery (ERAS®) society published perioperative care guidelines for gastrectomy [31]. These guidelines contain 25 care items compared to 23 items in our CP. Comparing the two documents, 17 recommendations are very similar, while six recommendations given by the ERAS guidelines are not included in our CP. Examples are as follows: surgical access type, transversus abdominis plane (TAP) block or the use of wound catheters, skin preparation, preanesthetic medication, prophylaxis for postoperative nausea and vomiting (PONV), and oral bowel preparation. In contrast to the ERAS guidelines, our CP comprises recommendations regarding vitamin B12 substitution, catheters, transfusion and nursing and rehabilitation. Possible future revisions of the CP should incorporate the evidence-based ERAS guidelines.

While the results regarding process quality were encouraging, three parameters related to outcome quality deteriorated after CP implementation. The LOS in the intermediate and intensive care units was significantly longer in the CP group. Moreover, the median of the highest visual-analogue-scale (VAS) pain score was significantly lower in the pre-CP group. This result is rather unexpected, given that the CP included a dedicated analgesia scheme according to recent recommendations. It also included epidural catheter placement, which was carried out in the overwhelming majority of patients. Additional oral analgesics were administered in a stepwise, pain-adjusted manner, so that there is no obvious explanation for higher pain levels in patients treated according to the CP. Therefore, a clear explanation for higher pain levels in the CP group is lacking. Hypothetically, nursing staff might have been more aware of possible postoperative pain after CP implementation, and consequently tended to carry out more accurate pain assessment, leading to a higher reported pain level. This would also explain why the stipulated goal of epidural catheter removal on day 3 was not met. This scenario could be regarded as ascertainment bias. On the other hand, extra requests for analgesics from patients did not differ between the groups treated with or without CP. This indicates that the stipulated analgesia scheme was quite sufficient. Inadequate pain management can lead to impaired mobilization, an increase in LOS, and ultimately, to elevated perioperative morbidity, particularly with regard to pulmonary complications.

CPs should also avoid excessively long LOS without medical reasons. In this study, we did not observe a decrease in LOS after CP implementation. However, a relevant variation in LOS was seen between individual patients. The stipulated goal for LOS in our CP might have been too ambitious, because it was clearly below the LOS reported in larger studies [14]. Moreover, the study comprised all consecutive patients, including those with severe postoperative complications. This may explain the large variation and exceedingly long LOS of some patients. The readmission rate was higher in the pre-CP group, which shows that patients treated with the CP were not discharged inappropriately early.

In summary, the implementation of a CP for oncological gastrectomy at our institution did not lead entirely to the results that were expected based on studies on gastrectomy CPs in Asian settings [32,33], and on studies on CPs for other procedures in abdominal surgery at our institution and in other settings [13,34–40]. The reasons for this apparent difference in the efficacy of gastrectomy and other abdominal surgery CPs can only be speculated on. It is known that the biology of the disease and care for patients undergoing gastrectomy for gastric cancer in Asia shows important differences compared to European settings [41], but it remains unclear which specific factors might have determined the lack of efficacy of our CP. Moreover, oncological gastrectomy potentially demands more complex perioperative care than other abdominal procedures, for which CPs have led to pronounced improvements in process and outcome quality [13,34–40,42,43]. From the results of this study, it is difficult to conclude if the lack of efficacy was mainly due to limited adherence to the CP, or due to its suboptimal content and design for the given setting.

One of the strengths of our study is that it included all consecutive patients undergoing oncological gastrectomy before and after CP implementation. This is comparable to the "intention to treat principle" in randomized trials. In the case where the individual goals of the CP were not met, the patient was not taken "off the pathway". All patients who entered the study were analyzed regardless of deviations from the CP or possible complications. Therefore, selection bias is highly unlikely.

There are several methodological limitations inherent to the study. Its design is retrospective and included a single center. Moreover, we used chart review to collect data. This could compromise the validity of the data. Furthermore, the small sample size could bias the results. Documentation was not fully complete for all patients with regard to some variables and consequently, these could not be used for the analyses. Although selectively missing documentation is unlikely, bias could result. A crossover or, in other words, contamination bias could have occurred during the development and implementation phase of the CP. Health professionals who were part of the development team could have used their knowledge of the CP content prior to its implementation in October 2012. To counteract such issues, the CP was actually designed and implemented over only three months. Due to the study design with two groups of patients operated on before and after a defined time point, i.e., implementation of the CP, patients were operated on during different periods. The treatment during these periods might have been different (beyond the usage of the CP) because of other factors that influenced the process and outcome quality. For example, it is indisputable that surgical technique, and the skills as well as the experience of the individual surgeon have an effect on perioperative

outcomes [44]. During the four-year study period, the surgeons who were in charge of and operated on patients changed. Therefore, surgical performance bias cannot be excluded.

Another weakness of our study is that not all stipulated goals were achieved after CP implementation. This suggests that not all team members adhered to the CP protocol. The main reasons for non-adherence to the main subitems have been explained above. Possibly, the addition of a dedicated study nurse to the CP team, and the introduction of an electronic CP could overcome non-adherence to the CP recommendations. The study nurse could promote protocol adherence and discuss the reasons for non-adherence with the appropriate caregivers. The use of an electronic CP checklist, with reminders in case of protocol deviation, could increase adherence, and thus potentially improve process and outcome quality.

Most of these limitations would have been avoidable if the study had been conducted as a randomized controlled trial. However, this is hardly feasible for studies evaluating CP usage in a single center because it usually requires cluster randomization [36,45].

#### **4. Materials and Methods**

#### *4.1. CP Design, Implementation, and Content*

Since 2006, the Department of Surgery, University Medical Center Mannheim, Medical Faculty Mannheim, Heidelberg University has implemented CPs for different surgical procedures in a stepwise manner [34–40,42,43,46,47]. In October 2012, a CP for oncological gastrectomy was introduced.

This CP is based on CPs for colorectal and bariatric surgery that incorporate ERAS elements. Both have been previously evaluated [36,43]. Specific elements were adapted to modify the CP for use in oncological gastrectomy. Both the original colorectal and the gastrectomy CP are based on published treatment and nursing recommendations. Furthermore, the best available evidence at the time of CP design was incorporated. The CP was designed and then implemented by a multi-hierarchical and interdisciplinary (anesthesiology, surgery, nutritional services, physiotherapy) team.

A literature review was done to identify current evidence on perioperative treatment elements. Subsequently, institutional standards that existed before, were integrated. Finally, all project participants agreed to the final CP version in a consensus meeting. Prior to the definitive implementation, all involved disciplines were trained to use the CP. After implementation of the CP, continuous efforts were made to enable further development and improvements of the CP based on suggestions made by staff.

A full version of the CP is provided in the online Supplementary Materials (Table S1). Its main contents are as follows: (1) hospital admission scheduled for the day before surgery; (2) epidural catheter placement; and (3) a stepwise oral pain medication scheme, based on non-opioids for all patients and on demand medication of potent opioids. Postoperatively, patients were transferred to a surgical intermediate care unit for at least one night. ICU admission took place only if deemed necessary by the surgeon and/or anesthesiologist. All patients were encouraged to drink sweetened tea until two hours prior to scheduled full anesthesia. An oral toluidine blue swallowing test was stipulated for postoperative day five. Drains were removed in case of a negative blue test and if respective enzyme levels in the drain fluid were not elevated (target drain: amylase <250 U/l in drain fluid). Detailed instructions on how to use an incentive spirometer were provided to patients. The stipulated day of discharge was postoperative day seven. Outpatient follow-up appointments were scheduled within 14 days after discharge. Patients were told to consult our emergency room in case of clinical irregularities. The rationale for incorporating the individual elements into the CP was that they were thought to either enhance recovery and thus shorten hospital stay, or to improve perioperative outcomes such as decreasing the risk of complications. Some of the elements (preoperative nutrition and smoking cessation, preoperative fasting and treatment with carbohydrates, epidural catheter placement, antithrombotic prophylaxis, antimicrobial prophylaxis, avoidance of hypothermia, glycemic control, urine catheter management, fluid balance, early and scheduled mobilization, and stimulation

of bowel movement) are recommended in the consensus guidelines for enhanced recovery after gastrectomy of the Enhanced Recovery After Surgery (ERAS®) Group [31]. The perioperative analgesia scheme was endorsed by national guidelines. Other CP elements such as the oral toluidine blue swallowing test and abdominal drain management were based on pre-existing institutional standards, which were not backed by higher-level evidence. The targeted length of hospital stay was based on the minimum stay for oncological gastrectomy defined in the German DRG system [48].

The CP was designed as a four-page paper-based document containing all designated treatment steps for each pre- and postoperative day. CPs were kept with patients' treatment charts, and therefore they were constantly available for all involved staff members.

#### *4.2. Study Design*

The study used a single-center retrospective cohort design. All consecutive patients undergoing elective oncological gastrectomy were included. The intervention group (CP group) comprised all patients operated on after CP implementation in October 2012 until September 2014. The control group (pre-CP group) included patients operated on before CP implementation (May 2010 to September 2012). No formal sample size calculation was done. Data were obtained by means of retrospective chart review.

Patients in the pre-CP group were treated according to the individual judgment and decisions taken by the treating surgeons. Several semiformal standards for selected elements of care (e.g., early removal of catheters, epidural analgesia and early mobilization) had been in place and were used prior to CP implementation, but there was no comprehensive tool covering the entire treatment continuum. In the CP group, all patients were treated according to the CP.

The study was approved by the ethical committee of the Medical Faculty of Mannheim of the University of Heidelberg (2015-823R-MA). Because of its retrospective nature, the requirement for informed consent to review medical records was waived by the ethical committee. Confidentiality of patient data was ensured. The study was conducted in compliance with the Declaration of Helsinki. Neither the individual de-identified participant data, nor the specific data are intended to be shared by the authors. The CP documents will be accessible indefinitely as online supplementary data. The study has been registered with the German Clinical Trials Register (DRKS00020323).

#### *4.3. Patient Characteristics*

Demographic and clinical characteristics included age, sex, and preoperative status of patients according to the American Society of Anesthesiologists (ASA) physical status classification [49], underlying disease, administration of neoadjuvant chemotherapy, tumor location, and serum albumin levels upon preoperative admission. Histopathological data were analyzed by the Department of Pathology, Universitätsmedizin Mannheim, Mannheim, Germany according to the 7th version of the TNM-classification [50].

#### *4.4. Surgery*

Both before and after CP implementation, surgery was carried out by dedicated upper GI surgeons with more than four years' experience. To achieve R0-resection, patients received either total, distal or completion gastrectomy, depending on the anatomic location of the tumor and possible previous gastric operations. There were no laparoscopic resections. Associated procedures were performed when necessary. The gastrointestinal passage was preferably reconstructed using a long Roux-en-Y loop. Esophagojejunostomy was performed with a 25 mm circular stapler whereas gastrojejunostomy was hand sewn. A D2 lymphadenectomy according to the guidelines of the Japanese Gastric Cancer Association should be performed in all patients.

#### *4.5. Study Outcomes*

Process and outcome quality were defined according to the Donabedian model [51,52]. Process quality was considered as the adherence to treatment specifications as detailed in the CP and was assessed using the following parameters: day of removal of the foley catheter and epidural catheter, placement of central venous line and epidural catheter, postoperative mobilization, day of removal of intra-abdominal drainage and nasogastric tube, day of oral toluidine blue test, and day of resumption of liquid and solid diet.

Outcome quality was measured with the following parameters: morbidity, mortality, reoperation rate, LOS stratified by the presence or absence of complications, day of first postoperative defecation, pain level on a numeric rating scale and readmission rate. Morbidity was assessed according to the Clavien-Dindo classification of postoperative complications [53]. Deaths were counted as postoperative if they occurred during the hospital stay or during readmission. Surgical site infections were ascertained according to the Centers for Disease Control and prevention (CDC) definition [54]. Readmission was counted as such if it occurred no later than 30 days after initial discharge and if it was considered to be related to a postoperative problem.

#### *4.6. Statistical Analysis*

All outcomes were compared between the CP and pre-CP group. Missing values were not counted in the analyses with no imputation of missing values having been performed. Dichotomous variables were compared between groups using the chi-square test. Ordinal variables were compared using the Student's *t*-test if they were normally distributed and the Mann-Whitney U-test if they were not normally distributed. For not normally distributed variables, the median was used for descriptive analyses. For normally distributed variables, the mean was used. *p*-values <0.05 were considered statistically significant. There was no adjustment for multiple testing. SAS 13.2 (Cary, NC, USA) was used for all statistical analyses.

#### **5. Conclusions**

This study showed that using a CP for oncological gastrectomy affects several aspects of perioperative treatment. A high degree of process standardization was achieved and the uptake of respiratory training and the timely initiation of enteral nutrition was ensured. Other expected changes such as better pain control, earlier mobilization and shorter LOS were not realized after CP implementation. Outcome quality measured with perioperative morbidity and mortality did not change after CP implementation. In conclusion, an oncological gastrectomy CP can be used to standardize perioperative care, but its utility must be carefully weighed against the anticipated cost and effort required for implementation and continuous development.

**Supplementary Materials:** The following are available online at http://www.mdpi.com/2072-6694/12/2/434/s1, Table S1: Clinical Pathway for oncological gastrectomy used in the CP group of the study.

**Author Contributions:** P.T. and U.R. participated in the conception and design of the study. P.T. performed data collection, analyzed the data and drafted the manuscript. P.T., U.R., C.M., M.S., E.B., S.P., N.N.R. and C.R. participated in the analysis and interpretation of data, and revision of the manuscript for important intellectual content. All authors have read and agreed to the published version of the manuscript and are in agreement to be accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

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

**Acknowledgments:** Sylvia Büttner from the Department of Medical Statistics, Biomathematics and Informatics Universitätsmedizin Mannheim, Medical Faculty Mannheim, Heidelberg University, Mannheim, Germany participated in the analysis of data.

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

### **References**


© 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* **The Influence of Pretherapeutic and Preoperative Sarcopenia on Short-Term Outcome after Esophagectomy**

### **Johanna Grün 1, Lea Elfinger 1, Han Le 1, Christel Weiß 2, Mirko Otto 1, Christoph Reißfelder <sup>1</sup> and Susanne Blank 1,\***


Received: 20 October 2020; Accepted: 14 November 2020; Published: 17 November 2020

**Simple Summary:** Although introducing minimally invasive surgery reduced postoperative morbidity after esophagectomy esophageal cancer still is a malignancy with poor prognosis. This study aimed to investigate whether preoperative sarcopenia has an influence on short-term postoperative outcome after esophagectomy in esophageal cancer patients. Our findings suggest that preoperative sarcopenia is no independent prognostic factor for postoperative outcome after esophagectomy but that patients' nutritional status consists of more factors than only body mass index (BMI) and muscle mass. Prehabilitation and preoperative optimization of the patients' nutritional status seems to be an important factor for short-term postoperative outcome after esophagectomy.

**Abstract:** By introducing minimally invasive surgery the rate of postoperative morbidity in esophageal cancer patients could be reduced. But esophagectomy is still associated with a relevant risk of postoperative morbidity and mortality. Patients often present with nutritional deficiency and sarcopenia even at time of diagnosis. This study focuses on the influence of skeletal muscle index (SMI) on postoperative morbidity and mortality. Fifty-two patients were included in this study. SMI was measured using computer tomographic images at the time of diagnosis and before surgery. Then, SMI and different clinicopathological and demographic features were correlated with postoperative morbidity. There was no correlation between SMI before neoadjuvant therapy (*p* = 0.5365) nor before surgery (*p* = 0.3530) with the short-term postoperative outcome. Regarding cholesterol level before surgery there was a trend for a higher risk of complications with lower cholesterol levels (*p* = 0.0846). Our findings suggest that a low preoperative SMI does not necessarily predict a poor postoperative outcome in esophageal cancer patients after esophagectomy but that there are many factors that influence the nutritional status of cancer patients. To improve nutritional status, cancer patients at our clinic receive specialized nutritional counselling during neoadjuvant treatment as well as after surgery.

**Keywords:** skeletal muscle index; esophagectomy; nutritional status; sarcopenia

### **1. Introduction**

Esophageal cancer is the eighth most common type of malignancy worldwide. As diseases like reflux and obesity are increasing worldwide, the incidence of Barrett's esophagus as well as the incidence of esophageal adenocarcinomas have been increasing during the last decades. Although the outcome after curative multimodal therapy for esophageal cancer improved in the last decades, esophageal cancer still has a poor prognosis with a 10-year survival rate of approximately 16–17% and 5-year survival rate of approximately 20% [1,2].

By introducing minimally invasive procedures including the robot-assisted esophagectomy the rate of postoperative morbidity after esophagectomy could be reduced [3,4]. But esophagectomy is still associated with a relevant risk of postoperative morbidity, especially respiratory complications and anastomotic leakage including the risk of mediastinitis [5,6]. Thirty-day mortality after esophagectomy ranges between 1% and 3% and is mostly a result of postoperative complications. Postoperative complications affect short-term but also long-term survival as well as quality of life [6–8].

The Esophageal Complication Consensus Group presented an evaluation of 2704 patients with an overall complication rate of 59%. In an evaluation of 1057 total minimally invasive transthoracic esophagectomies, 56% of the patients developed at least one complication after surgery of which 26.9% were classified as Clavien Dindo grade III or more [6,9].

There are a number of studies identifying risk factors for a poor short-term outcome after surgical resection of esophageal carcinomas, such as high age, congestive heart failure, coronary artery disease, peripheral vascular disease, hypertension, body mass index <25 kg/m2, and insulin dependent diabetes [10,11].

It has also been shown that the patient's nutritional status (including serum albumin, body mass index, muscle mass) plays an important role in determining patient outcome after surgery [12–14]. Patients with esophageal cancer often present with advanced disease and an impaired nutritional status due to dysphagia and cancer cachexia [15–17]. A good nutritional status before surgery has been proven to reduce the hypermetabolic response to surgery and to optimize wound and anastomotic healing and recovery [18–20].

Sarcopenia is defined by the European Working Group on Sarcopenia in Older People as low skeletal muscle mass and strength and is a risk factor for surgical patients in general [21,22]. It is also known as a possible risk factor for morbidity and poor prognosis after esophagectomy and could already be shown with a widespread prevalence, ranging from 16% to 79% before surgery [23–26].

A tool to measure skeletal muscle depletion is the skeletal muscle index (SMI). The skeletal muscle mass can be measured as cross-sectional area of the total skeletal muscle volume (cm2) at L3 in computertomographic images. The SMI is calculated as total skeletal muscle volume (cm2)/height2 (m2) [27,28]. It has been shown to be a prognostic factor independent from the body mass index (BMI) in oncologic patients [29].

Due to these findings the oncological patients at the Surgical Department of the University Hospital Mannheim are encouraged to take part in nutritional consulting at our clinic or at other specialized practices before and after surgery. Our clinic uses standardized protocols, guided by "Enhanced recovery after surgery"(ERAS) protocols, to improve postoperative outcome and length of hospital stay [30–33].

As a patient's nutritional status before surgery seems to play a major role for short term outcome, the aim of our study was to investigate whether sarcopenia, at the time of diagnosis or before surgery as well as BMI and serum albumin levels are risk factors for postoperative mortality and morbidity after minimally-invasive or robot-assisted esophagectomy.

#### **2. Results**

52 patients were included in the study. The demographic and clinicopathological data of these patients are presented in Table 1. Of the 52 patients in one patient the tumor could not be resected.

Twenty-eight patients (54%) were seen by the nutritional expert at our department preoperatively. All patients were seen by the nutritional expert postoperatively. Shakes with high concentration of proteins were advised to all patients preoperatively. Ten patients (19.2%) needed intravenous nutrition preoperatively.


**Table 1.** Demographic and clinicopathological characteristics of patient cohort.

Mean albumin level was 35 ± 4.7 g/dL, mean cholesterol level (199.83 ± 48.18) mg/dL. In 6 patients (11.5%) a totally minimal invasive approach was not possible, two patients had to be converted to open surgery (conversion rate 3.8%).

Mean duration of surgery was 445 min (range 303–770).

#### *2.1. Skeletal Muscle Index*

The skeletal muscle index (SMI) ranged from 29.7 to 62.6 cm2/m2 at time of diagnosis and from 31.9 to 62.5 cm2/m2 before surgery (mean values (47.7 <sup>±</sup> 8.6 cm2/m2 and 42.1 <sup>±</sup> 7.0 cm2/m2 respectively)).

Applying the cut-offs for sarcopenia used by Prado et al [34]. Of the patients, 54.3% were sarcopenic at time of diagnosis and 87.5% before surgery.

*Cancers* **2020**, *12*, 3409

A moderate correlation was detected between BMI and SMI (r = 0.69855, *p* < 0.001, Pearson's correlation coefficient).

Results regarding SMI, weight and BMI are summarized in Table 2. BMI at diagnosis is missing in 10 patients.


**Table 2.** Skeletal muscle index (SMI), weight, and body mass index (BMI) at different time points.

#### *2.2. Short Term Postoperative Outcome*

Short term postoperative outcome was defined as complications during hospital stay and complication associated mortality.

The median length of hospital stay was 17.5 days (range 6–114 days), the length of stay at the Intensive Care Unit ranged from one to 77 days (median 17.5 days).

One patient died during the first 30 days after surgery (30-day mortality 1.9%). The 90-day mortality in our cohort did not differ from this value because there were no other patients who died within 90 days after surgery.

Four patients died because of complications of surgery, three of those after 90 days after surgery (complication associated mortality 7.69%).

The general complication rate was 59.6% (31 patients in total). Twenty-six patients had surgical complications (50%), of those 16 patients suffered from anastomosis insufficiency (30.8% anastomosis insufficiency rate). Twenty-three patients suffered from medical complications (44.2%), including cardiac complications (21.2%) and pulmonary complications (32.7%).

There was no correlation between SMI before neoadjuvant therapy nor before surgery with short-term postoperative outcome (*p* = 0.5365, *p* = 0.3530 respectively, *t*-test).

The BMI also had no influence on postoperative morbidity: *p* = 0.4228 (BMI before diagnosis), *p* = 0.1673 (BMI before neoadjuvant treatment), *p* = 0.2810 (BMI before surgery), *t*-test. Regarding cholesterol level before surgery there was a trend for a higher risk of complications with lower cholesterol levels ((207 ± 55) mg/dL vs. (191 ± 38) mg/dL, *p* = 0.0846, *t*-test).

Other factors such as patient age (*p* = 0.4361, *t*-test), sex (*p* = 0.6872, Fisher's exact test), history of cardiac disease (*p* = 0.8615, Chi-square-test), history of pulmonary disease (*p* = 1.0, Fisher's exact test), tumor stage ((y)pT: *p* = 0.7241, (y)pN: *p* = 1.0, Cochran–Armitag-test), and preoperative albumin levels (*p* = 0.3747, *t*-test) had no statistically significant influence on short-term postoperative outcome. The duration of surgery however correlated with the incidence of complications (*p* = 0.0034, *t*-test) as patients who had longer times of surgery showed a higher risk for complications (466 vs. 392 min). Patients who had another type of cancer before developing esophageal cancer also suffered more often from complications (*p* = 0.0049, Chi-square-test).

#### **3. Discussion**

Our findings suggest that a low preoperative SMI does not necessarily predict a poor postoperative outcome in esophageal cancer patients after esophagectomy. Neither BMI before neoadjuvant therapy nor before surgery shows a correlation with postoperative outcome in our patients' collective. Only one patient presented with a BMI <20 kg/m2. This might be due to patient selection. We included in our analysis only patients who were eligible for a thoracoabdominal surgical approach, which means that they were in a sufficient general status of health to tolerate one-lung-ventilation. We also checked all patients for nutritional deficiencies and sent them to a nutritional expert, which explains why the body weight difference between the time of diagnosis and the time of surgery is quite low and which could also explain the missing impact of BMI and SMI in our patients' collective. Often it is during neoadjuvant treatment that patients lose weight and muscle mass [35]. In our patients' collective the mean BMI before and after neoadjuvant treatment did not differ (25.87 and 25.34 kg/m2) and most of the patients (70.95%) did not have a significant reduction of SMI (70.95% had less than 5% reduction). Of the patients, formula increase of SMI during neoadjuvant treatment. This shows that an improvement in patients' general status during neoadjuvant treatment is feasible.

As the number of patients in this cohort was relatively low it is also possible that in a greater cohort an effect of preoperative SMI on postoperative outcome might also be greater. Nevertheless, we could not see even a trend to significant correlations between SMI and complications and also there are other studies that showed similar results concerning this topic [36,37].

The level of cholesterol seems to be associated with the short-term postoperative outcome in our patients' collective, which underlines the importance of the nutritional status. The results are not statistically significant but this could be due to the relatively low number of patients included.

Malnutrition in esophageal cancer patients is caused by dysphagia and a reduced food intake but also by cancer cachexia, which is induced by systemic inflammation. Cachexia leads to loss of body weight, body fat, and skeletal muscle mass. The molecular mechanisms of cachexia are still not fully understood, additionally there exists no consistent definition [38].

It should be mentioned that only one third of our patients had a squamous cell carcinoma (SCC). Patients with SCC normally present with a more advanced malnutrition than patients with an adenocarcinoma of the esophagus as malnutrition is often already present before diagnosis of the tumor.

Different studies showed significant relations between sarcopenia and poor outcome after surgery such as the study by Elliot et al. that showed that sarcopenia is significantly associated with major complications after surgery or the study by Järvinen et al. that showed a correlation with worse overall survival [14,25,35,39–41].

Grotenhuis et al. and Siegal et al. on the other hand could not find any correlations between sarcopenia and postoperative complications or survival rates [36,37]. Nakashima et al. could find a significant correlation between sarcopenia and higher anastomotic leakage rates and also between sarcopenia and worse overall survival but these significant results were only present in a subgroup of patients older than 65 years [42].

The definition for sarcopenia is not consistent. Some studies define their own cut-offs. We used the most common cut-offs suggested by Prado et al [34], resulting in a rate of sarcopenia of 87.5% after neoadjuvant therapy. There is a huge variation in the prevalence of sarcopenia in esophageal cancer patients in the literature, varying from 16% to 75% [43].

Nevertheless, the latest meta-analysis by Papaconstantinou et al. showed a significant increase in overall morbidity, respiratory complications and anastomotic leaks in esophageal cancer patients with sarcopenia after esophagectomy. There were no statistically significant differences in overall mortality or Clavien–Dindo grade III or greater complications between patients with low and patients with normal SMI [44].

Although our findings contradict many studies that showed significant correlations between sarcopenia and higher complications or lower over-all survival there are other studies that support our findings.

There are some other important aspects of our patient cohort that should be illuminated. It has already been proven that prehabiliation and preoperative optimization of nutrition in cancer patients leads to better outcome and better quality of life after cancer surgery [45,46]. Nutrition goals include an adequate nutritional intake to prevent loss of muscle mass, modulate inflammation and the immune response, optimize glucose control, reduce the hypermetabolic response to surgery, and provide nutrients to optimize wound and anastomotic healing and recovery [18–20,47–50].

As mentioned above our cancer patients are provided with professional advice by specialized nutritional counsellors before and after surgery to improve their fitness and nutritional. This includes provision of protein shakes, dietary plans and even parenteral feeding if necessary.

It also has been shown that standardized protocols like ERAS protocols for surgical cancer patients lead to shorter in-hospital stays, decreased complications and longer overall survival [31–33,51]. Therefore, our clinic uses standardized pathways for esophageal cancer patients to improve postoperative outcomes of esophageal cancer patients.

This aspect of our preoperative treatment might be the reason for the absence of significant increase of postoperative morbidity or mortality in sarcopenic patients. One aspect that should be evaluated in further trials is, whether the outcome of esophageal cancer patients can be improved by introducing prehabilitation programs.

The presented study has some limitations. It is retrospective in nature. This increases the risk for systemic errors and selection bias. The number of patients included in this study (*n* = 52) is relatively low which may affect its statistical power, especially in regard to complication rates and specific complications.

#### **4. Materials and Methods**

#### *4.1. Patient Selection and Study Design*

The Surgical Department of the University Hospital Mannheim of Heidelberg University is a certified centre for esophageal surgery. A detailed clinico-pathologic database is prospectively maintained for all patients with esophageal cancer since January 2018 including data on short-term postoperative outcome. In this analysis we included patients having been operated between January 2018 and July 2019 in the Surgical Department of the University Hospital Mannheim of Heidelberg University. The local ethical committee "Ethikkomission II, University of Heidelberg" gave approval for the analysis (ethic code 2020-803R) and all patients gave written informed consent.

There were 52 patients who underwent surgery for treatment of esophageal cancer in whom CT scans were sufficient to determine the skeletal muscle index. Of the 52 patients, 48 underwent neoadjuvant chemotherapy before surgery (FLOT regimen).

The skeletal muscle index was measured using computer tomographic images at the time of diagnosis as well as before surgery.

Out of the 52 patients 13 did not have two usable computer tomographies either because they did not undergo neoadjuvant treatment or because imaging was not done in-hospital and was not sufficient for measurement of the SMI.

The skeletal muscle index (SMI) was calculated as the cross-sectional area of the total skeletal muscle volume (cm2) at L3.

At the level of the third lumbar vertebra (L3) we measured the area (cm2) of the left and right psoas major muscles, the side abdominal muscles, rectus abdominis muscles, erector spinae muscles, and quadratus lumborum muscles using "syngo.share view diagnostic" (Siemens Healthineers, Erlangen, Germany). Two adjacent axial images within the same series were selected, and total muscle cross-sectional area (cm2) at L3 was determined and averaged for each patient (1).

$$\text{SMI} \left( cm^2 / m^2 \right) = \text{Lenn tisse } area\_{L.3} \text{ (} cm^2 \text{)} \text{height}^2 \text{ (} m^2 \text{)}\tag{1}$$

Low SMI was defined as <52.4 cm2/m2 for male patients and as <38.5 cm2/m<sup>2</sup> for female patients according to current literature [44].

#### *4.2. Preoperative Treatment*

All patients are seen by a surgeon specialized in upper GI surgery before admission to hospital. Education in nutrition as well as recommendation for daily physical activity and cessation of smoking is given during the consultation. Patients with nutritional deficiency are additionally sent to a nutritional consulting at the University Hospital Mannheim to compensate nutritional deficiencies before surgery.

Body weight, BMI, serum albumin levels, dysphagia, and subjective well-being are measured regularly by specialized nutritional consults. If the patients are suffering from dysphagia, weight loss or low BMI their nutritional status is treated by prescribing protein shakes, dietary plans and, if necessary parenteral feeding.

#### *4.3. Neoadjvuant and Perioperative Therapy*

Patients with locally advanced tumors (>cT2N0) receive neoadjuvant or perioperative treatment according to the FLOT protocol for adenocarcinomas and according to the CROSS protocol for squamous cell carcinomas [52].

After completion of preoperative treatment patients are reevaluated including a risk assessment and preoperative blood management.

#### *4.4. Surgery*

Surgery is performed by minimally invasive procedure, either laparoscopic and thoracoscopic approach or laparoscopic abdominal approach and robot-assisted thoracal approach. The decision for a robotic approach is taken due to availability of the DaVinci robot. Resection is done according to Ivor–Lewis procedure (right-thoracic approach). Lymphadenectomy is performed according to current guidelines including mediastinal lymphadenectomy and abdominal lymphadenectomy (D2 Lymphadenectomy). Reconstruction is done by gastric conduit and side-to-side esophagogastrostomy.

#### *4.5. Postoperative Management*

To ensure a standardized treatment of patients after surgery, the postoperative hospital management follows an in-house pathway which defines postoperative nutrition, mobilization, drainage management, laboratory tests, and dismissal after esophagectomy.

Directly after surgery patients stay in an Intermediate Care Unit (IMCU) for at least two days to ensure a balanced fluid management and to detect early postoperative complications as soon as possible.

Patients are encouraged to engage in light physical activity beginning on the day of surgery. Breathing exercises are started on the first postoperative day and should be done at least every hour for five to ten minutes.

Increased patient self-sufficiency in the general ward helps to improve mobilization and nutrition. Nutritional counselling is performed during hospital treatment to determine nutritional deficiencies and to arrange nutritional support after the dismissal.

#### *4.6. Postoperative Complications*

Postoperative complications were investigated separately as surgical and medical complications as well as complications in total. To measure complication rates the Comprehensive Complication Index (CCI) was used as well as Clavien–Dindo classification.

The Clavien–Dindo classification is a commonly used tool with which usually the single most severe complication occurring in a patient during a given episode of care is reported [53].

The CCI is a relatively new tool to measure complications based on the Clavien–Dindo classification. In contrast to the Clavien–Dindo classification the CCI integrates all postoperative complications with their respective severities, on a scale ranging from 0 (no burden from complications) to 100 (death). The CCI is calculated by adding the weights of different complications (wC) and using a defined formula (2) [54].

$$\text{CCI} = I \sqrt{(w\text{C}\_1 + w\text{C}\_2 \dots \dots + w\text{C}\_x)} \text{J} 24.7 \tag{2}$$

#### *4.7. Statistical Analysis*

For quantitative, approximately normally distributed variables, the mean and standard deviation have been calculated. Qualitative variables are given as absolute and relative frequencies. The median, together with range, are presented for skewed or ordinally scaled parameters. A Student's *t*-test was used for comparing approximately normally distributed quantitative variables. The Cochran–Armitage test for trend was used to determine the influence of (y)pT and (y)pN categories on major postoperative complications. For qualitative variables, a χ2-test or Fisher exact test was performed, as appropriate. To determine a correlation between BMI and SMI the Pearson correlation coefficient was determined. All statistical tests for the comparison of 2 groups were two-tailed. In general, a test result was considered statistically significant if *p* < 0.05. All statistical analyses were performed using the SAS statistical analysis software, release 9.4 (SAS Institute Inc., Cary, NC, USA).

#### **5. Conclusions**

Our findings suggest that the nutritional status and physical fitness of esophageal cancer patients consist of more than BMI and skeletal muscle mass but that albumin levels, cardiac health and subjective well-being, as a consequence of adequate nutrition preoperatively, are important factors for a good postoperative outcome.

In the era of multimodal approach to esophageal cancer, treatment should consist not only of oncological treatment and surgery but also of a more systemic approach including patient education, physical exercise, and nutritional support. The period between diagnosis and surgery should be used to reduce cachexia and to better the general status of health.

Therefor prehabilitation programs as well as nutritional plans should be evaluated in clinical studies with the aim to give a consistent recommendation for patients with esophageal cancer [55].

**Author Contributions:** Conceptualization, M.O., C.R., and S.B.; methodology, S.B.; software, C.W.; validation, S.B.; formal analysis, C.W., S.B.; investigation, H.L., L.E.; data curation, L.E., J.G.; writing—original draft preparation, J.G.; writing—review and editing, S.B.; supervision, M.O., C.R.; project administration, S.B.; 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*
