**Immediate Prosthetic Breast Reconstruction after Nipple-Sparing Mastectomy: Traditional Subpectoral Technique versus Direct-to-Implant Prepectoral Reconstruction without Acellular Dermal Matrix**

**Gianluca Franceschini 1, \* ,† , Lorenzo Scardina 1,† , Alba Di Leone 1 , Daniela Andreina Terribile 1 , Alejandro Martin Sanchez 1 , Stefano Magno 1 , Sabatino D'Archi 1 , Antonio Franco 1 , Elena Jane Mason 1 , Beatrice Carnassale 1 , Federica Murando 1 , Armando Orlandi 2 , Liliana Barone Adesi 3 , Giuseppe Visconti 3 , Marzia Salgarello <sup>3</sup> and Riccardo Masetti 1**


**Abstract:** Background: The aim of this study was to compare outcomes of immediate prosthetic breast reconstruction (IPBR) using traditional submuscular (SM) positioning of implants versus prepectoral (PP) positioning of micropolyurethane-foam-coated implants (microthane) without further coverage. Methods: We retrospectively reviewed the medical records of breast cancer patients treated by nipple-sparing mastectomy (NSM) and IPBR in our institution during the two-year period from January 2018 to December 2019. Patients were divided into two groups based on the plane of implant placement: SM versus PP. Results: 177 patients who received IPBR after NSM were included in the study; implants were positioned in a SM plane in 95 patients and in a PP plane in 82 patients. The two cohorts were similar for mean age (44 years and 47 years in the SM and PP groups, respectively) and follow-up (20 months and 16 months, respectively). The mean operative time was 70 min shorter in the PP group. No significant differences were observed in length of hospital stay or overall major complication rates. Statistically significant advantages were observed in the PP group in terms of aesthetic results, chronic pain, shoulder dysfunction, and skin sensibility (*p* < 0.05), as well as a trend of better outcomes for sports activity and sexual/relationship life. Cost analysis revealed that PP-IPBR was also economically advantageous over SM-IPBR. Conclusions: Our preliminary experience seems to confirm that PP positioning of a polyurethane-coated implant is a safe, reliable and effective method to perform IPBR after NSM.

**Keywords:** breast cancer; nipple-sparing mastectomy; immediate breast reconstruction; acellular dermal matrix (ADM); aesthetic and oncological outcomes; quality of life

**Citation:** Franceschini, G.; Scardina, L.; Di Leone, A.; Terribile, D.A.; Sanchez, A.M.; Magno, S.; D'Archi, S.; Franco, A.; Mason, E.J.; Carnassale, B.; et al. Immediate Prosthetic Breast Reconstruction after Nipple-Sparing Mastectomy: Traditional Subpectoral Technique versus Direct-to-Implant Prepectoral Reconstruction without Acellular Dermal Matrix. *J. Pers. Med.* **2021**, *11*, 153. https://doi.org/ 10.3390/jpm11020153

Academic Editor: Hisham Fansa

Received: 25 December 2020 Accepted: 19 February 2021 Published: 22 February 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/).

#### **1. Introduction**

Immediate prosthetic breast reconstruction (IPBR) is considered as an integral part of the surgical treatment of patients undergoing nipple-sparing mastectomy (NSM) for breast cancer, as it positively affects psychological health, sexuality, body image, and self-esteem.

Traditionally, IPBR has been performed by placement of the prosthetic implant in a submuscular (SM) pocket created beneath the pectoralis major muscle, in order to protect the integrity of the implant and reduce its visibility and palpability [1,2]. Although this technique has shown increasingly good results, it still yields a higher risk of undesirable outcomes such as significant postoperative pain, injury-induced muscular deficit, breast animation deformity, lateral deviation of the breast mound with poor inframammary fold definition, and insufficient lower pole fullness [3,4].

In recent years, placement of the implant in a prepectoral (PP) plane has been increasingly employed. When this technique is performed, the implant is usually covered with an acellular dermal matrix (ADM) to shield it in the subcutaneous space underneath the skin flaps; however, the use of ADM has been reported to increase risks of seroma, infection, and skin/nipple-areola complex (NAC) necrosis, and associated with higher medical costs [1]. To limit these inconveniences, the use of implants with a special micropolyurethane-foamcoated shell surface (microthane) that does not require ADM coverage has recently been proposed [2,5].

The aim of this study was to compare outcomes between traditional SM-IPBR and a PP technique using microthane implants without ADMs in patients undergoing NSM.

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

After approval from the Institutional Review Board of our hospital, a retrospective review of the medical records of breast cancer patients who underwent NSM followed by IPBR over the two-year period of January 2018–December 2019 was performed. Patients treated before January 2018 were not enrolled because before that date, PP-IPBR in our institution was routinely performed with ADMs, which would have added heterogeneity to our population.

Patients were divided into two cohorts based on the site of implant placement: in SM-IPBR, anatomical textured implants were positioned in the subpectoral pocket according to a previously described standardized technique, while in PP-IPBR, a definitive Polytech implant with a micropolyurethane-foam-coated shell surface was placed in the subcutaneous plane [5,6].

#### *2.1. Operative Protocol and Surgical Technique*

A complete preoperative workup including clinical assessment, ultrasonography, mammography, breast MRI, and disease staging was performed in all patients; surgical planning was always discussed in a multidisciplinary dedicated surgery board. Common indications to NSM included large tumor-to-breast size, inability to obtain clear surgical margins with breast-conserving surgery, extensive or multicentric disease, contraindications to adjuvant radiotherapy, and patient preference; absolute contraindications to NSM with both types of reconstruction were inflammatory carcinoma, a locally advanced tumor infiltrating the skin or NAC, and previous radiotherapy. Obesity (BMI > 30 kg/m<sup>2</sup> ), large breasts with severe ptosis, and active smoking were considered as relative contraindications due to the increased risk of skin or NAC necrosis, breast asymmetry, and nipple displacement [2–6]. Bilateral NSM was performed in patients with a bilateral breast tumor or in women with unilateral disease and a high risk of contralateral breast cancer, such as BRCA mutation carriers.

A specific algorithm shared with the plastic surgeons, based on anamnestic, morphological, functional, and oncological criteria, was used to define the most appropriate reconstruction technique [7,8]. The Rancati classification, based on digital mammographic imaging, was used to predict thickness of post-mastectomy skin flaps [9].

In the vast majority of cases, NSM was carried out through a radial incision on the external quadrants; axillary or inframammary crease incisions were used only in selected cases. Skin flaps and NAC were progressively elevated from glandular tissue. The entire gland was then separated from the muscular plane and removed, preserving the superficial pectoralis fascia. An accurate circumferential palpation of the surgical cavity after removal of the gland was always performed to rule out the possibility of residual breast tissue. Intraoperative pathology evaluation of retroareolar tissue was performed in all cases to confirm secure margins. The removed gland was always weighed to better determine the subsequent reconstruction volume.

The final decision on the type of reconstructive technique (SM versus PP) was made in the operating room based on flap thickness and perfusion assessment [2,10]. Skin-flap thickness was measured using pliers, and perfusion was assessed using indocyanine green dye fluoroangiography and a photodynamic eye (PDE) imaging system (Figures 1 and 2).

**Figure 1.** A case of nipple-sparing mastectomy and direct-to-implant prepectoral reconstruction without acellular dermal matrix. (**a**,**b**) Preoperative pictures of a 43-year-old right-breast cancer patient for whom right nipple-sparing mastectomy and direct-to-implant prepectoral reconstruction without acellular dermal matrix were planned. (**c**,**d**) Six-month postoperative pictures after right nipple-sparing mastectomy through a radial lateral incision (mastectomy specimen 190 g) and prepectoral reconstruction using a definitive anatomical implant (Polytech 30746, 295cc) with a micropolyurethane-foam-coated shell surface, placed in the subcutaneous plane.

A single-stage SM reconstruction was performed using total coverage of the implant beneath the pectoralis major and serratus anterior [7]; PP-IPBR was realized with the placement of the prosthesis into the same anatomical space of the excised mammary gland [2,5]; textured implants were used for SM-IPBR and Polytech implants with a micropolyurethanefoam-coated shell surface for PP-IBPR [2,5]; and a contralateral procedure to achieve better symmetry was performed when deemed necessary [10,11].

We chose to position a prepectoral implant every time we had good soft-tissue coverage after mastectomy (defined as flap thickness of at least 1 cm and good perfusion with indocyanine green dye fluoroangiography and the photodynamic eye imaging system). In SM-IBPR, we performed a submuscular–subfascial pocket dissection, which allows, with time, a good ptosis. In these cases, any exceeding skin can usually be nicely managed by intraoperative redraping with taping. In SM-IBPR, reduction–augmentation procedures were performed as previously reported.

Two Jackson Pratt drains were always placed in the reconstructive space, usually left in place at the time of hospital discharge and later removed when the amount of fluid

collected over 24 h was <30 mL. Patients received levofloxacin at a dosage of 500 mg every 12 h until drain removal and were advised to continue wearing a sports bra for 1 month.

The operative time (from incision to the end of skin suture) and length of hospitalization were recorded.

#### *2.2. Clinical Assessment and Statistical Analysis*

Patients were assessed at weekly intervals during the first month and then every 6 months by breast surgeons, plastic surgeons, and oncologists.

Major complications (requiring surgical revision), loco-regional recurrences (defined as local recurrence if involving the ipsilateral skin flap, chest wall, or NAC; or as regional recurrence if involving ipsilateral axillary, internal mammary, or supraclavicular nodes), cosmetic outcomes, quality of life, and economic costs were assessed in all patients.

An automated breast volume scanner (ABVS), a dedicated imaging system that can obtain full-field high-resolution views of skin flaps, was used to better evaluate possible local recurrence in the usually thicker skin flaps of patients with PP-IPBR [10].

The "QOL assessment PRO" is a questionnaire created through a multidisciplinary effort by all specialists working in the Breast Unit of Fondazione Policlinico Universitario Agostino Gemelli IRCCS. It was developed based on the experiences reported in the literature, and has been proficiently employed in our center for several years [12–17]. The questionnaire condenses in seven simple questions the essential patient-reported outcomes (PROs) involving pain, arm motility, aesthetic satisfaction, and general quality of life (QOL), and is therefore a practical tool that in our experience gives results more agreeable to patients than BREAST-Q, and increases their compliance to participate in the study [18]. The QOL assessment PRO was administered six months after surgery via a telephone call by a member of hospital staff, and consisted of five close-ended questions (requiring a yes/no answer) and two scoring questions (requiring a score between 0 and 5 as an answer) (Table 1).

#### **Table 1.** QOL assessment PRO survey.

#### **Smart QoL Assessment**

• Quality of life

• What score would you give to your pain, from 0 (no pain) to 5 (very intense)?

• Is arm motility impaired after surgery? (YES/NO)

• Did you do sports before surgery? (YES/NO)


Satisfaction

• How would you evaluate, from 1 (poor) to 5 (excellent), the aesthetic result of your operation?

Psychological and relational field

• Did the operation compromise your womanhood, sexuality, or relationship life? (YES/NO)

Abbreviations: QOL = quality of life; PRO = patient-reported outcomes.

Results were expressed as means with associated median and range. Statistical analysis was performed using SPSS (version 24.0 for Windows). A Fisher exact test was used for comparison of categorical variables. A *p*-value equal to or less than 0.05 was considered statistically significant. A cost analysis was performed according to a standardized method [19].

#### **3. Results**

Over the two-year study period from January 2018 to December 2019, 177 breast cancer patients with IPBR after NSM were included. SM-IPBR was performed in 95 cases, while PP-IPBR was performed in 82 cases. Patient characteristics are reported in Table 2. Ptosis degree, Rancati score, and intraoperative flap thickness assessment were decisive in determining the kind of reconstruction performed, and therefore differed significantly between the PP and SM group. The remaining aspects were similar in both populations. Adjuvant radiotherapy did not affect aesthetic and oncological outcomes.


**Table 2.** Patient characteristics.

Abbreviations: FUP = follow-up; BRCA = breast cancer gene. Statistically significant *p* values (<0.05) are marked in bold.

The mean ages were 44 (28–73) and 47 (27–73) years respectively. After unilateral NSM, a simultaneous contralateral symmetrization procedure was deemed necessary and carried out in 44/44 (100%) patients of the SM group and in 2/55 (3.6%) patients of the PP group. The type of surgical treatment is summarized in Table 3.


**Table 3.** Type of surgical treatment.

Abbreviations: NSM = nipple-sparing mastectomy; IPBR = immediate prosthetic breast reconstruction. Statistically significant *p* values (<0.05) are marked in bold.

#### *3.1. Duration of Surgery and of Hospitalization*

For patients undergoing unilateral NSM and IPBR, the mean total operative time was 319 min in the SM group and 247 min in the PP group; for patients undergoing bilateral NSM, it was 368 min and 306 min, respectively.

The longest surgery (510 min) was for a patient who underwent a transaxillary bilateral mastectomy with sentinel node biopsy, axillary dissection, and bilateral SM reconstruction. Operative times are summarized in Table 4. Length of hospitalization did not significantly differ between the two populations.

#### **Table 4.** Operative time (minutes).


Statistically significant *p* values (<0.05) are marked in bold.

#### *3.2. Perioperative and Oncological Outcomes*

Median follow-up was similar: 20 (6–28) months in the SM group and 16 (5–28) months in the PP group. There was no significant difference in length of stay, overall major complication rates, and oncological outcomes between the two reconstructive cohorts.

Implant loss caused by infection was observed in one patient in the SM group (1.05%) and one patient in the PP group (1.2%). One patient in the PP group (1.2%) developed a full-thickness NAC necrosis that required secondary excision.

During follow-up, NAC recurrence occurred in one patient of the SM group (1.05%), while in the PP group, no local relapse was observed. Regional recurrences occurred in 2/95 (2.1%) patients in the SM group and in 1/82 patients (1.2%) in the PP cohort.

Regarding disease-free survival, one patient in the SM group with triple negative breast cancer developed brain metastases six months after surgery.

#### *3.3. Cosmetic Outcomes and Health-Related Quality of Life*

A total of 126/177 patients completed our survey assessing their postoperative quality of life (64.2% and 78%, respectively, for the SM and PP groups).

Statistically significant (*p* < 0.05) advantages in terms of cosmetic results, chronic pain, shoulder dysfunction, and skin sensibility were observed in the PP group.

A not statistically significant difference in favor of the PP group was shown for sports activity and sexual/relationship life (Table 5).


**Table 5.** QOL assessment PRO survey replies.

Statistically significant *p* values (<0.05) are marked in bold.

#### *3.4. Economic Performance*

Whenever a surgical procedure is performed, different resources (including personnel, equipment, facilities, time, and materials) are utilized. A cost analysis was performed according to a standardized method and direct cost comparison [19]. The analysis showed better economic performances in the PP group due to shorter operative times, less-frequent need of contralateral breast symmetrization, and less-frequent use of contralateral implants. The average savings with PP-IPBR were EUR 1503 for unilateral NSMs and EUR 1568 for bilateral procedures (Table 6).


**Table 6.** Economic analysis.

Abbreviations: PP-IPBR = prepectoral immediate prosthetic breast reconstruction; SM-IPBR = submuscular immediate prosthetic breast reconstruction; OR = operating room; NSM = nipple-sparing mastectomy; ADM = acellular dermal matrix. Statistically significant *p* values (<0.05) are marked in bold.

#### **4. Discussion**

In our institution, we offer IPBR to all patients undergoing NSM. For many years, we have used only SM placement of the implants, but since 2016, we also started to perform PP-IPBR in selected cases, initially with ADM coverage and only recently without the use of matrices [2,6,20].

PP placement of the prosthesis into the space of the excised mammary gland allows a more natural breast appearance with a more harmonious breast slope and ptosis [21–23]. It also allows, in most cases of unilateral NSM, the avoidance of symmetrization procedures on the contralateral breast [24,25]. In our experience, a symmetrization procedure was performed for 44/44 (100%) patients in the SM group, compared to only 2/55 (3.6%) cases in the PP group with polyurethane-covered implants.

Initially, when performing PP-IPBR, we used ADM coverage of the implant. ADMs are biologic scaffolds of human, bovine, or porcine origin that lack immunogenic epitopes and are therefore easily revascularized and integrated into host tissue without encapsulation or contracture [23–26].

The use of ADM, however, may be hampered by surgical and economic issues. Some authors reported higher medical costs, with a variable additional expense between USD 2100 and USD 3400, depending on the size of the dermal sheet utilized [17,27].

For these reasons, in January 2018 we started to perform PP-IPBR using a Polytech implant with a micropolyurethane-foam-coated shell surface (microthane) that does not require further ADM coverage [2,5]. The 1.4 mm micropolyurethane sponge coating is reabsorbed by the body and contributes to form an ideal capsule that protects the implant and reduces capsular contracture, resulting in softer and more natural-appearing breasts. Furthermore, the extremely adherent texture of this implant reduces the risks of rotation and displacement, and consequently the possible need for revision surgery [5].

Careful patient selection and surgical conduct are mandatory to perform PP-IPBR successfully. This technique should be considered only for patients in which adequate thickness and perfusion of skin flaps can be ensured during mastectomy [2,24,28].

To minimize the risk of learning-curve-related complications and technical problems, we considered exclusion criteria of BMI > 30kg/m<sup>2</sup> , oversized breasts, ptosis of grade >2, obese patients, heavy smokers, and previous radiation therapy [24,29].

Regarding the surgical conduct, lateral–radial incisions or axillary or inframammary crease incisions are preferable in order to better preserve vascular integrity of the NAC [20,29,30]; skin flaps of adequate thickness should be separated from the mammary gland using blunt dissection and preserving medial perforators, and real-time skinperfusion testing with a fluorescence imaging system should be performed intraoperatively to assess skin-flap viability with immediate resection of potential ischemic tissues. Choice of implant size and shape should be based on evaluation of the breast and chest-wall

conformation and accurate weight of the surgical specimen (in this regard, we recommend using fill volumes similar to those of the removed gland).

With proper patient and implant selection and careful surgical conduct, PP-IPBR can be performed with results similar to SM-IPBR in terms of postoperative complication rates and oncologic safety [1,27,30,31]. In our series, there were no statistically significant differences in terms of implant failure and local, regional, or systemic recurrence between the two groups. We observed only two cases of major complications that led to implant loss: one case of infection in the PP group, and one in the SM group. One patient in the PP group developed NAC necrosis. We classified this complication as minor because it required no surgical revision and was treated successfully in outpatient regime, as the necrosis involved only a small portion of the NAC and was not full thickness.

Regarding patient quality of life, we observed statistically significant improvements in aesthetic results, chronic pain, shoulder dysfunction, and skin sensibility (*p* < 0.05) in the PP group, and a trend of better outcomes (even if statistically not significant) regarding sports activity and sexual/relationship life in this group.

These better results are probably explained by the avoidance of chest-wall musculature manipulation in PP-IPBR [1,2,5].

PP-IPBR significantly reduces operative time as there is no need for submuscular pocket creation, and, in most cases, for contralateral breast symmetrization. When using microthane-coated implants, operative time is further reduced by the avoidance of ADM coverage [2,5,27].

In our series, this shorter operative time, coupled with the reduced need for contralateral implants, generated an average saving of EUR 1500 for unilateral procedures; this saving significantly increased when using a Polytech implant, as the costs of ADM coverage are also avoided (the cost of a 30 × 20 cm sheet of ADM in our hospital is EUR 4056). Furthermore, because PP reconstruction averts the issues related to pectoralis major muscle manipulation, it also minimizes postoperative costs of painkillers and postoperative physiotherapy, with additional benefit for the healthcare system [32–35].

#### **5. Conclusions**

Our study presents several limitations, as it is a retrospective unicentric analysis with a relatively limited duration of follow-up, and may include a small selection bias, as PP-IPBR without ADM has been adopted in our institution only recently, and therefore grants less expertise and more potential for technical mistakes than SM-IPBR. However, this work provides encouraging preliminary data on the safety and efficacy of PP positioning of microthane-coated implants without ADM in patients undergoing NSM with IPBR.

PP-IPBR can represent a valid alternative to traditional IPBR, improving outcomes and patient quality of life; it is easier to perform, reduces operative time, and minimizes complications related to manipulation of the pectoralis major muscle, while also contributing to the containment of costs.

Careful patient selection, adequate surgical experience, and repetitive practice of specific tasks are mandatory to optimize the outcomes and reduce the risk of minor and major complications. Further prospective trials with a larger number of patients and a longer follow-up are necessary to draw more validated conclusions.

**Author Contributions:** Conceptualization, G.F. and L.S.; methodology, G.F., A.D.L., L.S. and A.O.; software, A.F. and S.M.; validation, G.F., R.M., D.A.T.; formal analysis, E.J.M. and B.C.; investigation, A.M.S.; resources, L.B.A. and G.V.; data curation, S.D. and F.M.; G.F. and L.S.; writing—review and editing, G.F. and R.M.; visualization, G.F. and G.V.; supervision, G.F., R.M. and M.S., G.F. and L.S. contributed equally to this work. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board and Ethics Committee of Fondazione Policlinico Universitario Agostino Gemelli IRCCS; Università Cattolica del Sacro Cuore, Rome, Italy.

**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.

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

#### **References**


### *Article* **The Assisi Think Tank Meeting Breast Large Database for Standardized Data Collection in Breast Cancer—ATTM.BLADE**

**Fabio Marazzi 1 , Valeria Masiello 1, \* , Carlotta Masciocchi 2 , Mara Merluzzi 3 , Simonetta Saldi 4 , Paolo Belli 5 , Luca Boldrini 1,6 , Nikola Dino Capocchiano 6 , Alba Di Leone 7 , Stefano Magno 7 , Elisa Meldolesi 1 , Francesca Moschella 7 , Antonino Mulé 8 , Daniela Smaniotto 1,6 , Daniela Andreina Terribile 8 , Luca Tagliaferri 1 , Gianluca Franceschini 6,7 , Maria Antonietta Gambacorta 1,6 , Riccardo Masetti 6,7 , Vincenzo Valentini 1,6 , Philip M. P. Poortmans 9,10 and Cynthia Aristei 11**

	- Istituto di Radiologia, Università Cattolica del Sacro Cuore, 00168 Roma, Italy; nikoladino.capocchiano@unicatt.it (N.D.C.); gianluca.franceschini@policlinicogemelli.it (G.F.); riccardo.masetti@policlinicogemelli.it (R.M.)

**Abstract: Background:** During the 2016 Assisi Think Tank Meeting (ATTM) on breast cancer, the panel of experts proposed developing a validated system, based on rapid learning health care (RLHC) principles, to standardize inter-center data collection and promote personalized treatments for breast cancer. **Material and Methods:** The seven-step *Breast LArge DatabasE (BLADE)* project included data collection, analysis, application, and evaluation on a data-sharing platform. The multidisciplinary team developed a consensus-based ontology of validated variables with over 80% agreement. This English-language ontology constituted a breast cancer library with seven knowledge domains: baseline, primary systemic therapy, surgery, adjuvant systemic therapies, radiation therapy, followup, and toxicity. The library was uploaded to the *BLADE* domain. The safety of data encryption and preservation was tested according to General Data Protection Regulation (GDPR) guidelines on data from 15 clinical charts. The system was validated on 64 patients who had undergone post-mastectomy radiation therapy. In October 2018, the *BLADE* system was approved by the Ethical Committee of Fondazione Policlinico Gemelli IRCCS, Rome, Italy (Protocol No. 0043996/18). **Results:** From June 2016 to July 2019, the multidisciplinary team completed the work plan. An ontology of 218 validated variables was uploaded to the *BLADE* domain. The GDPR safety test confirmed encryption and data

**Citation:** Marazzi, F.; Masiello, V.; Masciocchi, C.; Merluzzi, M.; Saldi, S.; Belli, P.; Boldrini, L.; Capocchiano, N.D.; Di Leone, A.; Magno, S.; et al. The Assisi Think Tank Meeting Breast Large Database for Standardized Data Collection in Breast Cancer—ATTM.BLADE. *J. Pers. Med.* **2021**, *11*, 143. https://doi.org/ 10.3390/jpm11020143

6

Academic Editor: Enrico Capobianco

Received: 30 December 2020 Accepted: 8 February 2021 Published: 19 February 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/).

preservation (on 5000 random cases). All validation benchmarks were met. **Conclusion:** *BLADE* is a support system for follow-up and assessment of breast cancer care. To successfully develop and validate it as the first standardized data collection system, multidisciplinary collaboration was crucial in selecting its ontology and knowledge domains. *BLADE* is suitable for multi-center uploading of retrospective and prospective clinical data, as it ensures anonymity and data privacy.

**Keywords:** breast cancer; large database; standardized data collection; networks

#### **1. Introduction**

Breast cancer, one of the main causes of women's mortality, is characterized by highly complex presentation patterns [1]. Even though population-based screening programs [1], new therapies [2], advanced technologies [3], and multidisciplinary approaches [4] have improved survival and quality of life [4] in the previous decades, cure remains a challenge in some sub-groups of patients. Consequently, hypothesis-based tailored treatments that are adapted to each individual patient's specific features are being explored in an approach termed personalized medicine. Due to complex information systems, personalized medicine overcomes uncertainties about particular conditions in small sub-groups of patients, which increase the complexity of decision-making [5,6]. Despite growing interest, a literature review revealed no consensus on how to define and apply personalized medicine [5]. Semantic approaches include patient stratification and treatment tailoring. In the former, individual patients are grouped into subpopulations according to the probability that a specific drug or treatment regimen will be of benefit, whereas in the latter, the individual patient's status is used as the rationale for treatment choice [6,7].

The application of personalized medicine may be limited in clinical practice by the results of randomized controlled trials (RCTs). Patient selection, as defined by inclusion and exclusion criteria, leads to adaptive randomization, so outcomes refer only to the RCTeligible population [8]. Furthermore, since the selected patients are usually in good clinical condition, with few or no comorbidities, the results cannot be extrapolated to all cases that physicians may encounter in clinical practice [9]. Additionally, due to long recruitment and follow-up times, RCT evidence may be out-of-date when it is made available, and progress may have already been made in developing treatments beyond old standards. Lambin et al. [10,11] reported that high quantity, low quality data from clinical charts reflected reality better than RCT data, and therefore provided valuable information for applying personalized medicine in clinical practice [9,12]. However, new instruments are needed to include the data and address uncertainties in clinical decision-making.

Rapid learning health care fills this gap, since it extracts and applies knowledge from routine clinical care data rather than RCT evidence alone. Since data management of crosslinked information from diverse sources is complex, data analysis should be managed by machine learning to create decision support systems, i.e., software applications that apply knowledge-driven healthcare to clinical practice. Another rapid learning principle is that these systems need constant updating.

In February 2016, a group of expert radiation oncologists organised the Assisi Think Tank Meeting (ATTM) to discuss research, controversies, and grey areas in breast cancer [13], and proposed a validated system based on rapid learning health care for standardized data collection to generate evidence for personalized medicine. In one of the participating centers, the Fondazione Policlinico Gemelli IRCCS, an umbrella protocol [14,15] was already approved by the Ethical Committee. The Beyond Ontology Awareness (BOA) platform (Figure 1) had been developed and implemented in close collaboration with physicians and informatics technology researchers [8,13]. It safely stores, analyzes, and shares data on diverse cancer types in a standardized manner [9,16] as well as reproducing the ontology structure and managing data legacy and privacy. BOA software converts the center's

legacy archives in accordance with a global data dictionary and anonymously replicates clinical data in a large cloud-based database.

In the present project, the BOA platform was expanded for specific use in breast cancer care. A multi-disciplinary panel of experts from the Fondazione Policlinico Gemelli IRCCS, Perugia University, and General Hospital designed a standardized data collection system and developed the *Breast LArge DatabasE (BLADE)*. Its primary objective was to offer radiation oncology centers worldwide treating breast cancer the opportunity to collect and share data in a standardized large database, and thus develop descriptive, predictive, and prognostic models for supportive care, survival, and toxicity. Its long-term aim is to build decision support systems to personalize treatments, use resources in terms of cost-effectiveness, and make therapies more effective and less toxic.

**Figure 1.** General beyond awareness ontology (BOA) architecture, with both the BOA.Local and BOA.Cloud servers. An infinite number of external institutions without a BOA.Local installation can be added at needed to this infrastructure. Double-line arrows represent non-anonymized patient data, dashed arrows represent anonymized patient data, and dotted arrows represent aggregate data.

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

After a review of breast cancer literature and current guidelines, a multi-step process was set up for data collection, analysis, application, and evaluation. Benchmarks were the rapid learning criteria by Lambin et al. [11]. The project was organized in a 7-step working plan as defined in a GANNT chart, and the time-frame for each step was established [17] (Figure 2). Data collection was structured to capture volume, variety, velocity, and veracity [11] and aimed to achieve a standardized ontology and overcome privacy issues. Approval was acquired from the Ethical Committee.

**Figure 2.** Timeline framework for ATTM.BLADE project.

#### *2.1. Data Collection Methodology*

*Working Plan and Team (Step 1).* Members of the working group from the Fondazione Policlinico Gemelli and Perugia University, and General Hospital included 6 radiation oncologists; 1 medical oncologist; 1 pathologist; 3 breast surgeons; 1 radiologist; 2 informatics experts; 1 data manager. The working group established a timeframe of 12 months for developing the *BLADE* system. Responsibilities and times to complete each step were defined. Progress was updated every 3 months via live meetings or conference calls.

*Variable Selection and Organization (Step 2).* Each team member reviewed the literature, focusing on RCTs and international guidelines, e.g., NCCN, ASTRO, ESTRO, and AIRO for radiation oncology [18–20] and established 7 domains of knowledge: baseline, primary systemic therapy, surgery, adjuvant systemic therapies, radiation therapy, follow-up, and toxicity. Major variables were chosen for each domain to create a shared-language ontology (terminology system). Variables were related to patients (e.g., age, sex, and gene profiling), clinical presentations (e.g., disease stage, markers, and pathology findings), treatments (e.g., surgery, systemic therapies, radiation therapy, and palliative care), and imaging (at diagnosis, treatment, and follow-up).

Variables were validated by a consensus panel that indicated the response type for each variable (yes/no, single, or multi-options), selected and voted on multi-options. Consensus was reached with 80% agreement.

*Setting up the BLADE domain (Step 3).* BOA was configured to include *BLADE* and process breast cancer data. It is equipped with local and cloud servers (Figure 1**)** depending on the desired configuration package. Users can access the BOA services through an intranet or internet connection and need only a standard web browser to connect, with no additional software. In the BOA.Local configuration, which only allows access through the local intranet, each institution has complete control over its data repository, and collected records are saved without any automated pseudo-anonymization procedures. The internetfacing server installation on the BOA.Cloud has the same features as the BOA.Local service, but it automatically and mandatorily pseudo-anonymizes all data. Before storage, each patient is assigned an ad hoc universally unique identifier (UUID), and all personal data or connections to existing records are severed. BOA.Cloud and BOA.Local store and process data in accordance with General Data Protection Regulation (GDPR). BOA.Local data can be dynamically cloned, automatically anonymized, and consolidated onto the BOA.Cloud server through a research manager—research node connection algorithm, and the data are then ready to be processed or analysed as needed. Figure 3 illustrates the underlying data model used in the databases of both BOA services.

**Figure 3.** Underlying BOA data model visualized through an entity–relationship model that highlights all relationships between the different objects in the database. As an example (and using imaginary values), the archive named *BLADE* would contain a patient named John Doe, affected by a pathology of breast cancer, for which he was treated through a treatment of first treatment. This treatment would have a compiled version of the case report form (CRF) radiotherapy linked to the phase called neoadjuvant, and an answer of prone, to the question of radiotherapy treatment position present in the previously mentioned CRF.

> To create the *BLADE* domain, Excel spreadsheet files with all ontology-related variables were uploaded on to the BOA platform. *BLADE*'s 7 specific case report forms (CRFs), which were devised according to OpenClinica system criteria [21], are compatible with the BOA ontology framework. CRFs are available in Supplementary Materials file 1, with explanations of CRF definitions in Supplementary Materials file 2.

> *Inclusion Criteria (Step 4).* The working group defined patient selection criteria, agreeing that retrospective and prospective data from all selected breast cancer patients can be included in *BLADE.*

> *Retrospective data* : When *BLADE* is installed on the BOA platform, patient data will be derived from existing retrospective electronic or paper databases in each participating center. The data will be anonymized and shared only for research purposes.

> *Prospective data* : Patients whose data are eligible for enrolment in prospective *BLADE* studies will be informed about the opportunity to share their data for research purposes at their first medical examination, and invited to participate. The patients' written informed consent will be obtained and archived.

> *Patients' privacy protection (Step 5).* Privacy needs to be guaranteed according to GDPR guidelines [22] for data protection. *BLADE* and BOA manage data using an AES-256 encryption system and an automatic data pseudo-anonymization algorithm. Each case is associated with a UUID code number with no reference to the individual's identity, and is only accessible to specifically authorized health operators through their personal access codes and accounts. All changes in *BLADE* are automatically tracked and logged, including past and present values for form fields and the account identifiers of operators that modified existing data or inserted new data into CRFs.

#### *2.2. Testing the BLADE Domain for Coherency and Reliability (Step 6)*

A data entry expert in the CRF system inserted data from 15 clinical charts of breast cancer patients that were randomly selected from Policlinico Gemelli records. According to GDPR principles, informatics verified accuracy, data conservation, limitations, and integrity during uploading. Criteria for coherency and reliability tests of the *BLADE* domain were the following (Article. 32 of GDPR):


#### *2.3. System Validation (Step 7)*

*BLADE* was validated after checking adhesion to the GDPR criteria, and uploading and extracting data for statistical analysis from the clinical records of 64 patients who had undergone post-mastectomy radiation therapy (RT). All patients gave permission for their data from local databases to be transferred to *BLADE*.

Physicians asked the informatics expert to extract the following data from *BLADE*:


Records were automatically extracted and the output was structured according to the standard needs of a data science team (e.g., a .csv file with all selected records processed on a flat table with specific column names and without any identifying information).

Validation benchmarks were:


#### **3. Results**

#### *3.1. Setting up BLADE (June 2016)*

The 12-month timeline for completing *BLADE* overran by more than 1 year due to the quantity and complexity of the information. For example, Step 2 lasted 18 months, during which the working group met three times for variable selection and three times for variable validation. In July 2018, after reaching 80% consensus, a total of 218 variables were successfully uploaded to constitute the *BLADE* domain. Figure 4 reports as an example, the definition of the radiotherapy variable according to OpenClinica criteria.


**Figure 4.** Example of a CRF configuration file. The columns represent various mandatory configuration settings for BOA and are to be interpreted as follows: The ID column represents an internal identifier and is generated automatically when the CRF is first uploaded. CRF\_NAME refers to the name by which the CRF is to be visualized in the UI. QUESTION\_NUMBER can either be automatically assigned or manually set, and refers to the ordering of the various questions inside the CRF, with SECTION\_NAME and SECTION\_LABEL working as visual dividers when the questions are displayed in the interface, with the former being the name to be used in the UI code, and the latter being the name to be displayed. ITEM\_NAME and DESCRIPTION\_LABEL work in a similar manner, with the former being the identifier in the underlying code and the latter being the name of the text to be displayed with the question in the UI.

> The variables were organized into seven main CRFs corresponding to the knowledge domains, which were the interfaces for uploading encrypted patient data. In parallel with the data entry expert's work, automatic testing tools in BOA tested specific characteristics in reference to the *BLADE* domain and generated synthetic patients. BOA tested both itself and the linked infrastructure by generating 5000 synthetic patients with a variable number of CRFs, and randomly created data in the space of nearly 20 min. To test performance, 30 fake user agents were connected to the interface and random pages from the web-service were requested for deletion or modification. Numbers for testing tool input were over a hypothetical maximum simultaneous workload for the *BLADE* project. Throughout these tests, no noticeable performance degradations were revealed, no abnormalities in the data structure or integrity were found, and no information leaked in the fake user sessions due to, for example, wrongly configured page-caching settings.

> The privacy protection protocol was initially approved by the Ethical Committee of Fondazione Policlinico Gemelli IRCCS with protocol no. 0043996/18 in October 2018.

#### *3.2. BLADE Data Safety Tests (January 2019)*

To check that the *BLADE* domain was uploaded correctly, informatics analyzed accuracy, conservation limitation of data, data integrity, and data flows between application and data processing on 15 charts from randomly chosen patients. They completely adhered to EU GDPR criteria as reported in Article 32 Security of Processing [22,23]. Uploaded data were not linked to individual patients. Technical and organizational effectiveness measures did not break confidentiality, integrity, availability, and resiliency. Simulated physical and technical accidents showed no loss of data.

#### *3.3. Validation (February–July 2019)*

The physician's review increased 81.5% of uploaded data from 64 patients to 84% and corrected 10% of uploaded and missing data. The following were corrected: compiletime errors due to the data manager's lack of experience with *BLADE* (7.5%); missing data (8.5%).

For statistical analysis, 100% of clinical, treatment and tumor-related data, 80% of reconstruction data, and 98% of dosimetric data were available. Mean available data ranged from 92.6% to 94.5%, corresponding to <20% validation benchmarks. All the planned statistical analyses were performed.

#### **4. Discussion**

The *BLADE* project was set up to support ATTM research into breast cancer, with the aim of providing decision support systems to facilitate clinical decision-making and treatment tailoring. In the 2016 ATTM [13], attention focused on developing such a system from the potentially large database that was available from clinical records, not only in radiation oncology centers, but in many other specialty units (e.g., surgery, pathology, medical oncology, etc.) that are dedicated to the diagnosis and treatment of breast cancer.

The present results showed that *BLADE* is a valid system for collecting data anonymously, as its encryption system successfully passed the tests, satisfying GDPR criteria and benchmarks. Data managers were accountable for only 7.5% of errors, some of which were corrected during the physician's review. Regarding radiation therapy, *BLADE* uniquely focuses on clinical, technical and dosimetric parameters, which makes it particularly suitable for analyzing radiation-therapy-related outcomes and toxicity.

One of the strengths of *BLADE*'s ontology lies in its validated variables that were uploaded after a multi-step process involving the consensus of a multidisciplinary team. Unlike other large databases for breast cancer, *BLADE* provides health workers with the opportunity to focus on diverse fields in the diagnosis and treatment of breast cancer, as it is based on the acquisition of the pathways and the heterogeneity characterizing breast cancer [24–28]. Although several large national databases were set up, none were based on validated, published ontologies [25–28], and few could offer decision support systems [29–33]. Most were developed to investigate long-term survival outcomes such as, for example, the Surveillance, Epidemiology, and End Results (SEER) database, which was set up by the U.S. National Cancer Institute (NCI) and reports annually on the data it has collected on breast cancer from nine American oncological centers [29–31].

Another strength of the *BLADE* system is its capacity to incorporate new, validated variables or mathematical algorithms for assessing, for example, the success of treatment or a strategy for monitoring clinical outcomes and cost-effectiveness. In the future, it might include accreditation or valuation indicators for associated centers, update evidence or guidelines, and incorporate new sectors such as proteomics, complementary medicines, etc.

One limiting factor of the present study was linked to *BLADE*'s small homogeneous sample and its inability to upload digital imaging and communications in medicine (DI-COM) data, which will be very relevant when *BLADE* is used to develop prediction models. DICOM data and RT planning information will be uploaded with the 2022 *BLADE* upgrade, which will create a unique data repository [34]. A lack of testing of *BLADE*'s ability to perform machine learning analysis, an upcoming modality in cancer care, especially for predicting response to treatment, is a current limitation that is expected to be eliminated in the future. Using algorithms that iteratively learn from data, machine learning allows computers to find hidden insights without being explicitly programmed where to look, while inferential statistics need different tools to achieve this purpose, such as Bayesian networks, support vector machines, neural networking, and Cox regression. Machine learning is now starting to flank inferential statistical models (e.g., linear models, generalized linear models, and survival models), and its success over inferential statistics has already been reported together with the first promising results of its use in building predictive models of cancer survival [10,15,19]. We are confident that when *BLADE* is expanded to systematic multi-center use, machine learning analysis will become a reality and systems for decision-making support will be developed and validated, as *BLADE* is projected for a huge number of patients who will provide millions of data for analyses.

In the near future, we will use *BLADE* in our clinical daily practice to collect retrospective and prospective data and analyze outcomes to assess the role of post-mastectomy

radiation therapy in ductal in situ patients. This approach is derived from a 2019 survey by an ATTM research group [35], identifying this topic as a grey area in current practice.

#### **5. Conclusions**

*BLADE*, one of the projects emerging from the 2016 ATTM [13], is a support system for breast cancer care. In successfully developing and validating it as a standardized data collection system, multidisciplinary collaboration was crucial for selecting its ontology and knowledge domains. *BLADE* is suitable for multi-center uploading of retrospective and prospective clinical data, as it ensures anonymity and data privacy.

Finally, *BLADE* may constitute an international instrument for research purposes to be used by ATTM-like research groups [36].

**Supplementary Materials:** The following are available online at https://www.mdpi.com/2075-442 6/11/2/143/s1, Supplementary Materials file 1: CRFs, Supplementary Materials file 2: Explanations of CRF Definitions.

**Author Contributions:** Conceptualization, C.A., V.V. and P.M.P.P.; methodology C.A., N.D.C., F.M. (Fabio Marazzi), C.M., M.M. and V.M.; software, N.D.C.; validation, C.A., V.V. and P.M.P.P.; formal analysis C.A., F.M. (Fabio Marazzi), V.M., M.M. and N.D.C.; investigation C.A., F.M. (Fabio Marazzi), V.M., M.M. and N.D.C.; resources V.V.; data curation C.A, F.M. (Fabio Marazzi), V.M., M.M. and N.D.C.; writing—original draft preparation C.A., F.M. (Fabio Marazzi), M.M. and V.M.; writing review and editing, F.M. (Fabio Marazzi), V.M., M.M., S.S., P.B., L.B., N.D.C., A.D.L., S.M., E.M., F.M. (Francesca Moschella), A.M., D.S., D.A.T., L.T., G.F., M.A.G., R.M., V.V., P.M.P.P. and C.A.; visualization F.M. (Fabio Marazzi), V.M., M.M., S.S., P.B., L.B., N.D.C., A.D.L., S.M., E.M., F.M. (Francesca Moschella), A.M., D.S., D.A.T., L.T., G.F., M.A.G., R.M., V.V., P.M.P.P. and C.A.; supervision C.A., P.M.P.P. and V.V.; project administration, C.A., V.V., P.M.P.P., F.M. (Fabio Marazzi) and V.M.; funding acquisition V.V. All authors have read and agreed to the published version of the manuscript.

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

**Institutional Review Board Statement:** The study was conducted according to the guidelines of the Declaration of Helsinki, and approved by the Institutional Review Board (or Ethics Committee) of Fondazione Policlinico A. Gemelli IRCCS (protocol code N. 0043996/18—31.10.2018).

**Informed Consent Statement:** Not applicable.

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

**Information to Join ATTM.BLADE Network and/or Propose Research Project:** MD. Vincenzo Valentini—vincenzo.valentini@policlinicogemelli.it; Prof. MD. Cynthia Aristei—cynthia.aristei@unipg. it; MD. Fabio Marazzi—fabio.marazzi@policlinicogemelli.it; MD. Valeria Masiello—valeria.masiello@guest. policlinicogemelli.it.

#### **References**

