The Prognostic Role of β-Catenin Mutations in Desmoid-type Fibromatosis Undergoing Resection Only

Supplemental Digital Content is available in the text Objective: This meta-analysis (PROSPERO CRD42018100653) uses individual patient data (IPD) to assess the association between recurrence and CTNNB1 mutation status in surgically treated adult desmoid-type fibromatosis (DTF) patients. Summary of Background Data: The majority of sporadic DTF tumors harbor a CTNNB1 (ß-catenin) mutation: T41A, S45F, and S45P or are wild-type (WT). Results are conflicting regarding the recurrence risk after surgery for these mutation types. Methods: A systematic literature search was performed on June 6th, 2018. IPD from eligible studies was used to analyze differences in recurrence according to CTNNB1 mutation status using Cox proportional hazards analysis. Predictive factors included: sex, age, mutation type, tumor site, tumor size, resection margin status, and cohort. The PRISMA-IPD guideline was used. Results: Seven studies, describing retrospective cohorts were included and the IPD of 329 patients were used of whom 154 (46.8%) had a T41A mutation, 66 (20.1%) a S45F mutation, and 24 (7.3%) a S45P mutation, whereas 85 (25.8%) patients had a WT CTNNB1. Eighty-three patients (25.2%) experienced recurrence. Multivariable analysis, adjusting for sex, age, and tumor site yielded a P-value of 0.011 for CTNNB1 mutation. Additional adjustment for tumor size yielded a P-value of 0.082 with hazard ratio's of 0.83 [95% confidence interval (CI) 0.48–1.42), 0.37 (95% CI 0.12–1.14), and 0.44 (95% CI 0.21–0.92) for T41A, S45P and WT DTF tumors compared to S45F DTF tumors. The effect modification between tumor size and mutation type suggests that tumor size is an important mediator for recurrence. Conclusions: Primary sporadic DTFs harboring a CTNNB1 S45F mutation have a higher risk of recurrence after surgery compared to T41A, S45P, and WT DTF, but this association seems to be mediated by tumor size.

D esmoid-type fibromatosis (DTF) was first described 185 years ago by MacFarlane, and it was named desmoid in 1838 by Müller who referred to the Greek word ''desmos'' meaning ''a tendon-like structure''. 1,2 Ever since, the understanding of this non-metastasizing and histologically benign tumor has grown remarkably. 3 Its potential to arise in musculoaponeurotic structures at virtually any body site and to invade surrounding structures poses therapeutic challenges. A histological biopsy, with nuclear ß-catenin staining, can confirm the diagnosis. 4 In recent years, there has been a tendency for an ''active surveillance'' approach in asymptomatic patients and several prospective clinical trials (NCT02547831, Italy, NTR 4714, the Netherlands, and NCT01801176, France) are conducted to evaluate the safety of this approach. [5][6][7][8][9] For progressive patients, surgical resection, isolated limb perfusion, radiotherapy, and systemic therapy are available treatment options. 10 The variable growth behavior with the possibility of tumor progression, growth arrest, or regression without treatment, makes this tumor unpredictable. 5,11,12 Local recurrence after surgery, especially in case of tumors located in the extremities, the head/neck region and intraabdominal, occur frequently. 13,14 The genetic roots of DTF have been extensively studied. 15 Desmoid tumors can occur as part of the inherited condition familial adenomatous polyposis (FAP) and the FAP subtype; Gardner syndrome. 16 Both conditions are associated with mutations found in the adenomatous polyposis coli gene on chromosome 5, and are known for the development of hundreds of pre-malignant colonic polyps. The development of mainly intra-abdominal DTF tumors is one of the associated manifestations, with a cumulative lifetime risk reaching 21%. 17,18 Both syndromes will not be further discussed in this meta-analysis because the origin and clinical course of these diseases and the DTF tumors for which they predispose, differ from the sporadic variant of DTF. The sporadic variant is associated with extra-abdominal or abdominal wall desmoid tumors and finds its origin in the CTNNB1 (ß-catenin) gene. 15,19 -22 b-catenin is involved in several downstream signaling pathways, functions as a transcriptional activator and is involved in cell-cell adhesion. 19 The mutations are located on exon 3, causing mostly the following amino acid changes: T41A, S45F, and S45P. 19,23 The remainder of tumors, less than 5%, that lack a mutation in the CTNNB1 gene, and of which the underlying genetic aberrations are not entirely clear yet, are called wild-types (WT). 15 The use of the CTNNB1 mutation as a prognostic factor for recurrence after resection has been the subject of several studies. Some studies report that S45F-mutated DTF tumors exhibit a higher recurrence rate after primary resection than WT or other CTNNB1-mutated tumors. 19,[22][23][24] However, others report conflicting results and could not reveal an impact of specific CTNNB1 mutations on outcome. 20,25 These contradictory results are almost certainly due to the fact that these are relatively small retrospective studies including heterogeneous patient cohorts, which hinders the assessment of the true prognostic value of specific CTNNB1 mutations on outcome. Because the insight into prognostic factors can be crucial to come to a more personalized treatment approach for DTF patients who undergo a resection, we performed a meta-analysis using individual patient data (IPD) to study the impact of the CTNNB1 mutation on the risk of local recurrence in a large series of sporadic DTF patients. The hypothesis is that S45F mutated DTF tumors have a higher risk of local recurrence than WT or other CTNNB1-mutated DTF tumors. The objective of this metaanalysis with IPD is to evaluate the impact of the CTNNB1 mutation type on recurrence and recurrence-free survival (RFS) in adult patients with primary DTF tumors undergoing surgical resection alone.

Protocol and Registration
This study was approved by the local Medical Ethics Committee of the Erasmus Medical Center (MEC-2018-1386) Rotterdam, the Netherlands. The protocol of this meta-analysis was registered on PROSPERO (CRD42018100653) and can be accessed at www.crd. york.ac.uk/PROSPERO. 26

Eligibility Criteria
Studies with surgically treated sporadic DTF as a main subject were included. Papers describing follow-up, risk of recurrence, or RFS in primary DTF tumors with known CTNNB1 mutation type were included in this meta-analysis. The flowchart, depicting the study selection procedure, is available in Supplementary

Study Selection
Two independent review authors (M.J.M.T. and M.R.) assessed the retrieved articles of the search for potential inclusion by reviewing title and abstract. Next, full articles were evaluated according to the predetermined inclusion and exclusion criteria for this meta-analysis (listed in Supplementary Table 3, http://links.lww. com/SLA/B832).

Assessment of Study Quality and Risk of Bias
The Oxford levels of Evidence (Oxford Centre for Evidence Based Medicine) were used to assess the quality of included articles. 27 The risk of bias of included studies was assessed using the Quality In Prognostic Studies tool. 28 This tool consists of 6 domains of potential biases: study participation, study attrition, prognostic factor measurement, outcome measurement, study confounding, statistical analysis and reporting, rated as ''low,'' ''moderate,'' or ''high'' risk of bias. When all domains were considered to be ''low'' or ''moderate'' risk of bias, the risk of bias was considered to be low. If 1 or more domains were rated as ''high'' risk of bias, it was considered to be a high risk of bias. Two authors scored the risk of bias (M.J.M.T. and M.R.) and resolved discrepancies by discussion.

Data Collection Process
Two authors (M.J.M.T. and M.R.) independently extracted clinical and genetic information from the full text using a predefined extraction sheet. Subsequent cross-checks were performed. Inconsistencies were discussed and resolved. Patient overlap was defined as ''the description of a cohort from 1 institute with overlapping time periods''. In the latter case, the largest cohort was included and IPD was requested from the corresponding author.

Data Items
For each eligible article, the corresponding authors were contacted via email to retrieve the IPD. They were asked to provide the data either in a template database or by providing their database. Requested variables included: sex, date of birth, date of diagnosis, age at diagnosis, tumor site (extra-abdominal, intra-abdominal, abdominal wall), tumor site specified, tumor size, treatment type (active surveillance, surgery, radiotherapy, systemic therapy, or combination therapy), date of surgery, resection margin status, presentation (primary or recurrent tumor), FAP status, CTNNB1 mutation (WT, T41A, S45F, S45P, or other), recurrence (yes or no), manner of identifying the recurrent tumor (radiological and/ or pathological confirmation), date of recurrence on radiology and/or pathology, date of last follow-up, vital status (death from disease, death from other cause, alive without evidence of disease, alive with evidence of disease, unknown), date of death, date of the last update of the database. Studies for which IPD was not provided were excluded from this analysis.

Inclusion of Patients Using IPD
Inclusion criteria for individual patients to be selected included: primary DTF tumor, age equal to or above 18 years at diagnosis, treated with surgery alone, and known CTNNB1 mutation. Patients receiving neoadjuvant or adjuvant treatment, patients with rare CTNNB1 mutations, patients with intra-abdominal tumors or FAP, patients younger than 18 years, patients with missing data about recurrence, and patients with unknown CTNNB1 mutation type were excluded from the analysis (Supplementary Table 2, http://link-s. lww.com/SLA/B831). With respect to tumor sites, all papers applied different definitions for tumor sites. For data synthesis the following categories of tumor sites were applied: trunk/back (including the following terms: trunk, trunk superficial, abdominal wall, rectus abdominis, chest aperture, chest wall, flank, back, dorsal, prevertebral, scapular area, vertebra), head and neck (including the following terms: head/neck, neck, cervical), and extremity (including the following terms: extremity, upperextremity distal, upperextremity proximal, ulna, arm, elbow, shoulder, axilla, groin, inguinal, buttock, gluteal, calf, knee, lower leg, leg, popliteal, thigh, lower extremity proximal, lower extremity distal). One study failed to specify the extra-abdominal tumor site and these patients were categorized as having extra-abdominal DTF not specified.
In case the date of surgery was provided, age was calculated using date of birth and date of surgery. In case only the date of birth was provided, age was calculated using the date of birth and the date of diagnosis. In case age was provided in the cohort, it was defined in almost all cases as ''age at surgery.'' The median and mean age in years, the standard deviation (SD), and the interquartile range (IQR) were calculated for the total group and per cohort.
Tumor size was obtained from the provided databases; however, specifics on how this variable was measured were lacking for every cohort. A single measurement of tumor size reported was assumed to be the largest diameter of the tumor. In case tumor size was given in multiple dimensions, the largest tumor size was used. All tumor sizes were converted into millimeters (mm). The mean tumor size (in mm), SD, median and IQR were calculated for the total group and per cohort. In case of nonnormal distribution, log-transformed values were used. Additionally, the percentiles (25th, 50th, and 75th) were used to create age and tumor size categories.
RFS was calculated using the date of surgery and the date of recurrence. In case the date of surgery was not provided, the date of diagnosis and the date of recurrence were used to calculate this variable as we assumed that most patients underwent surgery within 2 months after diagnosis. End of follow-up was considered as ''last date of follow-up,'' or ''date of life or death''.

IPD Integrity
Data of individual patients were not subjected to data checking because cohorts were from various countries and often included patients from various hospitals due to the rarity of DTF. Data integrity was checked by comparing published articles with shared databases containing data of individual patients.

Statistical Analysis
An 1-stage approach was used for this meta-analysis with IPD. All variables were collected in a single database. Categorical variables are summarized as frequencies and corresponding percentages. Continuous variables are summarized as mean values with SD or as median with IQR. Analysis of variance and the Chi-square test were used to evaluate the associations between the variables tumor size, tumor site, sex, mutation type, and cohort.
The Kaplan-Meier method was used to calculate RFS, which was defined as the time between the date of surgery (or the date of diagnosis in case the date of surgery was not available) and the date of recurrence (or the end of follow-up).
Univariable and multivariable Cox proportional hazards analyses were performed to assess the association between the outcome (recurrence) and the independent variables [age, sex, mutation type, tumor site, tumor size (log-transformed), resection margin status, and cohort]. The proportional hazards assumption was tested for each independent variable by including an interaction effect of the independent variable with time since surgery in a Cox regression with time-dependent covariates. In case of significance of the interaction effect, the proportional hazards assumption was considered not to be met and the variable was included as a stratification variable. An interaction effect between tumor size and CTNNB1 mutation type was tested in the multivariable Cox model adjusting for age, sex, mutation type, tumor site, and tumor size; subgroup analyses based on tumor size were considered in case of considerable interaction. Two-sided P < 0.05 was considered statistically significant. SPSS Statistics (version 24) was used for all statistical analysis (IBM, Armonk, NY).

Study Selection and Study Characteristics
Results of the search strategy are presented in the flowchart provided in Supplementary Table 2, http://links.lww.com/SLA/B831. A total of 47 articles were screened based on full text for eligibility. Studies were excludedforthe following reasons: describing apediatric cohort (n ¼ 1), no CTNNB1 mutation type data available (n ¼ 14), no full text available (n ¼ 3), conference abstract (n ¼ 1), no recurrence data described (n ¼ 6), no DTF as main subject (n ¼ 1), and a review article (n ¼ 3). Based on the origin of the patient cohort, 8 studies had to be excluded for describing patient sets having a large overlap with series already published. 19,24,25,[29][30][31][32][33] The papers describing the largest cohort were included to request the IPD. From others we were not able to receive the IPD. 15,34,35 All included articles concerned retrospective cohort studies and received a score of 2b at the Oxford Levels of Evidence 2011. 27 The Quality In Prognostic Studies tool was used for assessing the risk of bias (Supplementary Table 4, http://link-s.lww. com/SLA/B833). 28 A detailed description of the published reports of the included studies can be found in Supplementary

IPD and Clinical Characteristics
A total of 10 corresponding authors were contacted for exchanging IPD. Seven out of 10 authors were willing to share the data. Data from individual patients were provided in a cohort template (n ¼ 1), a database containing a selection of patients (n ¼ 2) or by sharing the entire database (n ¼ 4), and the cohorts were analyzed by 1 author (M.J.M.T.). Patients were screened using the aforementioned inclusion and exclusion criteria for IPD, leaving a total number of surgically treated adult patients with primary DTF tumors of 329. The majority of the patients were females (n ¼ 247, 75.1%) with a median age of 38 years (IQR 31-50 years), and the most common tumor site was the trunk/back (n ¼ 194, 59%). The majority (n ¼ 154, 46.8%) of patients had a T41A mutated DTF tumor. Other clinical characteristics are summarized in Table 1. Missing values included unknown resection margin status for 18 (5.5%) patients, missing tumor size (in mm) for 7 patients (2.1%), and an extra-abdominal tumor site which was not further specified for 49 patients (14.9%). The latter was included as separate category for the variable tumor site.

Follow-up and Recurrence
The median follow-up time was 49 months (IQR, 21-94 months). Of 329 patients, 83 patients (25.2%) experienced a recurrence. Of these 83 patients, the median time to recurrence was 16 months (IQR, 10-31 months) and the mean time to recurrence was 26 months (SD 30 months). Table 2 summarizes the number and corresponding frequencies of local tumor recurrence according to the CTNNB1 mutation type.

Survival
There was a statistically significant (P ¼ 0.019) violation of the proportional hazards assumption for the variable cohort in the multivariable Cox regression, suggesting that baseline hazards differed between studies. Therefore cohort was used as a stratification variable in the Kaplan-Meier analysis and in the univariable and multivariable Cox regression models. Supplementary Figure 1, http://links.lww.com/SLA/B828 shows the Kaplan-Meier survival curves with the cohort as a stratification variable.

Univariable Cox Regression Models
In the univariable Cox regression models, CTNNB1 mutation type, tumor site, and tumor size (log-transformed) were significant prognostic factors for local tumor recurrence after surgical treatment (Table 3).

Multivariable Cox Regression Models
In the first multivariable analysis, resection margin status was left out to reduce model complexity. Because none of the included cohorts indicated tumor size as a prognostic factor, the first multivariable model was only adjusted for sex, age, tumor site, and CTNNB1 mutation, stratified by cohort. In this multivariable analysis, the S45P mutation and WT DTF were significantly less likely to recur compared to S45F mutated DTF with a hazard ratio (HR) of 0.32 [95% confidence interval (CI) 0.11-0.97], P ¼ 0.043 and a HR of 0.34 (95% CI 0.17-0.69), P ¼ 0.003, respectively. A tumor located in the extremities was shown to be an adverse   Table 4.

Tumor Size
Despite the fact that none of the included studies described tumor size as a prognostic factor, a significant P-value for tumor size (log-transformed) was found in the univariable analysis. To gain more insight into this variable, we tested with an analysis of variance whether there were differences in tumor size (as a log-transformed variable) between the cohorts and we found a statistically significant difference (P < 0.001). The cohort from Van Broekhoven et al and Mussi et al contained smaller tumors. Next, we investigated whether the tumor sizes differed between the various mutation groups and found a statistically significant difference (P ¼ 0.001) in tumor size between mutation groups. From this analysis, we concluded that tumors harboring a S45F and S45P mutation were larger compared to T41A and WT tumors. Additionally, we looked whether the tumor size differed between the tumor sites, but this difference was not statistically significant (P ¼ 0.392). No significant association could be found between tumor site and CTNNB1 mutation type (P ¼ 0.261), and between mutation type and sex (P ¼ 0.643), using the Chi-square test. Taken together these findings indicated an important association between tumor size and CTNNB1 mutation and led to the development of a second multivariable model (II) including tumor size as a continuous, log-transformed variable. This model led to nonsignificant results for the association between CTNNB1 mutation and recurrence (P ¼ 0.082). The results of both multivariable analyses are displayed in Table 4.  To examine the impact of tumor size on recurrence, we tested whether there was an effect modification of tumor size and mutation type and found a significant effect modification (P interaction ¼ 0.09) between those variables, adjusting for mutation, tumor size, tumor site, sex, and age. To account for potential effect modification, we used the median tumor size (55 mm) as a cut-off value to perform a subgroup analysis of multivariable analysis (Supplementary Table 6, http://links.lww.com/SLA/B835). Supplementary Figure 2, http:// links.lww.com/SLA/B829 presents the combined estimated effect (in terms of HR) of tumor size and CTNNB1 mutation for the multivariable model for all patients with the interaction between tumor size (log-transformed) and CTNNB1 mutation.

DISCUSSION
This meta-analysis uniquely combined the IPD of 7 studies to determine the effect of the CTNNB1 mutation type on recurrence rate in a cohort of surgically-treated DTF patients, who did not receive any additional perioperative therapy for their primary tumor. Although active surveillance is advocated in asymptomatic patients, a substantial number of patients still receive surgical treatment for their DTF during their course of disease. 10,36 Reported recurrence rates after surgical treatment remain high, between 20% and 68%, and highlight the need for prognostic factors to predict recurrence, to inform patients and to explore strategies to reduce recurrence risk in high-risk groups. 14,24,30,[37][38][39] Tumor site, age at onset, and CTNNB1 mutation type are most frequently mentioned as important prognostic factors. 14,40 Although several studies have reported that CTNNB1 mutation type had no prognostic value for recurrence compared to WT DTF, others have identified a correlation between the CTNNB1 S45F mutation and a higher risk of recurrence. 15,20,24,25,41,42 This study showed that CTNNB1 mutation type is indeed relevant for recurrence, and the observed risk of recurrence was highest for S45F tumors, although the association was not statistically significant (P ¼ 0.082). We also found that tumor size is a mediator for recurrence rate. The findings of this study are remarkable because none of the included studies found a significant correlation between recurrence and tumor size, and did not identify size as a prognostic marker recurrence. 20,22,[41][42][43] However, others did report that tumor size is an independent predictor of recurrence (Crago et al, P ¼ 0.004) or event-free survival (Huang et al, P ¼ 0.006), but they did not adjust for CTNNB1 mutation type. 44,45 Other larger series, not included in this meta-analysis because they did not meet our inclusion criteria or used overlapping patient cohorts, did not find a correlation between tumor size and progression-free survival, 31 time to recurrence, 19 mutation type, 34 and progression or recurrence. 30 Although the variable tumor size is prone to interobserver and intraobserver variability, the current multivariable model suggests that both tumor size and mutation type should be considered as predictors for recurrence in patients with primary, extra-abdominal, surgically treated DTF.
Tumor site, especially the extremity, is reported as a significant prognostic factor for recurrence in the univariable analysis of multiple studies. However, this statement often does not hold in the multivariable analysis. 19,24,37,45 These conflicting results could be explained by the fact that most of these studies comprise relatively small case series or cohort studies and include patients who received neoadjuvant/adjuvant chemotherapy, or radiotherapy, which influences the recurrence rates and impairs the assessments of the true prognostic value of these parameters. 46 The current study shows that a tumor located in one of the extremities, is an adverse prognostic factor (P < 0.001) for recurrence compared to tumors located on the trunk/back. Females represented the majority of patients in the current cohort (75%). This is in line with previous reports describing nationwide cohorts by Broekhoven et al 47 (71.3% females) and Penel et al 36 (72.6% females). In the current study, no statistically significant association between sex and mutation type (P ¼ 0.643) could be found. Additionally, univariable analysis did not identify sex as a prognostic marker for recurrence (P ¼ 0.290).
The added value and the strength of the current study is the use of IPD, which created a relatively large cohort of patients that receive the same treatment for their primary DTF, but there are some limitations to take into account. The use of IPD demanded the formation of new variables, for example, T0 which was defined as date of surgery. Unfortunately, this variable was not available for all patients and was solved by the use of ''date of diagnosis'' in a small part of patients. Another example of a new variable which had to be created was tumor site. Although a statistically significant result was found in both univariable and multivariable analyses, one should keep in mind that there is a relatively large group of patients having an extra-abdominal located tumor without any further specification (extra-abdominal not specified). A major limitation, considering the present knowledge, is the relatively large number of patients with WT tumors in the current cohort. Multiple studies report that CTNNB1 mutations are not always detected by Sanger sequencing due to the low frequency of mutant alleles and the relatively low sensitivity of the technique. This can lead to incorrect allocation to the ''WT group'' while having CTNNB1 or other, novel, mutations. 15,48,49 The CTNNB1 genotyping protocols, used in the included studies (Supplementary Table 5, http://links.lww.com/SLA/B834), focused on the presence of a CTNNB1 mutation in exon 3 and provided no information about the presence of other mutations that may be present in the tumors. Another limitation is the heterogeneity of included studies which differed in primary outcome and differences in follow-up procedures, possibly leading to missing a certain number of recurrent tumors. This is also reflected by the variable recurrence rates reported by the included studies. Moreover, the rarity of DTF is reflected by the limited number of available studies with available CTNNB1 type assessing recurrence, and led to the inclusion of studies with a high risk of bias. As the aim of the current study was to gain insight into the prognostic values of the CTNNB1 mutation, every study, regardless of their sample size, was included in this meta-analysis. Several studies could not be included in this meta-analysis due to potential cohort overlap and due to the impossibility to acquire the IPD. 15,19,[23][24][25][29][30][31][32][33][34][35] Although there might be a selection, this is the largest cohort describing the correlation between CTNNB1 mutation type and recurrence risk to date.
Despite these limitations, using the IPD of several international cohorts created a large pooled cohort of patient that received the same treatment, unique for a rare disease like DTF. The pooled data provided new insights into the prognostic value of the CTNNB1 mutation type in predicting local recurrence in surgically treated DTF patients, with tumor size as an important mediator. Currently, there are no prospective studies examining the value of adjuvant therapy after surgical resection of DTF. Future studies investigating the role of adjuvant therapy in patients with an S45F mutation, in whom re-resection would result in unacceptable morbidity, are recommended. Additionally, mechanistic studies, exploring how the different CTNNB1 mutations affect DTF tumor biology are warranted, especially because an increasing number of patients is being treated with active surveillance. Data from the 3 studies, that are currently investigating the active surveillance approach, will give valuable information regarding the association between CTNNB1 mutation type and DTF tumor behavior.

CONCLUSIONS
Tumor size and CTNNB1 mutation type should be considered as predictors for recurrence in patients with extra-abdominal, surgically treated primary sporadic DTF. Ongoing studies about upfront