Survival Benefit for Individuals With Constitutional Mismatch Repair Deficiency Undergoing Surveillance.

PURPOSE: Constitutional mismatch repair deficiency syndrome (CMMRD) is a lethal cancer predisposition syndrome characterized by early-onset synchronous and metachronous multiorgan tumors. We designed a surveillance protocol for early tumor detection in these individuals. PATIENTS AND METHODS: Data were collected from patients with confirmed CMMRD who were registered in the International Replication Repair Deficiency Consortium. Tumor spectrum, efficacy of the surveillance protocol, and malignant transformation of low-grade lesions were examined for the entire cohort. Survival outcomes were analyzed for patients followed prospectively from the time of surveillance implementation.
RESULTS: A total of 193 malignant tumors in 110 patients were identified. Median age of first cancer diagnosis was 9.2 years (range: 1.7-39.5 years). For patients undergoing surveillance, all GI and other solid tumors, and 75% of brain cancers were detected asymptomatically. By contrast, only 16% of hematologic malignancies were detected asymptomatically (P < .001). Eighty-nine patients were followed prospectively and used for survival analysis. Five-year overall survival (OS) was 90% (95% CI, 78.6 to 100) and 50% (95% CI, 39.2 to 63.7) when cancer was detected asymptomatically and symptomatically, respectively (P = .001). Patient outcome measured by adherence to the surveillance protocol revealed 4-year OS of 79% (95% CI, 54.8 to 90.9) for patients undergoing full surveillance, 55% (95% CI, 28.5 to 74.5) for partial surveillance, and 15% (95% CI, 5.2 to 28.8) for those not under surveillance (P < .0001). Of the 64 low-grade tumors detected, the cumulative likelihood of transformation from low-to high-grade was 81% for GI cancers within 8 years and 100% for gliomas in 6 years.
CONCLUSION: Surveillance and early cancer detection are associated with improved OS for individuals with CMMRD.

INTRODUCTION

Constitutional mismatch repair deficiency (CMMRD) syndrome (OMIM #276300)[1] is a cancer predisposition syndrome associated with inheritance of biallelic pathogenic variants in mismatch repair (MMR) genes (MLH1, PMS2, MSH2, and MSH6), leading to deficient MMR during DNA replication.[2] Loss of functional MMR results in a rapid accumulation of mutations and the development of cancers that are highly aggressive and display a hypermutant phenotype.[2-4] Because of the nature of these tumors, commonly used chemotherapies and radiation are largely ineffective.[5] Consequently, many patients with CMMRD develop symptomatic cancers and die regardless of treatment approach.[6-12]

CONTEXT

Key Objective

Do the current surveillance recommendations for constitutional mismatch repair deficiency (CMMRD) syndrome improve early tumor detection and overall survival in individuals with CMMRD? This has never been tested prospectively.

Knowledge Generated

Early presymptomatic detection of CMMRD cancers using the current recommended surveillance modalities was associated with improved survival. Overall survival was significantly better in individuals with CMMRD who fully adhered to surveillance guidelines compared with individuals who did not undergo surveillance (P < .0001). All low-grade brain and most GI lesions transformed to malignant cancers during the study period indicating a potential benefit of early detection and intervention.

Relevance

Current surveillance recommendations appear effective for early detection and intervention that can improve outcome for patients with CMMRD. These guidelines can be implemented in most centers around the world.

The most common tumors observed in these patients are early-onset CNS, GI, and hematologic malignancies with higher penetrance than other cancer predisposition syndromes.[13] Although other cancer types have been reported in children and adults with CMMRD,[13-16] the frequency and extent of these malignancies are not as well characterized.

Surveillance protocols have previously been developed for several cancer predisposition syndromes such as hereditary breast-ovarian cancer and Li-Fraumeni syndromes and have significantly improved survival for screened patients.[17-20] Effective surveillance relies on three central pillars.[21] First, a surveillance protocol should include modalities that identify most cancer types presenting in at-risk individuals. Second, the recommended modalities should prove effective at early detection of asymptomatic tumors that can be resected or managed effectively with reduced treatment related morbidity. Third, a surveillance protocol must demonstrate that early asymptomatic tumor detection ultimately improves overall patient survival.

Initial efforts to establish surveillance for individuals with CMMRD began with anecdotal cases incorporating GI surveillance[22,23] and serial brain magnetic resonance imaging (MRI), although surveillance approaches varied widely.[22,24,25] Proposed surveillance guidelines for CMMRD were recently revised[26,27] to include more comprehensive recommendations. However, current guidelines are based on expert opinion and no study has systematically evaluated the efficacy of surveillance and whether adherence affects overall survival (OS).

The International Replication Repair Deficiency Consortium (IRRDC) was formed in 2007 and currently includes pediatric and adult patients from more than 45 countries. Since 2008, surveillance recommendations were implemented as part of the management of these patients. Data from all registered IRRDC patients are collected prospectively from the time of enrollment.

The goal of this study was to systematically evaluate the proposed surveillance guidelines for CMMRD[27] to determine whether current modalities are appropriate to detect the full spectrum of tumors seen and whether detecting asymptomatic tumors results in improved outcome.

PATIENTS AND METHODS

Inclusion Criteria

All patients were enrolled in the IRRDC with patient consent and institutional research ethics board approval (SickKids, Research Ethics Board no. 1000048813). Patients were diagnosed as having CMMRD through established genetic, clinical, and other molecular diagnostic criteria by the consortium genetic counselor (M.A.) based on criteria established by an international consensus.[13,28] This group was termed full study cohort (Fig 1). Patients who did not completely fulfill these established criteria including suspected yet unconfirmed CMMRD were excluded. Since current evidence on CMMRD does not suggest an association between the cancer susceptibility of a patient and which of the four MMR genes is affected, we included all patients with CMMRD in this study and added the gene affected to our multivariable analysis.

FIG 1.

Study overview. CMMRD, constitutional mismatch repair deficiency; IRRDC, International Replication Repair Deficiency Consortium.

Tumor Spectrum

Cancer diagnoses were collected for the full study cohort (n = 110). These were compared with patients with CMMRD previously published in the literature through a computer-aided search of MEDLINE (2008 to December 2019). The search used CMMRD syndrome and biallelic mismatch repair deficiency syndrome as keywords and was restricted to English language publications. Malignant lesions and age of diagnosis were recorded. Authors were contacted to clarify patient identification to minimize duplication of cases.

Surveillance Data

The surveillance protocol used by the consortium was published in 2017 and includes the following: Brain MRI is to be implemented at diagnosis and repeated every 6 months. Whole-body MRI (WBMRI) should begin at age 6 years and is to be performed annually. Abdominal ultrasound and CBC should be performed at diagnosis and repeated every 6 months. Upper and lower endoscopy are recommended annually starting at 4-6 years of age. Further details of the protocol are outlined in the Data Supplement (online only).[27] Data were obtained through ongoing follow-up for the full study cohort. In addition, a surveillance questionnaire was sent to each patient's physician (Data Supplement). Medical documents, including pathology, surgical, endoscopy, and autopsy reports, were also reviewed. The end of patient follow-up date for the study was November 1, 2019.

To evaluate the efficacy of early cancer detection by the recommended guidelines, tumors from patients undergoing surveillance from the full study cohort were identified either as asymptomatic or symptomatic at diagnosis and the modalities that detected each tumor were reported. Data were also collected on all benign and low-grade lesions found in the full study cohort.

Prospective Cohort and Outcome Measures

To assess whether surveillance guidelines and early detection affect survival, we evaluated patients who were followed prospectively from 2008 (n = 89), the year surveillance was initiated (Fig 1). Patients in this prospective cohort were categorized as either full, partial, or no surveillance based on the level of surveillance they received. The full surveillance cohort (n = 33) comprised individuals who underwent routine screening using all the modalities recommended. The partial surveillance cohort (n = 20) included individuals who did not consistently undergo all screening modalities or from whom modalities were not performed at recommended intervals. Examples include individuals in whom only colonoscopy and CBC were performed, or for whom brain MRIs were performed at intervals > 6 months. The no surveillance cohort (n = 36) included individuals who did not receive any routine screening. Overall patient survival was determined for these three groups. Patients who were enrolled in the consortium before 2008 (n = 21) and whose data were retrospectively collected were excluded from survival analysis.

To test the role of early asymptomatic tumor detection and considering that patients with CMMRD develop synchronous and metachronous cancers, a secondary separate analysis of patient outcome was measured per tumor. Tumors were divided into two groups, symptomatic or asymptomatic, based on presentation at diagnosis. Tumor-specific survival was subsequently measured by comparing these two groups. To avoid further bias introduced by low-grade tumors, patient- and tumor-specific survival analyses were determined solely using malignant tumors from the prospective cohort. Patients who were enrolled in the consortium before 2008 (n = 21) and whose data were retrospectively collected were excluded from survival analysis. Patient country of origin was classified as low resource or high resource based on the World Bank Global Index LMIC List 2020.

Statistical Analysis

For analysis of asymptomatic versus symptomatic tumor-specific survival, each tumor was treated as an independent entity and the unit of analysis was the tumor rather than the individual. Tumor-specific survival was defined as time to death or last follow-up from the diagnosis of brain, GI, or other tumors. For participants with multiple tumors, cause of death was attributed to one of the tumors and we censored tumors that were not the cause of death at the time of death. For patients, time to OS was defined in years from date of CMMRD diagnosis to the date of death or last follow-up.

Categorical variables such as sex, gene affected, and tumor types were summarized with counts and percentages. Continuous variables such as age at first cancer diagnosis and follow-up duration were summarized with median and range. One-way analysis of variance, Fisher's exact test, and N–1 chi-squared test (for comparison of proportions) were used where appropriate for analysis of the prospective cohort demographics tables.

OS rates were calculated using the Kaplan-Meier product-limit method. Logrank test was used as a univariate analysis to assess potential prognostic factors. Cox proportional hazards model was also used to assess the joint effect of prognostic factors. All tests were two-tailed with a probability of < .05 considered statistically significance. Statistical analyses were performed using version 9.4 of the SAS System for Windows (Copyright 2002-2012 SAS Institute Inc, Cary, NC), and SPSS, Statistics for Windows, Version 22.0., IBM Corp (Armonk, NY).

RESULTS

Overall, 110 patients with confirmed CMMRD and 193 malignant tumors were identified (Fig 1). Two tumors were excluded from the remaining analysis as full information was not available. The full study cohort comprised 110 patients and 191 tumors that were used to determine the efficacy of the proposed surveillance modalities in malignant tumors and for analysis of low-grade lesions. Ongoing prospective data were collected for 89 patients. This established the prospective cohort for the survival analysis. Five patients did not develop cancer during the study with a median follow-up of 1.6 years (range, 0.68-4.98 years) and 84 developed 139 malignant tumors during the study time.

Tumor Spectrum for Patients With CMMRD

The spectrum of malignant tumors identified for IRRDC patients (full study cohort) is presented in Figure 2A. Additional demographics are detailed in the Data Supplement. Overall, the median age of first cancer diagnosis was 9.2 years (range, 1.7-39.5 years). CNS tumors were the most common cancers observed (44%, n = 85; median age at diagnosis 9.9 years, range 2.3-38.5 years) followed by GI (27%, n = 52; median age at diagnosis 15.9 years, range 8.5-49.9 years) and hematologic malignancies (19%, n = 37; median age at diagnosis 10.5 years, range 2.2-29.9 years). We compared the prevalence of cancers in individuals with CMMRD from the literature (Data Supplement) to our cohort and revealed similar ratios. Ten percent of tumors (n = 19; median age at diagnosis 16.1 years, range 1.7-52.3 years) occurred in other organs, such as bone, soft tissue, genitourinary, and embryonal cancers.

FIG 2.

Tumor spectrum and efficacy of the surveillance modalities to detect asymptomatic malignant tumors in the full study cohort. (A) Distribution of malignant tumors from individuals with CMMRD. (B) Tumors detected during surveillance per tumor type. *P = .03 (brain) compared with literature cohort (brain) determined using “N–1” chi-squared test. ALL, acute lymphoblastic leukemia; AML, acute myelomonocytic leukemia; CMMRD, constitutional mismatch repair deficiency; IRRDC, International Replication Repair Deficiency Consortium.

In addition, adult type tumors, such as colorectal, breast, and genitourinary tumors, were observed at young ages (31%, median age 17.9 years, range 8.5-52.3 years). Of the 52 patients who survived their initial cancers, 36 (69%) developed metachronous cancers (> 2 years between cancers). A detailed list of the malignant tumors identified for each IRRDC patient along with basic demographics is listed in the Data Supplement. Together, the high frequency of numerous cancers in multiple organ systems at a young age highlights the importance of implementing a multimodal surveillance protocol early in life.

Efficacy of Surveillance for Asymptomatic Tumor Detection

To test the efficacy of recommended modalities in detecting asymptomatic tumors, we looked specifically at individuals undergoing surveillance within the full study cohort (n = 56). We then analyzed the cancers developed during this period to determine which tumors were detected asymptomatically or symptomatically (Fig 2B). Twenty CNS tumors were identified, including both malignant gliomas and embryonal tumors such as medulloblastoma. Both brain-specific and WBMRI were able to detect 15 (75%) asymptomatic CNS cancers. The five brain tumors that developed symptomatically while on surveillance were in patients who had intervals longer than six months between scans (n = 1, 1.5 years), or interruptions because of access or availability to surveillance modalities (n = 4). All 24 GI tumors were identified asymptomatically by surveillance. Similarly, all five of the other solid tumors detected by surveillance were asymptomatic. By contrast, only two of 12 (16%) hematopoietic malignancies were detected by current surveillance methods, both of which were identified incidentally.

Survival Benefit for Asymptomatic Detection of Malignant Cancers

Having established the tumor spectrum and potential benefit of the surveillance recommendations for patients with CMMRD, we examined the impact of early detection of asymptomatic malignant cancers on survival. Within the prospective cohort, 139 tumors were detected. Of those, 39 tumors (28%) were detected by surveillance tools before symptoms arose. All other tumors (72%, n = 100) were diagnosed after initial symptoms or signs. Survival analysis was performed per tumor. Five-year OS was 90% (95% CI, 78.6 to 100) and 50% (95% CI, 39.2 to 63.7) for asymptomatic versus symptomatic detection, respectively (P = .001, Fig 3A). Having a large cohort in this study allowed for evaluation of the two most common cancer types. For CNS tumors (92% high-grade glioma, 6% medulloblastoma, and 2% CNS embryonal tumors), 5-year OS was 72% (95% CI, 49.5 to 100) for asymptomatic tumors (n = 15) and 33% (95% CI, 20.5 to 55.6) for symptomatic tumors (n = 53; P = .04, Fig 3B). For GI cancers (n = 33, Fig 3C), OS was 100% (95% CI, 100 to 100) for tumors detected asymptomatically (n = 18) during surveillance compared with 81% (95% CI, 71.1 to 100) survival for symptomatic tumors (n = 15; P = .18). Multivariable analysis for all tumors (Data Supplement) and for brain (Data Supplement) and colon (Data Supplement) separately confirmed the independent value of asymptomatic tumor detection in all cancers and brain tumors specifically.

FIG 3.

Survival for individuals with constitutional mismatch repair deficiency according to surveillance status. (A) Survival for all tumors stratified by symptomatic and asymptomatic detection. (B) Survival for brain tumors stratified by symptomatic and asymptomatic detection. (C) Survival for GI tumors stratified by symptomatic and asymptomatic detection. (D) Survival for patients stratified by surveillance adherence. P values were calculated by log-rank test. OS, overall survival.

Survival Benefit for Patients With CMMRD Undergoing Surveillance

We then tested the ability of the surveillance protocol to improve survival for patients with CMMRD. Patients in the prospective cohort (n = 89) were further divided into three groups: individuals undergoing full surveillance (n = 33), partial surveillance (n = 20), and no surveillance (n = 36; see Methods, Data Supplement). Patient demographics were similar between groups (Table 1). Only eight patients were older than 18 years when the surveillance was initiated.

TABLE 1.

Prospective Cohort Patient Demographics

Four-year OS was 79% (95% CI, 54.8 to 90.9) and 15% (95% CI, 5.2 to 28.8) for patients with CMMRD undergoing full surveillance versus those not undergoing surveillance (P < .0001, Fig 3D). Patients undergoing partial surveillance had favorable outcomes with 4-year OS of 54% (95% CI, 28.5 to 74.5; P = .001).

To examine whether the observed differences in OS was a result of resource differences that affect availability or adherence to the surveillance protocol, we analyzed the distribution of patients from the full, partial, and no surveillance groups by country of origin (Table 1). The distribution of patients who did or did not undergo surveillance did not differ between high- and low-resource settings.

To address the lead time bias, the mortality ratio was calculated for the median follow-up of 25.4 months (Data Supplement). For the full surveillance cohort, mortality was 30% (10 of 33), which was significantly less (P = .001) than the nonsurveillance cohort at 72% (26 of 36). The partial surveillance cohort had a mortality of 40% (8 of 20) compared with the no surveillance cohort (P = .02). There was no significant difference in mortality between the full surveillance and partial surveillance cohorts. Eighty-nine patients from the prospective cohort were from 72 families. The penetrance of this syndrome and familial clustering did not affect presentation, tumor type, and outcome of patients in this study.

In multivariable analysis, including age, sex, gene affected, and resources available, surveillance was the single variable associated with improved OS in patients with CMMRD (P < .0001; Table 2).

TABLE 2.

Multivariable Overall Survival Analysis Using Cox Proportional Hazards Model for the Prospective Cohort

Low-Grade CMMRD Tumors Transform to High-Grade Cancers

A total of 64 benign and low-grade lesions were detected in 54 patients (Fig 4A, Data Supplement). Polyposis and low-grade gliomas (LGG) were the most common tumors. To assess the risk of malignant transformation, we analyzed data for patients undergoing GI surveillance for at least 3 years (n = 19). Of these, 58% (n = 11) developed dysplastic polyps, all of which resulted in prophylactic colectomy (Fig 4B) by age 13.2 years (range, 9.2-25 years). Of the six nonresected LGG identified by surveillance, all revealed universal transformation to high-grade glioma at a median of 1.7 years (range, 1-5.7 years; Figs 4C and 4D). Cumulative likelihood of transformation over time revealed an 81% probability of colorectal polyps undergoing high-grade dysplastic changes within 8 years (Fig 4D). Similarly, for LGG, the likelihood of transformation into high-grade glioma was 100% in 6 years (Fig 4D).

FIG 4.

Low-grade tumors detected in patients with constitutional mismatch repair deficiency undergoing surveillance. (A) Distribution of low-grade tumors in the full study cohort (n = 64). (B) Percentage of polyps with high-grade dysplasia and the frequency of preventative surgery. (C) Magnetic resonance imaging scans showing tumor transformation and time to transformation of six nonresected low-grade gliomas. (D) Cumulative likelihood of high-grade transformation over time for brain (red) and GI (blue) tumors.

DISCUSSION

This is the first prospective report on the impact of surveillance for individuals with CMMRD and demonstrates that a multimodal approach is associated with a survival benefit when these individuals adhere to the recommended surveillance guidelines.[27] The large number of tumors that can develop in affected individuals during a short period enabled us to examine several important concepts associated with current surveillance recommendations.

With increased awareness and diagnosis of CMMRD, the tumor-risk spectrum of this syndrome is expanding (Fig 2A). This study confirms that CNS, GI, and hematologic cancers are the most common tumor types, but also highlights that a significant number of tumors in other tissues, including embryonal tumors such as Wilms tumor and neuroblastoma, are not uncommon during early childhood. In addition, malignancies such as breast cancer, sarcomas, and genitourinary tumors are observed particularly in patients beyond the pediatric years. Overall, the tumor spectrum for CMMRD, including 576 CMMRD-associated cancers (literature and consortium cohorts), supports the recommended screening modalities.

The benefit from early detection of asymptomatic cancers is thought to be related to complete surgical resection and possibly less aggressive adjuvant therapy. Although early detection of tumors has not shown to improve survival in other cancer syndromes and sporadic cancer types,[29-31] our study reveals that early asymptomatic tumor detection is associated with a significant OS benefit, including in malignant gliomas where the role of gross total resection is still debated (Figs 3A-3C).

Another important observation in this study is that some patients with CMMRD have developed multiple metachronous cancers and as a result of prior therapies, exceeded the maximum tolerated lifetime dose limits for radiation and certain chemotherapy agents. Surveillance enables these patients to be managed with surgery and follow-up alone.

Lead time bias must be considered, as it may skew the observed survival benefit for the surveillance cohorts. To address this, we reported mortality ratios for the prospective cohort at the median follow-up (Data Supplement) and further demonstrated a significant increase in survival for the full surveillance cohort compared with the nonsurveillance cohort (P = .001). In addition, lead time bias is typically related to cancers where there is a long time between diagnosis and death. As the majority of CMMRD-associated cancers are rapidly lethal, this bias is less likely in this study.

The proposed modalities and screening frequency identified a large majority of asymptomatic brain tumors (75%), and all of the GI and other solid tumors in patients with CMMRD undergoing surveillance (Fig 2B). Perhaps, the most disappointing part of this study is the inability of current methods to reliably detect asymptomatic hematologic malignancies. Similar observations have been described in other leukemia-related cancer syndromes.[32] Several options may be more successful including metabolic tests such as lactate dehydrogenase or the use of molecular assays such T-cell rearrangement in circulating tumor DNA. Strategies are needed as 43% of the deaths in patients undergoing surveillance were because of hematologic malignancies, whereas these cancers accounted for only 16% of the deaths in the nonsurveillance cohort (while brain tumors predominate 75%).

Low-grade malignancies are invariably detected in all surveillance protocols.[17] Controversy exists regarding whether the transformation risk of these lesions is high enough to offer survival benefit.[17,18,29] We observed a strikingly high rate of transformation into malignant and premalignant lesions for both LGG and colorectal polyps (Fig 4). This may be because of the extremely high rate of mutation accumulation especially with secondary mutations in POLE and TP53.[3,33,34] Complete surgical resection of low-grade tumors, and total colectomy for patients with multiple high-grade dysplastic polyps may be lifesaving by preventing the development to higher-grade tumors.

Although aggressive, the benefit of the currently recommended surveillance protocol on OS of patients with CMMRD is striking (Fig 3D). This is justified as many patients develop multiple cancers in a synchronous or metachronous fashion resulting in extremely low long-term survival. The increased survival benefit associated with partial surveillance for patients with CMMRD is an intriguing finding. This may have management implications for patients with CMMRD worldwide where not all of the recommended modalities are readily available. Relatively simple but consistent follow-up with fewer modalities may still improve outcome for this devastating syndrome.[35]

Although surveillance recommendations were available to all patients with CMMRD since 2008, many centers decided not to burden their patients with such a cumbersome protocol because of the lack of robust evidence supporting the benefit of CMMRD surveillance recommendations. Other reasons for nonadherence to full and consistent surveillance may include poor access to care, lack of education, awareness, social and cultural norms, and physician opinions regarding the role of these protocols in cancer predisposition. However, a lack of awareness and resources on a systemic level does not appear to have influenced the results of this study (Table 1). As observed in other cancer syndromes,[17,18] we hope this study provides strong evidence leading to a wider use and adherence to the current protocol.

There are several limitations to this study that should be discussed. First, as most patients with CMMRD do not reach adulthood, the spectrum of cancers that may present in adults is limited. As the cancer risks become more defined, modification of the protocol should be considered. Second, in our cohort, only one malignant tumor was identified with WBMRI. Nevertheless, the tumor spectrum observed in patients with CMMRD (Fig 2A) clearly demonstrates a need for extended modalities outside of CNS, GI, and hematologic screening. In Li-Fraumeni[17,18] and hereditary pheochromocytoma or paraganglioma sydndromes,[36] WBMRI has been studied prospectively for efficacy and has been shown to play an important role in the detection of both benign and malignant lesions.[37] Ultimately, a longer follow-up time and a larger cohort of long-term survivors would be required to determine efficacy of WBMRI for CMMRD. In addition, it is important to recognize that the level of surveillance adherence was the decision of the patient and clinician and may be based on various factors. Although we considered systemic differences (ie, resource setting) and found no correlation, future studies should investigate other potential factors that may be involved. Finally, it will be important to assess the psychosocial impact of living with this condition and following this aggressive surveillance protocol.

In summary, this study provides justification for the current protocol and demonstrates a significant survival benefit associated with undergoing surveillance in patients with CMMRD. Further work is required to enable better detection of some malignancies.