Mismatch Repair Status Correlates with Survival in Young Adults with Metastatic Colorectal Cancer

Abstract

Background:

Young adults with metastatic colorectal cancer (mCRC) may have higher rates of deficient mismatch repair (dMMR) than older patients. This study sought to assess patterns of MMR-testing and survival among young adult mCRC patients in the National Cancer Database (NCDB), hypothesizing that dMMR correlates with worse survival than in MMR-proficient (pMMR) patients.

Methods:

Stage-IV colorectal cancers were identified in NCDB (2010–2016). Demographic and clinical features were compared between younger (age ≤ 30) and older mCRC patients and tested for association with overall survival. Stage-IV disease without other recorded metastatic sites defined peritoneal metastasis (PM). Fisher-exact tests compared proportions and Cox models tested association with overall survival.

Results:

Of 124,587 stage-IV colorectal cancers, 1,123 (0.9%) were in young patients. Young patients were more likely to have mucinous histology, high-grade, rectal primaries, and isolated peritoneal metastases (p<0.001). Younger patients more often had MMR-testing (29.1 vs 16.6%), with dMMR found at similar rates in young and older patients (21.7 vs 17.1% of those tested, p=0.4). Despite higher rates of adverse prognostic features, younger patients had better survival (median 20.7 vs 14.8 months, p<0.001). In MMR-tested patients, dMMR correlated with higher mortality risk compared to pMMR (median 16.6 months vs 25.5 months, p=0.01). On multivariable analysis, grade and MMR-status remained independently associated with survival.

Conclusions:

Median survival was worse with dMMR by 8.9 months compared to pMMR in young adults with mCRC. Despite higher rates of familial syndromes in young patients and recommendations for universal MMR-testing, over 70% of young mCRC patients had no MMR-status recorded.

Introduction:

Colorectal cancer (CRC) is the fourth most common cancer in the United States and the second-highest cause of cancer death. While overall rates of CRC are decreasing, incidence in adolescents and young adults continues to rise, from 3.0 per 100,000 in 2000 to 5.2 per 100,000 in 2017.^1,2 Young-onset CRC is more often poorly-differentiated, displays high-grade features such as signet-ring or mucinous histology, presents at advanced stages, and follows an aggressive course.^3–7 A recent retrospective review by Hayes-Jordan et al. demonstrated shorter overall and recurrence-free survival in pediatric and adolescent patients with CRC, which was not influenced by predisposing conditions including polyposis syndromes. Younger patients also had significantly higher prevalence of peritoneal metastasis, which had not been previously recognized. Thus, CRC in the adolescent and young adult population is increasingly felt to represent a distinct disease entity, whose nature is incompletely understood.^2–9

Mismatch repair deficiency (dMMR) refers to any of several mutations or inactivations in proteins responsible for maintaining DNA integrity during replication, most notably MSH2, MSH6, MLH1, and PMS2.^10–12 These may be somatic mutations, as in sporadic CRC, or germline mutations, as in heritable CRC disorders. Up to 15% of sporadic CRCs are observed to involve dMMR.¹³ Mismatch repair-deficient tumors often have a right-sided anatomic distribution, are poorly-differentiated, and have inferior response to systemic chemotherapy when compared with mismatch-repair proficient (pMMR) tumors.¹⁴

Despite their worse response to chemotherapy, patients with early-stage dMMR CRC tend to have better survival outcomes, and dMMR has therefore been regarded as a positive prognostic feature.¹² More recent studies have challenged this, finding instead that in stage-IV CRC (mCRC), dMMR patients may fare worse.^15–18 A recent study using the National Cancer Database (NCDB) demonstrated worse survival in MMR-deficient mCRC compared with pMMR.¹⁴ How these findings apply to young-onset mCRC was not addressed, and the impact of dMMR on survival in this population remains incompletely understood. This study therefore sought to assess patterns of MMR testing and survival among young patients with mCRC in the NCDB, hypothesizing that survival would be worse in young patients and those with dMMR compared to older and pMMR patients.

Methods:

Design

This study used de-identified patient data and the Institutional Review Board deemed the study exempt from review and consent. Patients with stage-IV colon, rectal, and rectosigmoid junction tumors diagnosed in 2010–2016 were identified in the National Cancer Database (NCDB). Young adult patients were those ages 18–30 and were compared to older patients. This age range was chosen due to previous research identifying heterogeneity within the young-adult CRC population, with an increase in adverse prognostic features in patients 30 and younger.^2,4 The NCDB is maintained by the American Cancer Society and the American College of Surgeons Commission on Cancer and while not nationally representative, captures 70% of new cancers in the United States.¹⁹ This study period was chosen because 2010 was the first year offering data on both MMR status and sites of metastasis. Peritoneal metastases were defined as described.¹⁴ Patients with stage-IV disease, but with no metastases recorded in any other site (brain, bone, liver, lung, or distant nodal metastases), were defined as having isolated peritoneal metastases. Adenocarcinoma and mucinous histologies were included while neuroendocrine tumors and rare histologies were excluded.

Patient variables collected included age at diagnosis, sex, race, site (defined in NCDB as colon, rectum, or rectosigmoid junction), tumor grade, tumor sidedness when available, mucinous histology, presence of KRAS mutation, MMR-testing status and result, and presence of peritoneal metastases. The primary outcome was overall survival from diagnosis.

Statistical Analysis

Descriptive statistics were performed with Fisher-exact, Wilcoxon, and Chi-squared tests as appropriate. Clinicopathologic factors were compared based on patient age (older versus younger), whether mismatch repair testing was performed, and MMR-testing result (defined as deficient, proficient, or status-unknown). Analyses of right- versus left-sided tumors included only tumors with recorded sidedness by ICD code. Survival analyses evaluated association of patient factors with overall survival by Kaplan-Meier method with log-rank test for significance at p<0.05 as described.¹⁴ Analyses of survival based on MMR-status considered only patients who had MMR testing. Follow-up was determined by reverse Kaplan-Meier method.²⁰ P-values were corrected for multiple testing according to Benjamini-Hochberg method.²¹ A multivariable Cox model included factors with univariable p<0.1. Analyses used R version 3.6.1 (Vienna, Austria).

Results:

Of 124,587 stage-IV colorectal cancer patients in NCDB, 1,123 (0.9%) were 30 years of age or younger. Young adult patients in NCDB had median age of 27.0 (Interquartile Range [IQR] 25.0–29.0), and 48.5% were female (Table 1). Compared to patients older than 30, significantly greater proportions of young adults were Black, Hispanic, Asian, and races other than white (p<0.01 for all). Young patients had more rectal and rectosigmoid primaries than colon (34.6 vs 26.9%). Right-sided tumors were much less common among younger patients (25.2 vs. 42.1%). Young patients more often had KRAS testing (47.6 vs 32.8% with known status, p<0.001), and among tested patients, younger patients had lower rates of KRAS mutations (35.8 vs. 44.1%). Several adverse prognostic features were more commonly found in young adult compared to older adult patients. Young patients were more likely to have mucinous (16.4 vs 9.0%) and high-grade tumors (28.7 vs 20.5%), and isolated peritoneal metastases were more common than metastases to other sites among young adults (28.5 vs. 18.5%).

Table 1:

Demographic and clinical characteristics of stage-IV colorectal cancer patients by age.

	Age		P Value
Characteristic	Older (>30) n=123,464	Young (≤30 years) n=1,123
Age, median years (IQR)	65.0 (55.0–75.0)	27.0 (25.0–29.0)	<0.001
Sex, n (%) Female	57,389 (46.4)	545 (48.5)	0.2
Race, n (%)			<0.001
White	91,800 (74.6)	660 (59.1)	<0.001
Black	18,344 (14.9)	223 (20.0)	<0.001
Hispanic	7,550 (6.1)	158 (14.1)	<0.001
Asian	3,821 (3.1)	52 (4.7)	0.007
Other	1,493 (1.2)	24 (2.1)	0.009
Site, n (%)			<0.001
Colon	90,238 (73.1)	735 (65.4)	<0.001
Rectum	22,648 (18.3)	250 (22.3)	<0.001
Rectosigmoid junction	10,578 (8.6)	138 (12.3)	<0.001
Mucinous Histology, n (%)	11,079 (9.0)	184 (16.4)	<0.001
KRAS mutant, n (% of tested)	17,877 (44.1)	191 (35.8)	<0.001
Right-sided n (%)	38,032 (42.1)	185 (25.2)	<0.001
Grade, n (%)			<0.001
Low	5,417 (4.4)	37 (3.3)	0.1
Intermediate	56,097 (45.4)	491 (43.7)	0.25
High	25,314 (20.5)	322 (28.7)	<0.001
Unknown	36,636 (29.7)	273 (24.3)	<0.001
Isolated PM, n (%)	22,897 (18.5)	320 (28.5)	<0.001

Note: KRAS percentages are for those with known KRAS status. P-values represent comparisons between young patients and older patients. For all tables, P-values are false-discovery-rate adjusted for multiple testing where appropriate.

Abbreviations: IQR: Interquartile Range; PM: Peritoneal Metastases

When examining treatment received, young adult patients were more likely than older adults to receive any cancer treatment (94.9 vs. 86.1%), chemotherapy (88.2 vs. 66.3%), surgery of the primary site (53.9 vs. 48.2%), and immunotherapy (19.2 vs. 14.3%, p<0.001 for all). Use of immunotherapy was similar in young dMMR, pMMR, and MMR-unknown patients (19.3, 21.9, and 18.3%, respectively, p=0.5). The database does not record details on drugs used or duration of treatment. Immunotherapy use in young adult patients increased from 2.7% in 2010 to 39.6% in 2016.

Mismatch Repair Testing –

Among all stage-IV patients, only 16.8% had a mismatch repair testing result recorded in NCDB. Testing was more common in younger patients (29.1 vs. 16.6%), but more than 70% of young adult patients still had unknown MMR status (Table 2). Rates of testing increased over time. In 2010, only 17.8% of young adults had MMR testing, while 39.6% had a recorded MMR test result in 2016. Testing was more common for younger patients in all tumor site subsets, and in tumors with mucinous histology. In both young and old patients, colon and rectosigmoid tumors were tested more often than rectal tumors, and despite their association with mismatch repair deficiency, mucinous tumors were tested no more often than all colon tumors.

Table 2:

Patterns of Mismatch Repair testing in young versus older stage-IV colorectal cancer patients

	Age			P Value
	All n=124,587	Older (>30 years) n=123,464	Young (≤30 years) n=1,123
MMR Tested, n (%) Yes	20,878 (16.8)	20,551 (16.6)	327 (29.1)	<0.001
Colon		15,506 (17.2)	228 (31.0)	<0.001
Rectum		3,229 (14.3)	58 (23.2)	<0.001
Rectosigmoid junction		1,816 (17.2)	41 (29.7)	<0.001
Mucinous Histology		2,262 (20.4)	52 (28.2)	0.01

For each subgroup, numbers and percentages of patients with recorded mismatch repair testing are shown. P-values represent comparisons between young patients and older patients.

Abbreviations: MMR: Mismatch Repair

In patients with known MMR status, deficient mismatch repair was present in 17.1% (Table 3). In this cohort, young adult patients were equally likely as older patients to have dMMR (19.0 vs. 17.1%, p=0.4). There was no association between dMMR and primary tumor site in young patients (p=0.5), although older patients were more likely to have dMMR in colon as opposed to rectum and rectosigmoid tumors (18.9% of colon vs 11.1 and 13.0% p<0.001). When rates of dMMR were compared between young adult and older patients by primary tumor site and presence of mucinous histology, no significant differences existed. The highest rates of dMMR occurred in tumors with mucinous histology in both younger and older patients (26.9 and 27.8% of patients with testing).

Table 3:

Results of Mismatch Repair testing in young versus older stage-IV colorectal cancer patients

	Age			P Value
	All n=20,848	Older (>30 year) n=20,551	Young (≤30 years) n=327
dMMR, n (%)	3,580 (17.1)	3,518 (17.1)	62 (19.0)	0.4
Colon		2,924 (18.9)	46 (20.2)	0.6
Rectum		2,871 (11.1)	8 (13.8)	0.5
Rectosigmoid junction		236 (13.0)	8 (19.5)	0.2
Mucinous Histology		629 (27.8)	14 (26.9)	1

For each subgroup, numbers and percentages of patients who had mismatch repair testing and were found to have deficient mismatch repair are shown. P-values represent comparisons between young patients and older patients.

Abbreviations: dMMR: Deficient Mismatch Repair

Due to the large number of young patients with unknown MMR-status, clinical characteristics were compared between patients with known and unknown MMR-status to determine whether differences existed between these groups (Supplemental Table 1). Young patients with known MMR-status were similar to those with unknown MMR-status in distribution of age, sex, race, tumor site, tumor sidedness, mucinous histology, KRAS mutations, low and intermediate grade tumors, and isolated peritoneal metastases (p>0.05 for all). Patients with unknown MMR-status were more likely to have unknown grade (27.8 vs. 15.9%), and less likely to have high-grade tumors (25.9 vs. 35.5%, p<0.05 for both). Similar proportions of known and unknown MMR-status young patients received no treatment, chemotherapy, and immunotherapy, while surgery of the primary site was significantly more common in those with known MMR-status (69.7 vs. 47.4%, p<0.001).

Survival -

Despite these higher rates of adverse prognostic features, younger mCRC patients had better overall survival with a median of 20.7 months compared to 14.8 months in older patients (p<0.001; Table 4). Younger patients with colon and rectal primaries had significantly longer survival than older patients, while survival with rectosigmoid junction primaries was similar between the two age groups. When considered in terms of mismatch repair testing, older patients with unknown MMR-status had significantly shorter survival than younger unknown-status patients (median 13.4 vs. 19.5 months, p<0.001), while survival in pMMR and dMMR patients was similar between younger and older patients.

Table 4:

Overall survival in stage-IV colorectal cancer patients by age

	Age		P Value
	Older (>30 years) n=123,464	Young (≤30 years) n=1,123
Survival, median mo. (CI)	14.8 (14.7–15.0)	20.7 (19.4–22.4)	<0.001
Colon	13.2 (13.0–13.3)	19.4 (18.3–22.0)	<0.001
Rectum	19.6 (19.2–20.0)	24.5 (21.5–27.4)	0.02
Rectosigmoid junction	19.3 (18.6–20.0)	19.8 (18.4–23.7)	0.7
dMMR	19.7 (18.7–20.8)	16.6 (11.7–24.1)	0.4
pMMR	23.9 (23.4–24.4)	25.5 (22.6–30.0)	0.1
Unknown MMR	13.4 (13.2–13.5)	19.5 (18.7–21.7)	<0.001

P-values represent comparisons between young patients and older patients.

Abbreviations: mo.: months; CI:95% Confidence Interval; dMMR: deficient Mismatch Repair; pMMR: proficient Mismatch Repair; MMR: Mismatch Repair

To test the hypothesis that dMMR is associated with worse survival in young adult patients, clinical and tumor features were tested for univariable association with survival in young adult patients with known MMR-status (Table 5). In this analysis, mismatch repair deficiency correlated with an increased risk of death (HR 1.63, p=0.01; Figure 1). Grade and tumor sidedness also showed significant association with survival (p<0.05), while presence of isolated peritoneal metastases or mucinous histology showed non-significant correlation at p<0.1. After reverse stepwise selection, a parsimonious model based on 262 patients with complete data and 177 mortality events revealed that only low-grade (HR 0.44, 95% confidence interval [CI] 0.32–0.60), and deficient mismatch repair (HR 1.51, CI 1.04–2.19) remained independently associated with survival in young adult mCRC patients. Patients with low-grade tumors and pMMR had median survival of 41.1 months, compared to 22.4 months in dMMR low-grade tumors (Figure 2). Tumors that were not low-grade and had pMMR had a median survival of 19.6 months, while tumors not low-grade with dMMR had the worst survival at median 12.3 months, with survival significantly different across the 4 groups (p<0.001).

Table 5:

Multivariable Cox model of Overall Survival in patients ≤30 with known MMR-status

Factor	Univariable HR (95% CI)	Univariable P value	Multivariable HR (95% CI)	Multivariable P value
Peritoneal Metastases	1.34 (0.98–1.83)	0.07
MMR-Deficient	1.63 (1.12–2.37)	0.01	1.51 (1.04–2.19)	0.03
Site (Colon Ref.)	-	0.6
Rectum	0.86 (0.57–1.28)	-
Rectosigmoid junction	1.11 (0.71–1.72)	-
Sex (Female)	0.92 (0.68–1.23)	0.6
Age (per year)	1.02 (0.94–1.02)	0.3
Race (White Ref.)	-	0.9
Black	1.04 (0.70–1.55)	-
Other	1.09 (0.75–1.57)	-
Mucinous Histology	1.45 (1.01–2.09)	0.06
KRAS (Unknown Ref.)	-	0.8
KRAS WT	0.98 (0.70–1.37)	-
KRAS Mutant	1.13 (0.74–1.74)	-
Grade (Not Low)	2.32 (1.70–3.15)	6e-8	2.26 (1.66–3.08)	<0.001
Side (Right)	1.59 (1.09–2.34)	0.02

Note: For multi-level clinical factors, P-value of the reference represents the overall P-value for the category. The multivariable model includes MMR-status and grade. (n=262 patients).

Abbreviations: HR: Hazard Ratio for death; CI: 95% Confidence Interval; MMR: Mismatch Repair; Ref.: Reference; WT: Wild Type

Figure 1. Overall survival in stage-IV colorectal carcinoma patients age 30 and younger with known MMR status.

Blue solid line shows patients with proficient mismatch repair (pMMR). Red dashed line shows patients with deficient mismatch repair (dMMR). Patients with dMMR had worse survival than those with pMMR (median 16.6 vs 25.5 months, p<0.001).

Figure 2. Overall survival in stage-IV colorectal carcinoma patients age 30 and younger by grade and mismatch repair status.

Blue: Low-grade tumors with proficient mismatch repair (pMMR), median OS 41.1 months; Yellow: Non-low grade tumors with pMMR, median OS 19.6 months; Green: Low-grade tumors with deficient mismatch repair (dMMR), median OS 22.4 months; Purple: Non-low grade tumors with dMMR, median OS 12.3 months. Patients with low-grade dMMR tumors had significantly worse survival than low-grade pMMR tumors (p<0.001 for difference across groups).

Discussion:

In stage-IV CRC patients ≤ 30 years of age in NCDB, median survival with dMMR was 8.9 months shorter than with pMMR, indicating that dMMR represents a negative prognostic marker for young patients with metastatic CRC. Our results are consistent with literature demonstrating a higher prevalence of aggressive features including mucinous tumors and high-grade histology among younger patients.^8,22 Rates of dMMR were similar in younger and older patients. Young adults with stage-IV CRC represent a small but increasing and clinically-relevant population, given the greater proportion of young CRC patients that present with advanced disease.^2,4,7

The young adult population is not clearly defined in the colorectal cancer literature. Many studies use a cutoff of 40 years of age, but 50, 35, 30, and 25 have variably been used as well.^4,5 Despite varying cutoffs, authors have consistently described late-stage presentations, rectal or rectosigmoid disease, and aggressive pathologic features in young patients.²³ We chose a bound of 30 years due to findings by Chiang et al. and Li et al. who separately described heterogeneity within the young adult CRC population, with less favorable clinicopathologic characteristics and cancer-specific survival in patients 30 and younger.^2,4 Hayes-Jordan et al. similarly argue that very young CRC patients may display unique patterns of disease. Consistent with their findings in an institutional database, these NCDB results suggest a higher rate of isolated peritoneal metastases in the ≤30 mCRC population.⁵

Limitations of this study include the high proportion of patients with MMR-unknown status. Some of these patients undoubtedly had testing that went unrecorded in the database, yet with so many untested patients, it remains possible that selection biases for or against testing could influence survival results. To account for this high proportion of non-testing, clinical and treatment features were compared in young MMR-tested and untested groups (Supplemental Table 1), and were generally similar, except for a higher proportion of unknown-grade tumors and lower rates of surgery for the primary tumor in the MMR-untested group. Based on these results, we believe that survival findings in the MMR-tested group are likely generalizable to the untested patients, although it remains possible that unrecognized differences exist between these patients.

In older patients, untested patients had worse survival than tested patients,¹⁴ while in these younger patients, untested patients had survival intermediate between pMMR and dMMR patients (Table 4). A possible explanation for this phenomenon could be that older patients with metastatic cancer more often decline cancer-directed therapy and these patients may be less likely to receive testing. In the older cohort, 12.8% of patients are recorded as receiving no treatment, compared to only 3.7% of younger patients (p<0.001). Of older patients who received no treatment, median survival was 1.7 months and only 3.4% had MMR-testing. Older patients were also less likely to receive chemotherapy. Exactly how these testing biases would affect survival cannot be determined; however, our analyses of survival and MMR-status compared only patients who actually had testing to minimize effects of bias in the untested group.

Another limitation is the low number of young patients with dMMR, which limited statistical power in the multivariable model to define risk factors for worse survival. The reality may be that additional factors are independently associated with survival, as is true in older patients, but because of low patient and event numbers for young adults with mCRC even in this large national database, the model had insufficient power to detect it. Finally, while NCDB does allow comparison of rates of chemotherapy use, it does not record the drugs given or number of cycles received, which clearly could influence survival outcomes, and we are unable to account for these variables in this study.

Historically, dMMR has been believed to carry a better prognosis in CRC than pMMR, so much so that the World Health Organization classified all dMMR tumors as low-grade, regardless of histology.²⁴ This impression derived from findings of improved survival in dMMR patients in adjuvant chemotherapy trials, which included stage II and III patients.^12,14 Newer data question this, finding that dMMR patients with metastatic CRC have worse prognoses, and the present study extends these findings to the young adult mCRC population.^14,17,25 With both MMR and grade remaining as independent predictors of survival, our results further challenge the idea that all dMMR tumors behave as low-grade. In this population, patients with low-grade dMMR tumors had survival similar to high-grade pMMR tumors, indicating that dMMR status does not automatically predict a better prognosis. This agrees with the findings of Johncilla et al. who demonstrated that high-grade dMMR tumors often present at a later stage and follow a less favorable course than low-grade dMMR tumors, concluding that MMR status should not supersede histology in clinicopathologic grading of CRC.²⁴ The prevalence of high-grade histologic features in young patients including signet-ring, mucinous, and poorly-differentiated further supports the consideration of both MMR status and tumor architecture.

Recent data demonstrate worse overall stage-for-stage survival in young patients with colorectal cancer,^5,8,22 yet in this study, overall survival in mCRC patients ≤30 in NCDB was significantly better than for older patients. This apparent discrepancy may be due to differences in study design. Hayes-Jordan et al. used an institutional database, which may be influenced by referral patterns, and considered a younger population. In contrast, Poles et al. found that early-onset CRC patients (defined as 22–50 years) had better survival than older (>50) and pediatric (<22 years old) patients. Our young adult population with improved survival compared to older patients overlaps with Poles’ “early-onset” age group. Taken together, these results suggest that the true pediatric mCRC age group may have particularly poor outcomes, while very young adults fare better.

Although younger patients were twice as likely as older patients to have MMR-testing, over 70% of patients with metastatic CRC did not have MMR-status available in NCDB. The National Comprehensive Cancer Network has recommended MMR-deficiency testing in all CRC patients under age 50 since 2014, and universal testing since 2018.^26,27 Beyond this, other consensus statements from as early as 2008 recommended universal MMR-testing in CRC.²⁸ The low, but increasing rate of testing may reflect gradual adoption of these guidelines over time. Failure to obtain MMR-testing may also reflect and correlate with socioeconomic health disparities,²⁹ and our data especially support the rigorous testing of young patients.^13,30 Molecular studies of left-sided colorectal tumors have demonstrated greater frequency of mutations associated with CRC-related syndromes, such as Lynch syndrome, in younger patients.³¹ Deficient mismatch repair is increasingly recognized as a target for immunotherapy in colorectal cancer, with promising survival outcomes in stage-IV disease.^32–34 The landmark Keynote-177 trial shows dramatically improved progression-free survival in mCRC patients given pembrolizumab alone as first-line treatment, which will become the new standard of care for dMMR patients, and additional immunotherapy trials are ongoing.³⁵ Such a high rate of non-testing in this population suggests that many young patients with immunotherapy-treatable tumors are not recognized. The study period preceded widespread use of immunotherapy in dMMR CRC, and it will be interesting to see how survival with dMMR changes as immunotherapy becomes routine in these patients.

Conclusions:

Deficient mismatch repair is a poor prognostic factor in young patients with metastatic colorectal cancer with HR for death of 1.63 compared to intact MMR. Tumor grade also significantly correlates with overall survival, both alone and after multivariable adjustment for MMR status. Immunotherapy holds promise for management of MMR-deficient metastatic colorectal cancers, yet despite guidelines advising universal MMR testing, the majority of young patients in NCDB did not have MMR status reported. Increased MMR-testing will identify patients who may benefit from novel treatments.