Exploring the Melanoma and Pancreatic Cancer Phenotype of a Potential CDKN2A Founder Variant, I49T (c.146T>C; p.Ile49Thr), in Individuals of Predominantly Mexican Ancestry

Abstract

PURPOSE

Pathogenic/likely pathogenic variants (P/LPVs) in the CDKN2A gene cause an increased risk of melanoma (MEL) and pancreatic cancer (PANC). The CDKN2A variant I49T (c.146T>C), reported to be recurrent in Hispanics, has conflicting pathogenicity classifications at laboratories, affecting clinical care. Multiple genetics clinics collaborated to explore cancers associated with I49T.

METHODS

Institutional clinical databases were queried for the CDKN2A variants, I49T, known P/LPVs, and c.-2G>A (a benign variant [BV]), and history of PANC and MEL was abstracted. A combination of statistical tests was used to investigate cancer history associations.

RESULTS

Data on 203 individuals, with qualifying CDKN2A variants detected on multigene testing between 2012 and 2023, were analyzed (I49T, n = 101; known CDKN2A P/LPVs, n = 57; BV, n = 45). Those with I49T were 91% less likely to have MEL than known CDKN2A P/LPVs (odd ratio [OR] = 0.089 [95% CI, 0.031 to 0.025]; P < .001) and were also less likely to have PANC (OR = 0.45 [95% CI, 0.14 to 1.41]; P = .17). However, mean age at PANC diagnosis for I49T was 56.0 years, significantly younger than known CDKN2A P/LPVs (μ = 71.0 years; P = .026).

CONCLUSION

In the largest I49T study to date to our knowledge, MEL was significantly less frequent compared with known CDKN2A P/LPVs. Although a nonsignificant trend was observed for less PANC in I49T than known P/LPVs, individuals with I49T presented with PANC at a significantly younger age than those with known CDKN2A P/LPVs. The presence of I49T in Hispanics of mostly Mexican ancestry supports that it is a founder variant, relevant to understanding cancer risk in a large proportion of Hispanics in the United States.

INTRODUCTION

CDKN2A (cyclin-dependent kinase inhibitor 2A) is a tumor suppressor gene, and pathogenic/likely pathogenic variants (P/LPV) confer an increased risk of melanoma (MEL) and pancreatic cancer (PANC).^1-3 PANC lifetime risk ranges from 11% to 60% and MEL risk from 28% to 76%.^4-7 The International Cancer of Pancreas Screening Consortium and the National Comprehensive Cancer Network (NCCN) recommend PANC and MEL surveillance in individuals with CDKN2A P/LPVs.^8,9 However, CDKN2A I49T (c.146T>C; p.Ile49Thr) has conflicting pathogenicity classifications in ClinVar^10,11 as multiple commercial laboratories classify it as a LPV and others classify it as a variant of uncertain significance (VUS). This discordance can affect clinical care as cancer risk management depends on whether the laboratory classifies it as a VUS or LPV.^12,13

CONTEXT

• Key Objective

• What are the melanoma (MEL) and pancreatic cancer (PANC) risks associated with the CDKN2A variant, I49T (c.146T>C; p.Ile49Thr)?

• Knowledge Generated

• In our cohort, MEL was significantly less frequent in participants with I49T compared with known CDKN2A pathogenic/likely pathogenic variants (P/LPVs). A nonsignificant trend was observed for less PANC in I49T than other P/LPVs; however, individuals with I49T presented with PANC at a significantly younger age than those with other CDKN2A P/LPVs. The presence of I49T in Hispanics of mostly Mexican ancestry supports that it is a founder variant, relevant to understanding cancer risk in a large proportion of Hispanics in the United States.

• Relevance

• The data generated from this study add to the limited understanding and published literature regarding the cancer phenotype of I49T.

Most data informing CDKN2A cancer risks are derived from non-Hispanic Whites (NHW).^14-17 Notably, I49T has an allele frequency of 0.44% in Hispanic/Latinos (H/L) in genetics databases (gnomAD, dbGAP, and ExAC) and is reported as recurrent in individuals of Mexican ancestry.^18-21 H/L is a term applied to a heterogeneous group of individuals of Mexican, Central and Southern American, Caribbean, and Spanish ancestry who constitute one of the fastest-growing US demographics.^22,23 Given the prevalence of I49T, understanding its impact on cancer risk is key to informed care and imperative to increase opportunities for enhanced screening and cancer detection. To our knowledge, there are no large studies published on the cancer phenotype of I49T; therefore, a collaboration of cancer genetics clinics was formed to explore the clinical presentation of individuals with this variant. This effort will help guide care by addressing this gap in knowledge.

METHODS

This retrospective, multicenter study included six institutions with site-specific, institutional review board–approved protocols in which written informed consent was obtained or waived. A cross-center data-use agreement allowing for deidentified data sharing through a REDCap database was procured.^24,25 The database was developed by the study team and included birth year, sex assigned at birth, ethnicity, maternal and paternal ancestry, CDKN2A variant, variant lineage, year of testing, laboratory, indication for testing, personal cancer history, age at diagnosis, and history of PANC and MEL in first-, second-, and third-degree relatives. Data from genetic counseling documentation at the time of multigene-panel testing (MGPT) were extracted between September 2022 and September 2023.

Clinical databases of individuals who underwent genetic counseling and MGPT at each institution were queried to identify those with the following CDKN2A variants: I49T, known P/LPVs, and c.-2G>A. At the time of data collection, I49T was classified as a VUS or LPV by major commercial laboratories, and anyone with it was included, regardless of classification. The CDKN2A c.-2G>A variant, found in H/Ls at a population frequency of 0.0186, was included as a benign variant (BV) clinical comparator, as it was well represented in the queried databases and is not associated with cancer risk.^10,26 Given that at least one commercial laboratory classified it as a VUS, it could be queried within institutional databases. However, identification of individuals with the BV relied on MGPT reports generated during the study period, so ascertainment of c.-2G>A across all tested individuals was not possible, as not all laboratories reported it out.

In addition to having a qualifying CDKN2A variant, a pedigree documenting personal and family cancer history, obtained during genetic counseling before September 2023, was required for inclusion. Patients with more than one qualifying CDKN2A variant, and those with a qualifying CDKN2A variant and a P/LPV in a gene with elevated PANC or MEL risk, as per NCCN, were excluded from data analysis.¹²

To investigate the relationship between personal and/or family history of PANC and/or MEL and demographic and clinical characteristics, a combination of statistical tests were used. Chi-square testing was used to assess associations between categorical variables. For small sample sizes or low expected frequencies, Fisher's exact test was used, and Kruskal-Wallis testing was applied to continuous variables. The overall differences in mean age at PANC and MEL diagnosis were tested across the CDKN2A variant groups. In the event of a significant Kruskal-Wallis result, post hoc Tukey's honestly significant difference test with ANOVA for pairwise comparisons was used to determine which specific means differed significantly. Multinomial logistic regression modeling was used to evaluate the association between CDKN2A variant groups and the odds of personal and/or family history of MEL and/or PANC, while controlling for potential confounding variables. From the regression coefficients, odds ratios (ORs) with 95% CIs were calculated to quantify the association between each variant group and personal/family history of MEL/PANC, relative to a reference level. Cumulative incidence functions (CIFs) were used to estimate the age-specific risk of developing PANC or MEL, treating age at diagnosis as the time scale. Gray's test was used to compare CIFs between different genetic variant groups.

RESULTS

Overall, 219 individuals had qualifying CDKN2A variants detected on MGPT between 2012 and 2023: 108 with I49T, 61 with known CDKN2A P/LPVs, and 50 with the CDKN2A BV. Patients were from the following institutions: University of California, Los Angeles (n = 20, 9.1%), Stanford Medical Center (n = 34, 15.5%), University of California, San Francisco (n = 37, 16.9%), University of Texas, Southwestern (n = 49, 22.4%), and University of Southern California, Norris Comprehensive Cancer Center and Los Angeles General Medical Center (n = 80, 36.5%). Sixteen participants were excluded from the analysis: one had multiple CDKN2A P/LPVs, and 15 had a qualifying CDKN2A variant and a P/LPV in a PANC and/or MEL predisposition gene (Appendix Table A1). The remaining 203 individuals were analyzed: 101 in the I49T group, 57 in the known CDKN2A P/LPV group, and 45 in the BV group (Table 1). By variant group, 99% of I49T and 98% of the BV were H/L, while 19% of the known CDKN2A P/LPVs reported H/L ancestry (P < .001). Mexican ancestry was reported by 87% of the I49T group, 84% of the BV group, and 12% of the known CDKN2A P/LPV group (P < .001). Most patients were referred for counseling and testing for hereditary breast and ovarian cancer– (67%) or Lynch syndrome–related indications (20%).

TABLE 1.

Participant Demographic and Clinical Information, by CDKN2A Group

Characteristic	CDKN2A I49T (n = 101)	CDKN2A Known P/LPV (n = 57)	CDKN2A BV c.-2G>A (n = 45)	Total (N = 203)	P
Sex, female	84 (83.2%)	41 (71.9%)	42 (93.3%)	167 (82.3%)	.018
Hispanic/Latino	100 (99.0%)	11 (19.3%)	44 (97.8%)	155 (76.4%)	<.001
Non-Hispanic White	1 (1.0%)	44 (77.2%)	1 (2.2%)	46 (22.7%)
Other ethnicity		2 (3.5%)		2 (1.0%)
Mexican ancestrya	88 (87.1%)	7 (12.3%)	38 (84.4%)	133 (65.5%)	<.001
Mean age at genetic testing, years (range)	48.0 (22-73)	51.0 (27-81)	46.0 (21-75)	48.0 (21-81)	.058
Personal history of cancer	68 (67.3%)	40 (70.2%)	34 (75.6%)	142 (70.0%)	.61
MEL	5 (5.0%)	21 (36.8%)	0	26 (12.2%)	<.001
Mean age at 1st MEL diagnosis, years (range)	32.0 (25-55)	39.0 (16-70)	NA	38.5 (16-70)	.72
PANC	6 (5.9%)	7 (12.3%)	2 (4.4%)	15 (7.4%)	.24
Mean age at PANC diagnosis, years (range)	56.0 (44-66)	71.0 (60-80)	56.0 (45-67)	60.0 (44-80)	.027
Family history of MEL	9 (8.9%)	39 (68.4%)	1(2.2%)	49 (24.1%)	<.001
Family history of PANC	26 (25.7%)	21 (36.8%)	5 (11.1%)	52 (25.6%)	.013

Abbreviations: BV, benign variant; MEL, melanoma; NA, not applicable; PANC, pancreatic cancer; P/LPV, pathogenic/likely pathogenic variant.

^a Proportion of individuals with at least one parental lineage with self-reported Mexican ancestry.

Personal History of CDKN2A-Associated Cancer

MEL

There were 26 individuals with MEL in the cohort, all in the known CDKN2A P/LPV and I49T groups (n = 26/203, 12.2%, Table 2). Patients with MEL constituted 5.0% (n = 5/101) of the I49T group compared with 36.8% (n = 21/57) of the known CDKN2A P/LPV group (P < .001), such that those with I49T were 91% less likely to have MEL than those with known CDKN2A P/LPVs (OR = 0.089 [95% CI, 0.031 to 0.025]; P < .001). When comparing the cumulative incidence of MEL, the I49T group showed a lower probability of MEL at all time points compared with the known P/LPV group, while the BV group had no MEL diagnoses, remaining flat at zero. Gray's test confirmed that the differences were statistically significant (P < .001, Fig 1). There was no difference in mean age at diagnosis between the two groups, 32.0 years for the I49T group and 39.0 years for the known CDKN2A P/LPV group (P = .74; Table 2).

TABLE 2.

Comparison of MEL and PANC Personal and Family History Between CDKN2A Variants

Variable	Category	No. (%)	OR, 95% CI	P
Personal history of MEL (n = 26)	P/LPV	21/57 (36.8)	Reference	—
	I49T	5/101 (5.0)	0.089 (0.031 to 0.025)	<.001
	BV	0/45 (0)	NA
Personal history of PANC (n = 15)	P/LPV	7/57 (12.3)	Reference	—
	I49T	6/101 (5.9)	0.45 (0.14 to 1.41)	.17
	BV	2/45 (4.4)	0.33 (0.066 to 1.68)	.18
Family history of MEL (n = 49)	P/LPV	39/57 (68.4)	Reference	—
	I49T	9/101 (8.9)	0.045 (0.019 to 0.11)	<.001
	BV	1/45 (2.2)	0.010 (0.0013 to 0.082)	<.001
Family history of PANC (n = 52)	P/LPV	21/57 (36.8)	Reference	—
	I49T	26/101 (25.7)	0.59 (0.29 to 1.96)	.14
	BV	5/45 (11.1)	0.21 (0.073 to 0.63)	.004
Personal and family history of MEL (n = 22)	P/LPV	20/57 (35.1)	Reference	—
	I49T	2/101 (2.0)	0.019 (0.004 to 0.089)	<.001
	BV	0/45 (0)	NA
Personal and family history of PANC (n = 6)	P/LPV	3/57 (5.3)	Reference	—
	I49T	3/101 (3.0)	0.44 (0.085 to 2.32)	.34
	BV	0/45 (0)	NA
Personal and family history of MEL or PANC (n = 29)	P/LPV	23/57 (40.4)	Reference	—
	149T	6/101 (6.0)	0.04 (0.013 to 0.12)	<.001
	BV	0/45 (0)	NA

Abbreviations: BV, benign variant; MEL, melanoma; NA, not applicable; OR, odds ratio; PANC, pancreatic cancer; P/LPV, pathogenic/likely pathogenic variant.

FIG 1.

Cumulative incidence of MEL for CDKN2A variant groups. MEL, melanoma; P/LPV, pathogenic/likely pathogenic variant.

PANC

There were 15 individuals with PANC across the three groups (n = 15/203, 7.4%, Table 2). This constituted 12.3% (n = 7/57) of individuals in the known CDKN2A P/LPV group, 5.9% (n = 6/101) of those in the I49T group, and 4.4% (n = 2/45) of those in the BV group; post hoc pairwise comparisons were nonsignificant (P = .23). Although those with I49T were 55% less likely to have PANC than those with known CDKN2A P/LPVs, this difference was not statistically significant (OR = 0.45 [95% CI, 0.14 to 1.41]; P = .17). When comparing the cumulative incidence of PANC between the I49T and known CDKN2A P/LPV groups, using age at diagnosis as the time scale, Gray's test indicated that the difference was not statistically significant (P = .21, Fig 2). The same analysis, when done for all three groups, did not show a statistically significant difference (P = .41; Appendix Fig A1). However, the mean age at diagnosis for PANC in I49T was 56.0 years, significantly younger than known CDKN2A P/LPVs whose mean diagnosis age was 71.0 years (P = .027, Table 1).

FIG 2.

Cumulative incidence of PANC for I49T and known CDKN2A P/LP variants. PANC, pancreatic cancer; P/LP, pathogenic/likely pathogenic.

Family History of CDKN2A-Associated Cancer

MEL

Forty-nine patients reported a family history of MEL across the three groups (n = 49/203, 24.1%): 68.4% (n = 39/57) of those in the known CDKN2A P/LPVs group, 8.9% (n = 9/101) of the I49T group, and only 2.2% (n = 1/45) of the BV group (P < .001; Table 1). Patients with I49T were 95.5% less likely to report a family history of MEL than those with known CDKN2A P/LPVs (OR = 0.045 [95% CI, 0.019 to 0.11]; P < .001; Table 2). Similarly, a family history of MEL was 99.0% less likely to be reported in the BV group than in the known CDKN2A P/LPVs group (OR = 0.010 [95% CI, 0.0013, 0.082]; P < .001; Table 2; Figs 3A and 3C).

FIG 3.

(A) CDKN2A I49T variant (n = 101). (B) Known CDKN2A P/LPVs (n = 57). (C) CDKN2A c.-2G>A variant (n = 45). MEL, melanoma; PANC, pancreatic cancer; P/LP, pathogenic/likely pathogenic.

PANC

Fifty-two patients reported a family history of PANC across the three groups (n = 26/101, 25.7%): 36.8% (n = 21/57) of the known CDKN2A P/LPV group, 25.6% (n = 52/203) of the I49T group, and 11.1% (n = 5/45) of the BV group (P = .013; Table 1). The odds of reporting a family history of PANC were lower in the I49T group compared with the known CDKN2A P/LPV group but did not reach significance (OR = 0.59; 95% CI [0.29, 1.96]; P = .14; Table 2). A family history of PANC was 79.0% less likely to be reported in the BV group than in the known CDKN2A P/LPVs group (OR = 0.21 [95% CI, 0.073 to 0.063]; P = .004; Table2).

Personal and Family History of CDKN2A-Associated Cancer

MEL

Twenty-two (n = 22/203, 10.8%, Table 2) individuals across the cohort reported a personal and family history of MEL, and 90.9% (n = 20/22) were in the known CDKN2A P/LPV group. Over a third (n = 20/57, 35.1%) of the known CDKN2A P/LPV group had a personal and family history of MEL compared with 2.0% (n = 2/101) of the I49T group and no one (n = 0/45) in the BV group. Those with I49T were 80% less likely to have a personal and family history of MEL than those with known CDKN2A P/LPVs (OR = 0.019 [95% CI, 0.004 to 0.08]; P < .001).

PANC

Only six patients (n = 6/203, Table 2) reported both a personal and family history of PANC—three in the known CDKN2A P/LPV and three in the I49T group. Comparatively, this constituted 5.3% (n = 3/57) of the known CDKN2A P/LPV group and 3.0% (n = 3/101) of the I49T group. Although individuals in the I49T group were 56% less likely to have both a personal and family history of PANC than those in the known CDKN2A P/LPV group, this difference was not significant (OR = 0.044 [95% CI, 0.085 to 0.2.32]; P = .34).

MEL or PANC

Twenty-nine individuals (n = 29/203, 14.3%, Table 2) had a personal and family history of either CDKN2A-associated cancer and most (n = 23/29, 79.3%) were in the known CDKN2A P/LPV group. Only 6.0% (n = 6/101) of individuals in the I49T group had a personal and family history of either MEL or PANC, compared with 40.4% (n = 23/57) of those in the known CDKN2A P/LPV group. Those with I49T were 96% less likely to have a personal and family history of MEL or PANC than those with known CDKN2A P/LPVs (OR = 0.04 [95% CI 0.013 to 0.12]; P < .001).

DISCUSSION

To our knowledge, this is the largest study of CDKN2A I49T to date, and we found differences in the frequency of MEL and PANC when compared with individuals with known P/LPVs in CDKN2A. Individuals with I49T were much less likely to have a personal or family history of MEL than those with known CDKN2A P/LPVs. This finding may be explained by the difference in the ethnic background of the I49T group, which is predominantly H/L, compared with the known CDKN2A P/LPV group, which is predominantly NHW. The lifetime risk of MEL in H/Ls is estimated to be 0.5%, compared with 3% in NHWs.^27,28 This difference in risk is primarily due to variations in skin type and melanin content, which plays a photoprotective role by reducing the impact of ultraviolet radiation (UVR) exposure.^29,30 Other important risk factors remain understudied in H/Ls, including genetic susceptibility.^29,31,32 Notably, H/Ls are underrepresented in CDKN2A literature to date, but studies have shown that factors influencing MEL risk in the general population, including UVR exposure, sunburn history, nevus presence, and lighter skin tone, play an important role in CDKN2A-associated MEL risk.^33-38 The incidence of MEL is increasing in US H/Ls, and those diagnosed have higher morbidity and mortality compared with NHWs.^39-46 Misconceptions about MEL in H/Ls and health care barriers may result in later stages of diagnosis.^47-51 Therefore, H/Ls predisposed to MEL by having a CDKN2A P/LPV represent a group who may uniquely benefit from increased education and targeted interventions.

With respect to PANC, we found that the mean age at diagnosis was significantly younger in I49T compared with known CDKN2A P/LPVs (mean age of 56.0 years v 71.0 years), despite a nonsignificant trend toward lower rates of personal and family history of PANC. These results should be interpreted with caution, given the small number of PANC events (n = 15), which makes age-based comparisons statistically unstable. Also, it is important to note that the age at MGPT was 48 years and did not differ by variant group. This is young, as the average age at PANC diagnosis in the general population is 70 years, and a decade younger for individuals with CDKN2A P/LPVs.^52-55 As this study was retrospective, it does not account for any PANC diagnosed after genetic counseling, which could lead to an underappreciation of risk. Additional context for these findings comes from population-based data of PANC incidence in H/Ls, which, although limited, supports a lower incidence compared with NHWs.^56,57 Given that the I49T group is almost exclusively H/L, population-level differences may partially contribute to observed trends.

This underscores the need to explore these findings in larger, diverse cohorts and to do so prospectively to better understand differences and their drivers. These studies are necessary to refine PANC and MEL risk estimates and determine if modification of screening intervals or age at initiation is indicated.^52,58 Until then, cancer screening and prevention, as per standard CDKN2A protocols, should be offered to individuals with I49T. Existing literature estimates the lifetime risk of PANC in individuals with P/LPVs in CDKN2A to be approximately 12-fold that of the general population, which is approximately 1.5%.⁵³ Given the elevated risk, NCCN supports PANC screening, regardless of family history, starting at age 40 years for CDKN2A.^8,9 Individuals with I49T may represent a group of CDKN2A P/LPV carriers at risk for earlier-onset PANC, benefiting from initiation of screening at this age.

Like MEL, most of the current research on CDKN2A PANC risk has been in NHW cohorts; therefore, it is unclear what role, if any, race and ethnicity may play in PANC risk in CDKN2A.^5,54,55 A recent study from Rodriguez-Olivares JL and colleagues, based in Mexico City, found that 4.4% (n = 6/137) of their patients with PANC had a CDKN2A P/LPV, and five of those (n = 5/6, 83.3%) were I49T.²¹ Although this study did not estimate gene-specific PANC risk, it does highlight the frequency of this variant in cancer cohorts of Mexican ancestry and the need for more studies to clarify its contribution to PANC predisposition. Recent studies support an increase of early-onset PANC, diagnosed under age 55 years, in H/L women, including Samaan et al, where they had the highest incidence of PANC, compared with NHWs and Blacks.^52,57 Further investigations of this trend are warranted, especially in individuals with PANC predisposition.

Beyond defining cancer risks, future studies focused on the perspectives of individuals with I49T regarding its conflicting classification, associated cancer risks, and recommended screening can inform best practices for communicating complexities, providing cancer education, and supporting CDKN2A families.^39,59 A large, commercial laboratory recently upgraded I49T from VUS to LPV, leading to the recontacting of patients, often years after genetic testing. Although this change in classification guides future screening and prevention strategies, the interim years, when those strategies were not indicated, represent a missed opportunity for cancer-risk management in patients and their families. Recontact can be challenging as patients may not be in the health care system, living, or reachable. Given that VUS rates are higher in minority groups and genomic data less available, the time to definitive reclassification, either to a BV or P/LPV, may be delayed, decreasing the possible preventive power of genetic testing in these groups.^60-63 Understanding subsequent behaviors, screening uptake, and adherence in individuals with I49T may identify facilitators and barriers to inform future interventions and communication strategies for other VUSs that are reclassified.

Most individuals with I49T self-reported Mexican ancestry, reinforcing that this is a founder variant.^18,21,64-66 Individuals of Mexican ancestry make up 60% of the US H/L population or an estimated 37 million people.^67-69 Given the allele frequency of I49T in databases, this could translate to more than 175,000 people in the United States with this variant. Understanding I49T cancer risks, especially given its prevalence, could increase opportunities for enhanced screening and cancer detection in a sector of the population with known suboptimal cancer outcomes.^70,71 For providers counseling individuals of Mexican ancestry about MEL and PANC risk, awareness of I49T can inform counseling and laboratory selection to avoid uncertain results.

Despite this being the largest study of I49T to our knowledge, there are limitations, including the small sample size of some groups making those findings preliminary in nature, meriting exploration in larger cohorts with different study designs. Stratification of the known CDKN2A P/LPV group by ethnicity would be of value but was not feasible due to the low number of cancer events among the H/L individuals with known CDKN2A P/LPVs, limiting the interpretability of subgroup comparisons.

Patients in this study were clinically ascertained, predominantly female, and most were referred for hereditary breast and ovarian cancer (67%) or Lynch syndrome (20%) indications, not specifically for CDKN2A-associated MEL and PANC. There are inherent biases based on referral indication and the cancers in the probands and their family members. Although the clinical setting allowed for the extraction of cancer history from three-generation pedigrees, it was based on patient report, which can result in underreporting of cancers in their second- and third-degree relatives.⁷² The sensitivity of family history collection in Spanish-speaking H/Ls has been reported to be lower, possibly driven by assumptions made by language-discordant providers, miscommunication during interpreter-mediated sessions, and abbreviated history extraction when interpretation is used.^72,73 Although language concordance was not collected for this cohort, several of the participating institutions have bilingual genetic counselors, which may overcome linguistic challenges reported in other studies. The ancestry of the individuals with known P/LPV CDKN2A variants was primarily non-Hispanic. Therefore, it is challenging to account for the potential effect of ancestry on cancer risks of those with known P/LPV CDKN2A variants, as well as population-level differences in MEL and PANC risk that may exist.

The clinical data analyzed were based on what was known at the time of counseling and MGPT. Future longitudinal and prospective studies are needed to capture cancer outcomes over time. Finally, the ascertainment of the BV was dependent on laboratory-specific reporting during the study period. Individuals tested at laboratories classifying this variant as benign may have carried it without our knowledge. Although this may limit the use of the BV group, it is unlikely to confound the primary analyses comparing the I49T and known P/LPV group, as the BV is not associated with cancer risk. Importantly, our primary analyses are not dependent on the BV group.

In conclusion, in the largest I49T study to our knowledge, MEL was significantly less common in I49T than in known P/LPVs in CDKN2A, and a nonsignificant trend for fewer PANC was observed. Strikingly, individuals with I49T presented with PANC at a significantly younger age than individuals with known CDKN2A P/LPVs. I49T was identified in H/Ls of mostly Mexican ancestry, supporting its role as a founder variant relevant to informing cancer risk in a large proportion of H/Ls in the United States and Mexico. More research is needed to refine cancer associations, penetrance, and to achieve concordant variant classification across laboratories. Clarifying the clinical impact of I49T is vital for risk-appropriate management and for reducing existing cancer disparities faced by many H/L. This study advances our understanding and adds to the literature regarding the cancer phenotype of I49T. Furthermore, it highlights areas of focus for future research.

APPENDIX

TABLE A1.

Participants Excluded Based on Multiple Variants (n = 16)^a

CDKN2A Variant Group	Other PANC Risk Gene With P/LPV	H/L	Personal Cancer History	Family History of Cancer
BV	APC	Y	Ovarian	Stomach and rectal cancer
BV	BRCA1	Y	Breast and ovarian	Mother with breast
BV	MLH1	Y	Breast	Colorectal, stomach, and breast cancer
BV	TP53	Y	Breast and sarcoma	None
BV	TP53 and BRCA2	Y	Breast	Mother with breast and ovarian, relatives with stomach and breast
I49T	APC	Y	Thyroid and colon polyps	Colorectal and colon polyps
I49T	ATM	Y	Pancreatic	Pancreatic in sibling
I49T	BRCA1	Y	None	Pancreatic in sibling
I49T	BRCA1	Y	Breast	Bone cancer
I49T	BRCA1	Y	None	Pancreatic in second-degree relative
I49T	BRCA2	Y	None	Concerning for hereditary breast and ovarian
I49T	MLH1	Y	Colorectal	Concerning for Lynch syndrome
I49T	Other CDKN2A P/LPV	Y	MEL and uterine	Second-degree relative with pancreatic
Known CDKN2A P/LPV	ATM	N	MEL and breast	Concerning for hereditary breast and ovarian
Known CDKN2A P/LPV	BRCA2	N	Prostate	None
Known CDKN2A P/LPV	BRCA2	N	None	MEL in second-degree relative

Abbreviations: BV, benign variant; H/L, Hispanic/Latino; MEL, melanoma; PANC, pancreatic cancer; P/LPV, pathogenic/likely pathogenic variant.

^a No P/LPVs in other PANC and melanoma risk genes were identified in conjunction with a CDKN2A variant of interest.

FIG A1.

Cumulative incidence of PANC for CDKN2A variant groups. PANC, pancreatic cancer.

AUTHOR CONTRIBUTIONS

Conception and design: Daisy Hernandez, Ashlie Browning, Amber Gemmell, Karlena Lara-Otero, Mariana Niell-Swiller, Jacqueline Mersch, Kathryn A. Mraz, Amie Blanco, Jacob G. Comeaux, Julie O. Culver, James M. Ford, Natalia Gutierrez, Ming Li, Charité N. Ricker

Financial support: Charité N. Ricker

Administrative support: Ashlie Browning, Jacob G. Comeaux, Jenny Lester, Charité N. Ricker

Provision of study materials or patients: Daisy Hernandez, Ashlie Browning, Daniella Kamara, Amie Blanco, Jacob G. Comeaux, James M. Ford, Natalia Gutierrez, Beth Y. Karlan, Charité N. Ricker

Collection and assembly of data: Daisy Hernandez, Ashlie Browning, Amber Gemmell, Karlena Lara-Otero, Mariana Niell-Swiller, Jacqueline Mersch, Kathryn A. Mraz, Daniella Kamara, Amie Blanco, Emmeline Chang, Jacob G. Comeaux, James M. Ford, Natalia Gutierrez, Jenny Lester, Beth Y. Karlan, Jesus Resendiz, Charité N. Ricker

Data analysis and interpretation: Daisy Hernandez, Ashlie Browning, Amber Gemmell, Karlena Lara-Otero, Mariana Niell-Swiller, Jacqueline Mersch, Kathryn A. Mraz, Daniella Kamara, Yvonne Cardona, Beth Y. Karlan, Ming Li, Qi Nie, Maria L. Wei, Charité N. Ricker

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/po/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).