Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2022 Dec 1.
Published in final edited form as: Lancet Oncol. 2021 Nov 12;22(12):1787–1798. doi: 10.1016/S1470-2045(21)00580-5

Cancer Incidence, Patterns, and Genotype-Phenotype Associations in an Observational Cohort Study of Individuals with Pathogenic or Likely Pathogenic Germline TP53 Variants

Kelvin Cesar de Andrade 1,*, Payal P Khincha 1,*, Jessica N Hatton 1, Megan N Frone 1, Talia Wegman-Ostrosky 1,2, Phuong L Mai 1,3, Ana F Best 4, Sharon A Savage 1
PMCID: PMC8915249  NIHMSID: NIHMS1758686  PMID: 34780712

Abstract

Background:

Li-Fraumeni syndrome (LFS), caused primarily by pathogenic/likely pathogenic (P/LP) germline TP53 variants, is a variably penetrant, rare cancer predisposition syndrome with very high risks of cancer starting in childhood, including the risk of multiple primaries over the lifespan. This study aimed to characterize and quantify cancer incidence, patterns, and genotype-phenotype associations in individuals with P/LP germline TP53 variants.

Methods:

This observational cohort study was conducted on 480 carriers of P/LP germline TP53 variants enrolled in the National Cancer Institute’s referral-based longitudinal LFS study between August 1, 2011 and March 24, 2020. Data on personal and family history of cancer were obtained through study questionnaires and validated by medical records. Variants were categorized based on both loss-of-function (LOF) and dominant-negative effect (DNE) properties. LFS-associated cancer incidences were compared with the general population using the Surveillance, Epidemiology, and End Results (SEER) 1975–2017 registry. Cancer incidences were evaluated using family-clustered Cox-regression models and competing risk methods.

Findings:

The standardized incidence of any cancer in carriers of P/LP germline TP53 variants compared with SEER is highest up to age 30 years (>60-times higher), remaining ~10-times higher after age 50 years. In females, the probability of a non-breast first cancer, considering breast cancer as a competing risk, was substantially lower than that of any first cancer (24·4% vs 50·4% by age 33·7 years). Overall, DNE_LOF and notDNE_LOF variants were associated with earlier age at first and second cancer compared with notDNE_notLOF and DNE_notLOF variants. The time interval from first to second cancer was shorter among carriers whose first cancer diagnoses were later in life. Multiple cancers were diagnosed within a short time frame in some participants.

Interpretation:

This study adds important granularity to the understanding of cancer incidence and patterns in individuals with P/LP germline TP53 variants. Integration of age-range-specific cancer incidence estimates, cancer-free survival by functional variant group, impact of risk-reducing mastectomy on female cancer incidence, and data on subsequent malignancy will be important as strategies are developed to optimize cancer screening and management for these individuals.

Funding:

Intramural Research Program, Division of Cancer Epidemiology and Genetics, NIH.

INTRODUCTION

Li-Fraumeni syndrome (LFS; OMIM 151623) is an autosomal dominant, variably penetrant inherited cancer predisposition disorder characterized by elevated risks of cancers, beginning in infancy. The most frequent LFS-associated ‘core’ cancers are pre-menopausal breast cancer, osteosarcoma and soft-tissue sarcomas (STS), brain tumors, and adrenal cancer among many others, most occurring at ages earlier than expected.1,2 The median age at first cancer diagnosis in LFS is approximately 31 years in females and 46 years in males. Individuals with LFS are also at risk of developing multiple primary malignancies.3 Traditionally, LFS is a clinical diagnosis based on personal and family history of cancer. Multiple criteria exist to diagnose LFS and to identify individuals who should be tested for pathogenic/likely pathogenic (P/LP) germline TP53 variants, the only known genetic cause of LFS to date.1,2 Due to the high cancer risks and heterogeneity, current cancer screening recommendations aimed at early cancer detection are multimodal, high-frequency evaluations centered around whole-body MRI.4,5 Cancer screening in LFS has been shown to improve survival, but may induce anxiety, uncertainty, and screening burden.6,7

Variants in TP53 affect different functions of the p53 protein and, consequently, its tumor suppressive activity. TP53 variants can be grouped based on their functional consequences: 1) variants associated with loss-of-function (LOF) leading to haploinsufficiency; 2) variants that endow the p53 protein with gain of functions (GOF); and 3) variants associated with a dominant-negative effect (DNE) and impairment of transactivation activities.8 Variable phenotypes, expressivity, and penetrance of cancer are frequent in carriers of P/LP germline TP53 variants.9 The shift from single-gene testing to large multi-gene cancer panels (almost of all which include TP53), has led to the unexpected identification of individuals with P/LP germline TP53 variants in the absence of typical cancer family history.10 A limited number of genotype-phenotype studies have evaluated how TP53 variant functional properties impact cancer risk.8 For example, DNE variants are typically associated with earlier age at cancer onset and considered highly penetrant11,12 while LOF variants are more likely to occur in families meeting classic clinical LFS criteria.13 While previous studies have directly compared DNE with LOF variants based on functional assay-specific characteristics, these properties are not mutually exclusive. A recent assay investigated these properties for TP53 variants, stratifying variants based on DNE and LOF properties, allowing the assessment of both features simultaneously.14

Further refinement and quantification of cancer risks and patterns are needed to optimize the care of P/LP germline TP53 variant carriers. This study comprehensively evaluated cancer incidence, patterns, and genotype-phenotype associations by functional TP53 groups in a large cohort of individuals with P/LP germline TP53 variants to facilitate development of personalized cancer risk assessment.

METHODS

Study Participants and Data Collection

Individuals in this study were part of the National Cancer Institute’s (NCI’s) Institutional Review Board (IRB)-approved longitudinal LFS Study (NCT01443468; http://lfs.cancer.gov) and enrolled between August 1, 2011 and March 24, 2020. Participants or their legal guardians signed informed consent, completed questionnaires, and provided medical records, including pathology and genetic testing reports among others.3 Metastatic cancers and recurrences were excluded. We included confirmed and obligate carriers of germline TP53 variants classified as P/LP (hereby referred to as individuals with LFS) in ClinVar15 by one or more major genetic testing laboratories or by the ClinGen TP53 Variant Curation Expert Panel (VCEP).16 Exclusion criteria were (1) confirmed germline mosaics or individuals with low variant allele fraction suggestive of mosaicism or clonal hematopoiesis (n = 9); (2) individuals with P/LP germline TP53 variants and personal and/or familial cancer history suggestive of another hereditary cancer predisposition syndrome (n = 1); and (3) individuals who were confirmed to carry additional P/LP germline variant(s) in a different high-penetrance cancer susceptibility gene (n = 1).

TP53 Variant Categorization

Missense and nonsense variants were classified into four groups based on a recently published systematic functional assay.14 Thresholds to determine DNE (p53WT_Nutlin, Z-score ≥ 0·61) and LOF (p53NULL_Etoposide, Z-score ≤ −0·21) properties were utilized as currently adopted by the International Agency for Research on Cancer (IARC) TP53 database (version R20, July 2019)17 and by the ClinGen TP53 VCEP revised classification criteria.16 Based on the combination of scores, functional groups were designated: DNE_LOF, notDNE_LOF, notDNE_notLOF, and DNE_notLOF. A fifth group designated as “Not Included”, consisted of variants that were not investigated in the original assay (e.g., insertions, deletions, frameshifts, and splice-site variants). For completeness, additional analyses comparing DNE vs. notDNE, and LOF vs. notLOF were performed for time-to-event analyses examining age at first cancer and time interval between first and second cancer. Of note, these individual comparisons should be carefully interpreted given the strong correlation between LOF and DNE positivity reflected by the P/LP germline TP53 variants included in this study.

Statistical Analyses

Outcomes assessed included order of cancer occurrence, cancer incidence compared with the general population, overall survival, cumulative incidence of first cancer diagnoses, cancer patterns including age at first cancer, age at second cancer, time interval between first, second and subsequent cancers, and probability of death after first cancer before the development of a second cancer. Cancer types were grouped into 16 morphologic categories. Non-melanoma skin cancers and HPV-associated high-grade dysplasia of the anus, cervix, and vulva were excluded. Synchronous cancers in the same individual were treated as one event for all non-cancer specific time-to-event analyses, and as independent events for order of cancer occurrence, cumulative and standardized incidences, and first-cancer-specific survival analyses. For the overall survival (OS) analyses, participants were censored at last follow-up. For time-to-event analyses of age at first cancer, individuals were censored at death or last follow-up. In some time-to-event analyses, sex-specific cancers (including breast, gynecological, and prostate cancers) were jointly considered as a competing event to other first-cancer diagnoses for direct comparisons of females with males, and to ensure that proportionality of cause-specific hazards was a reasonable assumption between groups (assessed with Chi-Square tests, Appendix p 1). For cancer-specific estimates, other first cancer diagnoses were treated as competing risks to avoid any potential treatment-associated bias. For breast cancer-specific analyses, prophylactic bilateral mastectomy was considered as an additional competing risk, when data were available. Sensitivity analyses were performed for time-to-event analyses for breast cancer as a first malignancy, excluding the women for whom mastectomy data were incomplete. Time-to-event analyses examining second-cancer risk included analysis of the time interval from first to second cancer diagnosis, as well as age at second cancer. Analyses of age at second cancer used delayed entry (entry at first cancer diagnosis) methods adapted for left-truncated and right-censored survival data, and all analyses accounted for death prior to second cancer as a competing risk.

Correspondingly, analyses of the estimated probability of death after first cancer diagnosis but prior to second cancer diagnosis were conducted accounting for second cancer diagnosis as a competing risk. All nonparametric survival or cumulative incidence estimates additionally accounted for family clustering due to potential shared risk factors within families. Family-clustered proportional cause-specific-hazards models were used to compare survival between groups. Baseline groups were established based either on the largest category (i.e., DNE_LOF) or on the ease of interpretation (i.e., “0–17” age-range category). Survival curves were created with Kaplan-Meier (KM).18 Cumulative incidence with competing risks19 were calculated using the Aalen-Johansen estimator.20 P-values less than 0.05 were considered statistically significant. Analyses and plotting were conducted using Python 3 (version 3.8.3) programming language utilizing the Lifelines library (version 0.25.8) and R version 4.0.2 with the survival package version 3.1.12 and tidyverse.

Standardized incidence ratios (SIRs) were obtained by comparison of observed cancer incidences in the cohort against expected incidences based on the Surveillance, Epidemiology, and End Results (SEER) 1975–2017 data (seer.cancer.gov). Individuals were stratified by sex, 5-year attained age group, and 5-year attained calendar year group. Individuals with missing date of birth or date of last follow up were excluded from analyses, resulting in 469 individuals analyzed (Appendix p 2 includes ICD-O definitions for each cancer category).

Role of the Funding Source

The funder of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. All authors had access to the data and the corresponding author had final responsibility to submit for publication.

RESULTS

The study included 480 individuals with P/LP germline TP53 variants from 143 families, after exclusion of nine individuals with low allele fraction, one individual with family history suggestive of another hereditary cancer predisposition syndrome, and one individual with an additional P/LP germline variant in BRCA1. The majority of participants were female (299/480, 62·3%). Clinical familial criteria met were: 33/143 (23·0%) families with Classic21 LFS, 56/143 (39·2%) Chompret22,23, 11/143 (7·7%) Revised Chompret 201511, 10/143 (7·0%) Eeles24, 1/143 (0·7%) Birch25, and no criteria met in 32/143 families (22·4%). The majority of participants on whom information was available were of European ancestry (n = 307/322; 95·3%).

There were 95 individuals aged 0–17 years, 84 were 18–29 years, 141 were 30–44 years, and 148 were >45 years old at last known follow-up/death. Of these, 28/95 (29·5%), 49/84 (58·3%), 99/141 (70·2%) and 123/148 (83·1%) had been diagnosed with at least one cancer, respectively. Data on age at last follow-up or death were not available for 12 participants. Median age at last follow-up or death was 36·7 years (Q1: 25·4, Q3: 47·2 years) for females and 34·3 years (Q1: 17·8, Q3: 52·5 years) for males. The majority of individuals (305/480, 63·5%) had at least one cancer (Table 1). Of the 619 cancers diagnosed, 62·0% (384/619) were validated by medical records, 54·0% (334/619) were pathology-confirmed. The most commonly observed cancers were breast cancer (188/619 cancers in 140 females, 30·3%), STS (124/619, 20·0%), brain cancer (62/619, 10·0%), osteosarcoma (41/619, 6·6%), and hematologic cancers (33/619, 5·3%) (Table 1). A total of 117 participants, including 64 females and 53 males, had died as of the end of the study period. The median OS of study participants was 65·5 years (95% CI, 60·8–68·9 years), with median OS of 66·0 years (95% CI, 62·4–71·3 years) in females and 63·3 years (95% CI, 57·0–68·9 years) in males (Appendix p 3).

Table 1.

NCI LFS cohort demographics stratified by functional variant group, gender, and cancer status. Data as of study close date (March 24th, 2020). A total of 10 cancer types were categorized as “Other” and includes: unknown primary (3), head and neck cancer (2), carcinoid tumor (2), thymus cancer (2), bladder cancer (1), parotid gland cancer (1), neuroblastoma (1), malignant peripheral nerve sheath tumor (1), mesothelioma (1), and chordoma (1). Abbreviations: LOF, loss-of-function; DNE, dominant-negative effect, GI, gastrointestinal.

Functional Variant Group
Cancer Status DNE_LOF Not included nothNE_notLOF nothNE_LOF DNE_notLOF Total
Female Male Female Male Female Male Female Male Female Male
Unaffected individuals (n) 40 27 22 25 28 20 1 4 5 3 175
Median Age at Last Follow-Up (range) 20·4 (3·1−66·0) 19·0 (1·4−58·2) 18·4 (1·3−60·7) 20·9 (3·1−73·0) 30·2 (0·7−68·8) 31·4 (0·4−76·7) 13·9 (13·9−13·9) 17·3 (15·1−46·5) 44·6 (7·3−70·4) 42·3 (41·1−46·4) 21·8 (0·4−76·7)
Affected individuals (n) 100 44 43 27 32 20 20 4 8 7 305
Median Age at Last Follow-Up (range) 37·8 (3·4−72·1) 29·0 (4·2−75·4) 40·3 (11·8−71·6) 41·1 (5·9−69·2) 47·0 (22·0−72·9) 60·5 (3·1−71·9) 43·0 (18·3−73·0) 41·7 (9·0−64·2) 48·9 (32·8−78·0) 65·2 (29·1−75·3) 41·6 (3·1−78·0)
Total Number of Cancers (individuals, n)
Breast 92 (67) 2 (2) 40 (30) ·· 25 (22) ·· 23 (15) ·· 8 (6) ·· 190 (142)
Soft-tissue sarcoma 38 (27) 18 (13) 21 (17) 12 (10) 10 (8) 8 (5) 12 (10) 2 (2) 2 (1) 1 (1) 124 (94)
Brain 16 (16) 25 (25) 3 (3) 6 (5) 4 (4) 5 (5) 1 (1) ·· ·· 2 (2) 62 (61)
Osteosarcoma 13 (10) 8 (7) 7 (7) 8 (8) 1 (1) 2 (2) 2 (2) ·· ·· ·· 41 (37)
Hematologic 11 (10) 6 (6) 3 (3) 5 (4) 1 (1) 4 (4) 1 (1) 1 (1) ·· 1 (1) 33 (31)
Lung 11 (9) 1 (1) 3 (3) ·· 4 (4) 1 (1) 2 (2) ·· 2 (2) ·· 24 (22)
Colorectal 9 (9) 6 (6) 3 (3) 2 (2) 1 (1) 2 (2) 1 (1) ·· ·· ·· 24 (24)
Melanoma 7 (4) ·· 4 (3) 2 (2) 2 (2) 3 (3) 3 (3) ·· ·· 1 (1) 22 (18)
Thyroid 9 (8) 2 (2) ·· 2 (2) 1 (1) ·· 1 (1) 1 (1) ·· ·· 16 (15)
Prostate ·· 4 (4) ·· 1 (1) ·· 7 (7) ·· 1 (1) ·· 2 (2) 15 (15)
Other 7 (6) 2 (2) 1 (1) 3 (3) ·· 1 (1) ·· ·· 1 (1) ·· 15 (14)
Pancreatic/Liver 3 (3) 1 (1) ·· 3 (3) ·· 2 (2) 2 (1) ·· 1 (1) 1 (1) 13 (12)
Adrenal 6 (6) 3 (3) ·· 1 (1) ·· 1 (1) ·· ·· ·· ·· 11 (11)
Upper GI 2 (2) 5 (5) ·· 1 (1) ·· 1 (1) 1 (1) 1 (1) ·· ·· 11 (11)
Gynecological 5 (4) ·· 2 (2) ·· 1 (1) ·· 2 (2) ·· ·· ·· 10 (9)
Kidney 1 (1) 2 (2) ·· 1 (1) ·· 1 (1) 2 (2) ·· ·· 1 (1) 8 (8)
Total 230 (100) 85 (44) 87 (43) 47 (27) 50 (32) 38 (20) 53 (20) 6 (4) 14 (8) 9 (7) 619 (305)
Open in a new tab

There were 83 unique P/LP germline TP53 variants in the study. One family had two P/LP germline TP53 variants, p.E298* and p.G245A, identified in different individuals. Most variants were in the “Not included” or DNE_LOF categories (31 and 30, respectively), in addition to nine notDNE_LOF, eight notDNE_notLOF, and five DNE_notLOF variants. DNE_LOF variants were the most frequent, seen in 211/480 (44·0%) individuals (Appendix p 4–5).

The order of cancer occurrence was confirmed for 582 cancers. In females (Appendix p 6), breast cancer was the most common first (56·9%; 123/216) and second primary malignancy (40·4%; 40/99). Brain cancer (26·0%; 27/104) and STS (23·7%; 9/38) were the most frequent first and second cancers in males, respectively (Appendix p 6). Thirty synchronous pairs of primary cancers (60 cancers) were detected among females (20 as first primaries, of which 14 were bilateral breast cancers) and eight synchronous pairs (16 cancers) among males (seven first primaries including three pairs of STS).

Cumulative incidence of first cancer accounting for competing risks in females confirmed breast cancer as the most frequent first cancer (cumulative incidence of 56·0% by age 60 years), followed by STS (lifetime cumulative incidence of 15·0%) (Appendix p 7). In males, brain cancer and STS each had a lifetime first-cancer cumulative incidence of ~20%, with STS showing bimodal incidence in early childhood, and after age 40 years. Cumulative incidence of male osteosarcoma steadily increased to plateau at approximately 10% by the mid-late 30s. Other male cancers increased in cumulative incidence after age 40 years with prostate, colorectal, and hematologic cancers each exceeding lifetime probability of 5% as a first malignancy (Appendix p 7).

Individuals with LFS had a nearly 24-times higher incidence of any cancer (excluding ductal carcinoma in-situ [DCIS]) compared with the general population (SIR 23·9, 95% confidence interval (CI), 21·9–26·0), with the highest comparative incidence from childhood to 30 years of age (Figure 1, Appendix p 8). Though SIRs declined with age, the overall cancer incidence was still 10·3-times higher after age 50 years (95% CI, 7·9–13·2, Fig. 1A). The ‘core’ LFS cancers had the highest incidence magnitudes of up to 1000-times for adrenal cancer (95% CI, 530·1–1,902·7, Fig. 1B), >700-times for osteosarcoma (95% CI, 498·1–966·3, Fig. 1M), >200-times for STS (95% CI, 186·6268·7, Fig. 1P), >100-times for brain cancer (95% CI, 78·5–133·0, Fig. 1C), and >36-times higher for breast cancer (95% CI, 31·2–41·5, Fig. 1D). The elevated mastectomy-agnostic incidence of female breast cancer approached the population-level after age 60 years (SIR 0·9, 95% CI 0·0–5·1, Fig. 1D).

Figure 1.

Figure 1.

Open in a new tab

Standardized Incidence Ratios (SIR) and 95% Confidence Interval (CI) for the main cancer types included in the analysis. Synchronous cancers were counted independently. Dashed line at 1 represents the expected SEER-associated SIR. Results for any cancer (plus breast in situ), leukemia, lymphoblastic leukemia, MDS/AML, lymphoma, leiomyosarcoma, liposarcoma, and rhabdomyosarcoma are provided in Appendix p 8.

Among the 457 individuals with complete information for this analysis, the median age at first cancer (50% probability of cancer) was 36·1 years (95% CI, 34·4–38·2 years); 33·7 years for females (95% CI, 31·4–35·6 years), and 45.0years for males (95% CI, 40·6–50·2 years; Fig. 2A). When breast cancers were treated as a competing risk, the probability of a first cancer diagnosis of a non-breast cancer malignancy for females decreased to 24·4% by age 33.7 years (95% CI, 19·6%−30·5%, Fig. 2B). For males, the median age at first cancer minimally increased to 45·7 years (95% CI, 40·8–50·7 years, Fig. 2B) when considering male-specific cancers as a competing risk. Overall, the risk of non-sex-specific cancers was significantly higher for males than females (p=0·0027, Fig. 2B).

Figure 2.

Figure 2.

Open in a new tab

Probability of diagnosis of a first cancer by age, stratified by sex (A, B) and functional variant group (C, D). A, C: Any first cancer. B, D: Any non-sex-specific or non-breast first cancer, considering sex-specific (breast, gynecological, prostate) or breast cancers only as a competing risk. Synchronous cancers of the same category were counted as single events; synchronous cancers of different categories were counted independently. DNE_LOF class was considered as a baseline. P-values calculated using family-clustered Cox-proportional-hazard (A, C) and proportional causespecific hazard (B, D) models. Abbreviations: LOF, loss-of-function; DNE, dominant-negative effect.

Analyses stratified by TP53 variant functional classes showed that the notDNE_LOF, DNE_LOF, and “Not included” groups had overall earlier median ages at first cancer, with a difference of up to 30·9 years between the notDNE_LOF and DNE_notLOF groups (Fig. 2C). Similar findings were observed when treating sex-specific cancers as a competing risk (Fig. 2D) or when stratifying by sex (Appendix p 9).

Cancer-specific analyses estimate a 25% probability of diagnosis of breast cancer as a first malignancy by age 33·7 years in females (95% CI, 31·1–35·6, Appendix p 10); five females underwent prophylactic bilateral mastectomy prior to the diagnosis of a first cancer of any type (considered as a competing event for Appendix p 10, panels A, E). Sensitivity analyses excluding 78 women for whom mastectomy information was incomplete, showed comparable cumulative incidence curves. There was no statistical difference between sexes for age at diagnosis of STS as a first cancer (p=0·65, Appendix p 10). As first cancers, brain cancer and osteosarcoma developed earlier in males (p=0·0018 and p=0·0056, respectively, Appendix p 10). For breast cancer and STS, the differences by functional variant groups were similar to that of all cancers, with notDNE_LOF, DNE_LOF, and “Not included” classes showing earlier age at first-cancer diagnosis (Appendix p 10). However, the DNE_LOF group was associated with earlier age at brain cancer diagnosis (Appendix p 10) and the “Not included” group with earlier osteosarcoma (Appendix p 10). Parameter estimates and p-values for all between-functional group comparisons are shown in Appendix p 11–13.

Of 276 individuals with complete data and a first cancer diagnosis, 129/276 (46·7%) developed second primary cancers. Our data showed that by 20 years after first cancer diagnosis, there was a 22·2% (95% CI, 17·3–28·7) risk of death prior to second cancer (Appendix p 14); our results suggest that males had a higher risk than females of death after first cancer but prior to second cancer (Appendix p 14). DNE_notLOF had the shortest time from first cancer to death prior to second cancer (Appendix p 14), likely due primarily to this group’s older age at first cancer diagnosis (Fig. 2C); shorter time from first cancer to death prior to second cancer was also noted among individuals with a later first cancer diagnosis (Appendix p 14). Accounting for death prior to second cancer as a competing risk, females had a shorter interval between first and second cancer diagnoses than males, although this difference was not statistically significant when comparing cause-specific hazards (50% risk of second cancer by 12·0 years after the first (95% CI, 9·14–15·0) in females and 28·9 years (95% CI, 10·0–37·7) in males, p=0·19, Appendix p 15). Based on these analyses, females have an estimated probability of 66% of developing a second cancer, and males have a probability of 45·4% of second cancer, within 20 years after the first cancer (Appendix p 15). There was no difference between males and females in the time to second cancer when sex-specific cancers were considered as an additional competing risk (p=0·66, Appendix p 15). Analysis by functional variant group showed that, while the differences in time to second cancer were not considerably pronounced, DNE_LOF variants were associated with the shortest intervals between first and second cancers (Appendix p 15). Comparisons between all functional groups are shown in Appendix p 11–13.

When accounting for death prior to second cancer as a competing risk, the probability of developing a second cancer by 7.5 years after the first was comparable when stratifying by age-group at first cancer diagnosis (0–17: 32·0%, 18–29: 29·4%, 30–44: 32·0%, 45+: 29·6%). After this period, individuals whose first cancers were diagnosed before 17 years of age developed a second cancer after a longer time interval (median interval 21·0 years; 95 CI%, 8·6–37·7 years) than those with later first cancers (Fig. 3A). Comparisons between all age-range groups are shown in Appendix p 11–13. Similar analyses with delayed entry based on age at first cancer showed an estimated 25% probability of developing a second cancer by age 10·6 years (95% CI, 5·9–17·9 years) for those with a first cancer between 0–17 years of age; 30·0 years (95% CI, 24·8–36·2 years) for first cancer between 1829 years of age, 36·1 (95% CI, 33·0–39·6 years) for first cancer between 30–44 years of age; and 54·6 years (95% CI, 50·5–66·3 years) for first cancer after 45 years of age (Fig. 3B). Corresponding graphs for age at death prior to second cancer, stratified by age at first cancer, are shown in Appendix p 14.

Figure 3.

Figure 3.

Open in a new tab

A, B: Probability of diagnosis of a second cancer, by time after first cancer diagnosis (A) and age at second cancer (B), accounting for death prior to second cancer as a competing risk. Stratification by age at first cancer diagnosis. Individuals with a diagnosis at age 0–17 years were considered as a baseline. P-values calculated using family-clustered proportional cause-specific hazard models. C: Timeline plot showing time intervals between primary cancers and last follow-up or death, restricted to the 128 individuals with at least two primary cancers and known age at diagnosis for all cancers. Synchronous cancers were counted as single events. Blue lines represent time intervals between primary cancers; orange lines represent time interval between the last diagnosed primary cancer and either last follow-up or death.

The timeline plot of 128 individuals who developed at least two primary cancers with known age at diagnosis for all of their cancers shows that some carriers develop multiple cancers within a short time frame regardless of order of cancer occurrence (Fig. 3C).

DISCUSSION

Detailed understanding of cancer incidence, risks, and patterns is required to develop personalized risk assessment and tailored screening for individuals with P/LP germline TP53 variants. This study not only validates prior important studies of cancer occurrence in LFS, but further refines them, expands on the understanding of second and subsequent malignancies, genotype-phenotype associations, and puts LFS-associated cancer incidence in context with that of the general population. The exceedingly high overall and age-range-specific incidence of cancer in individuals with LFS compared with that of the general population (SEER) quantifies the significant cancer burden throughout their lifespan, even leading into the fifth and sixth decades of life. In contrast to other LFS studies that focused comparisons with the general population on a single cancer type26, or to median age at cancer onset in published literature (IARC database)27, we calculated age- and cancer-specific SIRs for 24 different cancer types and subtypes from a large, clinically curated LFS cohort. Our data showed the temporality of age distribution similar to the IARC comparison by Amadou et al.27 with the age-at-onset of many cancers being relatively similar to those in their sporadic counterparts.28,29 One notable exception is female breast cancer, which occurs significantly earlier in women with LFS than in the general population.30 The incidence of LFS-associated female breast cancer reached population levels after age 60 years and could represent a true decline of risk or be due to the limited numbers of older at-risk women with intact breast tissue. These data suggest that risk-reducing mastectomy may not substantially change breast cancer risk in females with LFS older than 60 years of age who have not had breast cancer and have intact breast tissue, though mastectomy counseling at any age should take multiple factors into consideration, including near-term and remaining lifetime risk, family history, screening adherence and accessibility, and patient concerns.

It is notable that treating first-primary breast cancers as a competing event significantly increased the age at first-cancer diagnosis in females, with risk dropping by 25·2% by age 33·7 years. There were also no appreciable differences in the time interval to second cancer or age at second cancer between males and females when sex-specific cancers were treated as an additional competing risk. These findings highlight the importance of early provider-patient discussion about cancer risk-reduction strategies, and the potentially significant impact that risk-reducing mastectomy may have in delaying cancer onset in females with LFS.

There are limited published quantitative data on second and subsequent malignancies in LFS. Bougeard et al.11 reported on the number of patients with multiple primaries and also within each time interval-range to second cancer diagnosis. Our family-clustered time-to-event analysis, which estimated second-cancer free survival as well as time interval to second cancer based on the age of the first-cancer diagnosis, is a different approach and further refines the previously published estimates. We report on risk of death being just over 20% before a second cancer is diagnosed. Regardless of age at first cancer, we observed that second cancers develop more frequently in the first 10 years after the first cancer. After 7·5 years post first cancer diagnosis, individuals with later age at first cancer diagnosis had a shorter interval to a second cancer compared to those with childhood diagnosis. In a novel approach, the timeline plot (Fig. 3C) also allowed us to visually demonstrate the time variability between subsequent cancers, beyond second cancers, showing that multiple cancers may arise within short time intervals regardless of the order of cancer occurrence. Based on this observation, we hypothesize that a “trigger-effect” may impact certain individuals with two or more cancers, leading to the development of multiple malignancies at relatively short time intervals. This “trigger effect” may be a combination of biological and/or therapy-related phenomena. Similar to Bougeard et al.11 there were some cancers in our cohort arising within radiation fields, but also many that were unrelated to radiation by site or occurred in the absence of radiotherapy. The relationship between chemotherapy and subsequent cancers is unknown in LFS and warrants further study.

Genotype-phenotype associations were analyzed using a functional assay that allowed assessments of both DNE and LOF properties simultaneously14, in contrast to prior publications comparing DNE variants to LOF or truncating variants.8,11 Our data show that specific functional groups may be utilized to characterize cancer incidence estimates among carriers of P/LP germline TP53 variants. Overall, DNE_LOF variants were associated with earlier ages at first- and second-cancer diagnoses than notDNE_notLOF variants. Our results also suggest that the variants from the “Not Included” category had similar impacts as of those from the DNE_LOF functional group. Individual analyses of functional groups (Appendix p 16) suggest that LOF status has more pronounced impact than DNE in all time-to-event analyses, and, in fact, when controlling for LOF status, it was apparent that the seemingly more severe effects of DNE in any DNE vs. notDNE analyses were driven by the excess of DNE_LOF variants in our data. Similarly, when controlling for LOF status, the presence of DNE actually appeared to decrease severity. This is opposite to what would be expected biologically and should be interpreted cautiously given the specific variants and the relatively lower number of participants in the notDNE_LOF and DNE_notLOF functional variant groups. The later median age at first cancer in the notDNE_notLOF group may be skewed due to overrepresentation of two of the eight included variants, p.R337H and p.A347D, that each account for ~30% of the 100 individuals in this group. The few variants within the DNE_notLOF group, generally associated with the highest median ages at cancer diagnoses, have counterintuitive functional properties and these results should be carefully interpreted. Due to the high incidence of breast cancers in LFS, it is still uncertain whether these rankings are consistent across different first-primary cancer types or if it is mainly driven by the breast-cancer-specific incidence estimates in females. We noted a relative enrichment for prostate cancer in carriers of notDNE_notLOF variants that was likely driven by a single family in which there were four men with prostate cancers (negative for other known prostate cancer predisposition genes) and additional cases of prostate cancer in non-carriers of germline TP53 variants, suggesting the potential for additional uncharacterized predisposition factors in this family. Some differences in our genotype-phenotype findings may reflect the fact that other studies used different assays to define the effect of variants on TP53 function and had different sample sizes and analytic approaches. Future large consortial studies will be helpful to understand variant-specific effects, and to characterize variant effects from the “Not Included” group (i.e., splice-site, frameshift and insertion/deletion variants).

The current study expands and refines existing knowledge and provides a foundation on which to build future cancer risk modeling for personalized management. Variables to consider integrating in future studies of cancer risk modeling in P/LP germline TP53 variant carriers include our data on age-range-based cancer incidence compared with the general population, cancer incidence patterns among carriers, cancer diagnosis chronology, risk of first and subsequent cancers, and variant functionality. For example, based on SIRs and cumulative incidence curves, a 25-year-old female with a P/LP germline TP53 variant has very high cancer risks compared with the general population and a specifically high breast cancer risk, in addition to other cancers. Regardless of her specific P/LP germline TP53 variant, a discussion about risk-reducing mastectomy would be warranted based on current recommendations31. If she has a DNE_LOF variant, history of a cancer 5 years ago, and intact breast tissue, she will likely develop a breast cancer before age 30 years and/or develop a second cancer in the next 5 years. In contrast, if a 25-year-old female had no prior cancer history, had a notDNE_notLOF variant and had already undergone risk-reducing mastectomy, her overall cancer-free survival would be similar to that of males, and her functional variant group also seems less prone to developing sarcomas. These two women could potentially be approached differently for their screening management in the future. These hypothetical examples illustrate some of the complexities of personalized cancer risk management that data in this study can begin to address.

This study has many strengths, including being one of the largest clinically curated LFS cohorts in the world, with extensive data on second and subsequent primary malignancies. To our knowledge, our study is the first to implement family-clustered cox-regression analysis in LFS to account for intrafamilial shared risk factors in our time-to-event analyses, and the first to comprehensively account for competing risks throughout the analyses. The ability to perform delayed entry second cancer analyses is novel and provides data for those truly at risk of second cancer. Additionally, we not only account for mortality risk in our analysis of second cancer, but also provide estimates of that risk, stratified by the same factors. We acknowledge the limitations of this study, including the referral-based enrollment that may contribute to ascertainment bias and high proportion of individuals of European ancestry with lack of representation of underrepresented minorities. While 62% of cancer diagnoses were validated by medical record review including pathology reports, this measure is not always reported but is similar to or higher than other large studies of LFS cohorts.3,11,22,32,33 Survival bias allowing certain individuals to be tested may inflate some of the adult cancer estimates and may explain why there are relatively fewer pediatric diagnoses in our cohort. Our study was not statistically powered to properly answer specific questions on second and subsequent cancers and was not able to account for prior cancer treatment. We acknowledge that the presented LFS cohort, similar to others, has female overrepresentation due to the high incidence of breast cancer. We dealt with this by stratifying our analyses by sex whenever possible.

Our study provides new data essential for future development of personalized cancer risk modelling and tailored screening approaches for individuals with LFS.

Supplementary Material

1

Research in context.

Evidence before this study:

Li-Fraumeni syndrome (LFS) is a variably penetrant autosomal dominant cancer predisposition disorder, primarily caused by pathogenic/likely pathogenic germline TP53 variants, resulting in extremely high risks of multiple primary malignancies over the individual’s lifespan. A PubMed search from database inception through March 2020 using the terms (“genotype-phenotype” or “cancer incidence” or “cancer pattern” or “cancer risk”) and (“Li-Fraumeni “or “germline TP53”) primarily identified literature reviews or studies with variable sample sizes and inconsistent approaches to genotype-phenotype characterization. About half of first cancer diagnosis in LFS occurs by the early third decade in females and mid-fourth decade in males. The very high risk of cancer in LFS has led to an intensive annual cancer screening regimen being developed, aimed at early detection and centered around whole-body MRI. This one-size-fits-all approach does not take into account the germline TP53 variant, personal or family cancer history.

Added value of this study:

This study quantifies the exceedingly high burden of cancer in individuals with LFS in comparison with their counterparts in the general population. It refines and expands on prior genotype-phenotype correlation studies by categorizing variants based on both loss-of-function and dominant-negative functional properties. Furthermore, it shows that risk-reducing mastectomy considerably delays first-cancer onset in females with LFS. These analyses are key pieces of data required for the future development of personalized cancer screening strategies. This study also provides previously undescribed granularity on second and subsequent malignancies in individuals with LFS and highlights the temporality of second cancer onset based on the age at first cancer diagnosis.

Implication of all the evidence:

Individuals with LFS have a nearly 90% lifetime risk of developing at least one, and often multiple, cancers. Cancer screening strategies offer the benefit of early cancer detection but are emotionally and resource intensive. By putting the cancer burden in LFS in context with that of the general population, examining genotype-phenotype correlations based on TP53 variant functional group, and evaluating the temporality of subsequent malignancies in LFS, this study, in conjunction with the existing literature, creates a foundation for the development of tailored cancer screening and personalized risk assessment in LFS.

ACKNOWLEDGEMENTS

This study was fully funded by the Intramural Research Program, Division of Cancer Epidemiology and Genetics, NIH. K.C.A, P.P.K., M.N.F., A.F.B., and S.A.S. are employees of NIH. We are grateful for the participation of individuals and families in the NCI’s LFS Study and thank their referring clinicians. Research support was ably led by Janet Bracci, RN, Kathryn Nichols, RN, and a team of research nurses and research assistants through contract HHSN261201800004C with Westat, Inc. We acknowledge Alejandro Lafuente’s work in creating the figure in Appendix p 6. We thank Cameron Davidson-Pilon, developer of the Lifelines library, for his assistance with this survival analysis package. Statistical analyses for the SEER comparisons were performed by Jeremy Miller, Information Management Services, Inc. (IMS).

Footnotes

DECLARATION OF INTERESTS

K.C. A, J.N.H, M.N.F, P.L.M, and S.A.S are unpaid members of the ClinGen TP53 Variant Curation Expert Panel. M.N.F is co-developer of CancerGene Connect and member of the National Accreditation Program for Breast Centers Board representing the National Society of Genetic Counselors. All other authors declare no competing interests.

DATA SHARING STATEMENT

Upon publication of the manuscript, participant data including germline TP53 variant data, age, sex, cancer history will be available, after appropriate de-identification. These data dictionaries will be shared after data sharing agreements are established between the NCI and requesting institution, for scientists with relevant expertise and hypotheses to test using these data, as long as the request is consistent with study consent forms.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

REFERENCES

  • 1.Guha T, Malkin D. Inherited TP53 Mutations and the Li-Fraumeni Syndrome. Cold Spring Harb Perspect Med 2017; 7(4). [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 2.Schneider K, Zelley K, Nichols KE, Garber J. Li-Fraumeni Syndrome. In: Adam MP, Ardinger HH, Pagon RA, et al. , eds. GeneReviews((R)). Seattle (WA); 1993. [Google Scholar]
  • 3.Mai PL, Best AF, Peters JA, et al. Risks of first and subsequent cancers among TP53 mutation carriers in the National Cancer Institute Li-Fraumeni syndrome cohort. Cancer 2016; 122(23): 3673–81. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 4.Villani A, Shore A, Wasserman JD, et al. Biochemical and imaging surveillance in germline TP53 mutation carriers with Li-Fraumeni syndrome: 11 year follow-up of a prospective observational study. Lancet Oncol 2016; 17(9): 1295–305. [DOI] [PubMed] [Google Scholar]
  • 5.Kratz CP, Achatz MI, Brugières L, et al. Cancer Screening Recommendations for Individuals with Li-Fraumeni Syndrome. Clin Cancer Res 2017; 23(11): e38–e45. [DOI] [PubMed] [Google Scholar]
  • 6.Ross J, Bojadzieva J, Peterson S, et al. The psychosocial effects of the Li-Fraumeni Education and Early Detection (LEAD) program on individuals with Li-Fraumeni syndrome. Genet Med 2017; 19(9): 1064–70. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.Young JL, Pantaleao A, Zaspel L, et al. Couples coping with screening burden and diagnostic uncertainty in Li-Fraumeni syndrome: Connection versus independence. J Psychosoc Oncol 2019; 37(2): 178–93. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Fortuno C, Pesaran T, Mester J, et al. Genotype-phenotype correlations among TP53 carriers: Literature review and analysis of probands undergoing multi-gene panel testing and single-gene testing. Cancer Genet 2020; 248-249: 11–7. [DOI] [PubMed] [Google Scholar]
  • 9.Gargallo P, Yanez Y, Segura V, et al. Li-Fraumeni syndrome heterogeneity. Clin Transl Oncol 2020; 22(7): 978–88. [DOI] [PubMed] [Google Scholar]
  • 10.Rana HQ, Gelman R, LaDuca H, et al. Differences in TP53 Mutation Carrier Phenotypes Emerge From Panel-Based Testing. J Natl Cancer Inst 2018; 110(8): 863–70. [DOI] [PubMed] [Google Scholar]
  • 11.Bougeard G, Renaux-Petel M, Flaman JM, et al. Revisiting Li-Fraumeni Syndrome From TP53 Mutation Carriers. J Clin Oncol 2015; 33(21): 2345–52. [DOI] [PubMed] [Google Scholar]
  • 12.Frebourg T, Bajalica Lagercrantz S, Oliveira C, Magenheim R, Evans DG, European Reference Network G. Guidelines for the Li-Fraumeni and heritable TP53-related cancer syndromes. Eur J Hum Genet 2020; 28(10): 1379–86. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Rana HQ, Clifford J, Hoang L, et al. Genotype-phenotype associations among panel-based TP53+ subjects. Genet Med 2019; 21(11): 2478–84. [DOI] [PubMed] [Google Scholar]
  • 14.Giacomelli AO, Yang X, Lintner RE, et al. Mutational processes shape the landscape of TP53 mutations in human cancer. Nat Genet 2018; 50(10): 1381–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Landrum MJ, Lee JM, Benson M, et al. ClinVar: improving access to variant interpretations and supporting evidence. Nucleic Acids Res 2018; 46(D1): D1062–D7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Fortuno C, Lee K, Olivier M, et al. Specifications of the ACMG/AMP variant interpretation guidelines for germline TP53 variants. Hum Mutat 2021; 42(3): 223–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 17.Bouaoun L, Sonkin D, Ardin M, et al. TP53 Variations in Human Cancers: New Lessons from the IARC TP53 Database and Genomics Data. Hum Mutat 2016; 37(9): 865–76. [DOI] [PubMed] [Google Scholar]
  • 18.Kaplan EL, Meier P. Nonparametric-Estimation from Incomplete Observations. Journal of the American Statistical Association 1958; 53(282): 457–81. [Google Scholar]
  • 19.Gaynor JJ, Feuer EJ, Tan CC, et al. On the Use of Cause-Specific Failure and Conditional Failure Probabilities: Examples from Clinical Oncology Data. Journal of the American Statistical Association 1993; 88: 400–9. [Google Scholar]
  • 20.Aalen OO, Johansen S. An Empirical Transition Matrix for Non-Homogeneous Markov Chains Based on Censored Observations . Scandinavian Journal of Statistics 1978; 5: 141–50. [Google Scholar]
  • 21.Li FP, Fraumeni JF Jr., Mulvihill JJ, et al. A cancer family syndrome in twenty-four kindreds. Cancer Res 1988; 48(18): 5358–62. [PubMed] [Google Scholar]
  • 22.Chompret A, Abel A, Stoppa-Lyonnet D, et al. Sensitivity and predictive value of criteria for p53 germline mutation screening. J Med Genet 2001; 38(1): 43–7. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Tinat J, Bougeard G, Baert-Desurmont S, et al. 2009 version of the Chompret criteria for Li Fraumeni syndrome. J Clin Oncol 2009; 27(26): e108–9; author reply e10. [DOI] [PubMed] [Google Scholar]
  • 24.Eeles RA. Germline mutations in the TP53 gene. Cancer Surv 1995; 25: 101–24. [PubMed] [Google Scholar]
  • 25.Birch JM, Hartley AL, Tricker KJ, et al. Prevalence and diversity of constitutional mutations in the p53 gene among 21 Li-Fraumeni families. Cancer Res 1994; 54(5): 1298–304. [PubMed] [Google Scholar]
  • 26.Ognjanovic S, Olivier M, Bergemann TL, Hainaut P. Sarcomas in TP53 germline mutation carriers: a review of the IARC TP53 database. Cancer 2012; 118(5): 1387–96. [DOI] [PubMed] [Google Scholar]
  • 27.Amadou A, Waddington Achatz MI, Hainaut P. Revisiting tumor patterns and penetrance in germline TP53 mutation carriers: temporal phases of Li-Fraumeni syndrome. Curr Opin Oncol 2018; 30(1): 23–9. [DOI] [PubMed] [Google Scholar]
  • 28.Mirabello L, Troisi RJ, Savage SA. Osteosarcoma incidence and survival rates from 1973 to 2004: data from the Surveillance, Epidemiology, and End Results Program. Cancer 2009; 115(7): 1531–43. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 29.Ferrari A, Sultan I, Huang TT, et al. Soft tissue sarcoma across the age spectrum: a population-based study from the Surveillance Epidemiology and End Results database. Pediatr Blood Cancer 2011; 57(6): 943–9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 30.Noone AM, Cronin KA, Altekruse SF, et al. Cancer Incidence and Survival Trends by Subtype Using Data from the Surveillance Epidemiology and End Results Program, 1992–2013. Cancer Epidemiol Biomarkers Prev 2017; 26(4): 632–41. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.National Comprehensive Cancer Network (NCCN). NCCN Clinical Practice Guidelines in Oncology: Genetic/Familial High Risk Assessment: Breast and Ovarian. Version v2.2021. http://www.nccn.org/professionals/physician_gls/pdf/genetics_screening.pdf. [DOI] [PubMed] [Google Scholar]
  • 32.Olivier M, Goldgar DE, Sodha N, et al. Li-Fraumeni and related syndromes: correlation between tumor type, family structure, and TP53 genotype. Cancer Res 2003; 63(20): 6643–50. [PubMed] [Google Scholar]
  • 33.Gonzalez KD, Noltner KA, Buzin CH, et al. Beyond Li Fraumeni Syndrome: clinical characteristics of families with p53 germline mutations. J Clin Oncol 2009; 27(8): 1250–6. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1
NIHMS1758686-supplement-1.pdf (5.4MB, pdf)

RESOURCES