Observational study of population genomic screening for variants associated with endocrine tumor syndromes in a large, healthcare-based cohort

Several autosomal dominant, hereditary syndromes are associated with an increased risk of developing tumors of the endocrine and neuroendocrine system, including multiple endocrine neoplasia type 1 (MEN 1; OMIM 131100), multiple endocrine neoplasia type 2 (MEN 2; OMIM 155240, 171400, 155240), hereditary paraganglioma and pheochromocytoma (PGL/PCC; OMIM 115310, 605373, 601650, 168000), and von Hippel-Lindau (VHL; OMIM 193300) syndromes [1]. These conditions are considered rare disorders, with a collective disease prevalence of approximately 1 in 8500 [2].

MEN 1 is caused by heterozygous, loss-of-function variants in the MEN1 tumor suppressor gene (OMIM 613733) and is associated with endocrine and non-endocrine features. Individuals with MEN 1 have an increased risk of developing parathyroid gland hyperplasia, pituitary neoplasms, pancreatic neuroendocrine tumors, and, less commonly, adrenocortical tumors, lipomas, angiofibromas, collagenomas, meningiomas, ependymomas, schwannomas, leiomyomas, and leiomyosarcomas [3, 4].

MEN 2 is caused by missense, heterozygous, gain-of-function variants in the RET protooncogene (OMIM 164761) and is associated with an increased risk for medullary thyroid cancer. MEN 2 includes two clinically defined subtypes—MEN 2A and MEN 2B. MEN 2A can be further subdivided into classical MEN 2A, MEN 2A with cutaneous lichen amyloidosis (MEN 2A and CLA), MEN 2A and Hirschsprung disease (MEN 2A and HD), and familial medullary thyroid cancer (FMTC). Depending on the MEN 2A subtype, individuals might be at an increased risk for pheochromocytoma, primary hyperparathyroidism, cutaneous lichen amyloidosis, and/or Hirschsprung disease [5]. MEN 2B is associated with an earlier onset of MTC and pheochromocytomas compared to MEN 2A and may present with additional physical features including mucosal neuromas, intestinal ganglioneuromatosis, and characteristic physical features. Associations between MEN 2 subtype, aggressiveness of MTC, and the protein codon impacted by the DNA variant have been established [5].

PGL/PCC syndromes are caused by heterozygous, loss-of-function variants in genes coding for proteins involved in the mitochondrial respiratory chain complex II, including, but not limited to, SDHAF2 (OMIM 613019), SDHB (OMIM 185470), SDHC (OMIM 602413), and SDHD (OMIM 602690) (hereafter referred to as “SDHx” genes). Individuals with PGL/PCC syndromes are at an increased risk of developing paragangliomas, pheochromocytomas, renal cell carcinoma, and gastrointestinal stromal tumors (GISTs) [6‐8]. Although heterozygous, pathogenic/likely pathogenic (P/LP) variants in these SDHx genes increase the risk for paragangliomas, the location of these tumors, biochemical phenotype, and risk for malignancy have been correlated with the gene impacted [9]. Additionally, SDHAF2 and SDHD exhibit a parent-of-origin effect in which tumor development primarily occurs in individuals that inherit the variant on the paternal allele [10]. Homozygous and compound heterozygous P/LP variants in SDHx genes are associated with biochemical disorders such as the mitochondrial complex II deficiencies [11]. Since these are severe, early-onset disorders that usually are diagnosed clinically, they are not considered in this study.

Finally, VHL syndrome is caused by heterozygous, loss-of-function variants in the VHL tumor suppressor gene (OMIM 602690). Individuals with such variants are at an increased risk for hemangioblastomas of the brain, spinal cord, and retina; renal cell carcinoma; pheochromocytomas; paragangliomas; pancreatic neuroendocrine tumors; and papillary cystadenomas of the epididymis and broad ligament [12].

It is recommended that individuals with P/LP variants in MEN1, RET, an SDHx gene, or VHL undergo periodic biochemical and imaging surveillance for associated phenotypes [4, 9, 13‐20]. Should endocrine tumors or other features be identified through this surveillance, additional symptomatic management may be indicated (e.g., hyperparathyroidism in a patient with MEN 1 may warrant parathyroidectomy). Additionally, prophylactic thyroidectomy is considered for individuals with P/LP RET variants [5, 18].

Germline genetic testing for these hereditary syndromes has historically been pursued when an individual’s personal and/or self-reported family history is suggestive of the condition (e.g., RET sequencing when a person is diagnosed with medullary thyroid cancer or SDHx testing after a paraganglioma of the head or neck is diagnosed), as the presence of a P/LP variant informs additional tumor risks, guides risk management, and enables testing of at-risk relatives [5, 21‐24]. However, recent studies examining clinical genetic testing practices for other hereditary cancer syndromes and familial hypercholesterolemia have shown that indication-based genetic testing fails to identify a substantial proportion of at-risk individuals [25‐30]. Furthermore, more comprehensive identification of variants in unselected individuals via genomic screening leads to risk management, can assist in syndrome-related diagnoses [26, 31], and may be cost-effective in some populations [32].

It is unclear whether these benefits of improved ascertainment of at-risk individuals or positive impacts on risk management are present when screening broader cohorts for ETS risk. Here, we describe an observational study examining a healthcare system that screens biobank participants for P/LP variants in genes associated with MEN 1, MEN 2, PGL/PCC, or VHL. In particular, we describe the personal and self-reported family history of syndrome-related features in patient-participants ascertained through the Geisinger MyCode® Community Health Initiative (MyCode), their risk management behaviors after receiving a result, and the rates of relevant post-disclosure diagnoses of neoplasms and other syndrome-related features.

This study aims to provide critical insights needed to inform future genomic screening of unselected populations for ETS and support studies of longer-term clinical outcomes of genomic screening for ETS risk.

This study reports on the experience screening a healthcare system population for genetic variants associated with ETS risk. We found that ETS are much more common than previously reported, genomic screening for ETS can ascertain at-risk individuals who should have already been ascertained but were not, as well as those who would not have otherwise come to clinical attention, and genomic screening facilitated diagnoses of endocrine neoplasms in several participants. Yet, further research is necessary to address key outstanding questions on the clinical utility of genomic screening for ETS.

By screening an unselected healthcare system cohort for P/LP variants in ETS genes, we found a considerably higher rate of ETS — 1:622 to 1:751 — than is commonly reported. Although this variant prevalence is only an estimate due to conservative pipeline filtering [36], this estimate is over eleven times greater than the previously reported cumulative disease prevalence for MEN 1, MEN 2, PGL/PCC, and VHL of approximately 1:8500 [2]. These results are consistent with a recent abstract summarizing a study of RET variants in a healthcare-based biobank that found such variants are more common than anticipated based on disease prevalence [45]. The high variant prevalence reported here is, in part, due to the genotype-first approach undertaken in this study and the expected reduced penetrance of some variants (e.g., penetrance of maternally inherited SDHAF2 and SDHD variants). However, these data suggest that variants associated with ETS risk are more common than previously reported. Such data inform future population screening efforts by providing an estimated number of returnable variants in a healthcare-based cohort. These data are needed to contribute to future studies (e.g., All of Us [45] and Genomics England [46]) performing genomic screening as well as cost-effectiveness analyses that will be necessary to further assess the feasibility of genomic screening for these conditions [47]. Furthermore, although some variants are expected to have reduced penetrance, such estimates are based on clinical cohorts that tend to overestimate prevalence based on the ascertainment of individuals with a personal or family history of disease. Identifying variant prevalence and exploring personal and family history data in broader cohorts such as this will be needed to expand our understanding of penetrance and specific genotype/phenotype correlations.

This study also provides data regarding how broader screening efforts could increase the identification of at-risk individuals and suggests how such screening could augment current clinical care. Our data provide evidence of under-ascertainment of ETS in current clinical practice in two ways. First, despite having a personal and/or self-reported family history consistent with established cancer genetics referral guidelines [24], none of the seven MyCode-identified participants who met referral criteria had been evaluated clinically pre-disclosure, suggesting a need to more effectively identify patients who meet current referral criteria. Provider education and clinical decision support in the EHR could improve the clinical ascertainment of these at-risk participants. Population screening should not replace clinical evaluations of those meeting referral guidelines [47], but these data suggest such screening could capture those who have been missed by current practices. Second, referral guidelines are not sufficiently sensitive to identify all individuals with ETS risk. The majority of participants with a variant associated with ETS risk (89%, n=58/65) did not meet genetics referral criteria pre-disclosure. These data suggest that, even if referral criteria were appropriately applied, individuals with P/LP ETS variants would be missed in current clinical practice. Moreover, the distribution across genes differs in the participants who were previously clinically identified compared to those identified via MyCode (Additional file 1: Table S1), suggesting that under-ascertainment via current clinical practice might differ by gene (e.g., increased under-ascertainment of SDHx compared to MEN1). Similar findings of under-ascertainment in current practice have been illustrated in other hereditary cancer syndromes and actionable conditions [28, 29]. This suggests that broader testing strategies, such as population genomic screening, could provide an approach to better identify at-risk individuals.

Improved identification of at-risk individuals will not improve health outcomes if no change in health behavior follows disclosure. This study provides further evidence that genomic screening can prompt changes to clinical care. The majority (55%) of individuals who were not previously aware of their result completed at least one associated risk management behavior after disclosure (biochemical/imaging surveillance or surgical intervention). The number of individuals who sought additional care post-disclosure is similar to those with other actionable genetic risks including variants associated with hereditary breast and ovarian cancer syndrome, Lynch syndrome, and familial hypercholesterolemia [26]. Participants who saw a genetics provider and those who met with a specialist were more likely to complete recommended risk management compared to those who did not meet with these providers. It is not clear whether genetic counselors and/or specialists facilitated the performance of recommended risk management or whether individuals who chose to meet with a genetic counselor and/or specialist are more engaged in health care generally, or more likely to complete recommended follow-up due to other factors. Follow-up with a PCP was not associated with the performance of risk-related management behaviors post-disclosure. With the EHR data available, it is unclear whether this is because PCPs did not recommend risk-related management or if patients did not act on such recommendations. Future studies should continue to explore facilitators and barriers to patient follow-up to determine how best to facilitate the complementary roles PCPs, genetic counselors, endocrinologists, and other specialists can play in communicating and coordinating risk management for individuals with a P/LP ETS variant.

The diagnosis of MTC in eight of nine participants who underwent post-disclosure thyroidectomy illustrates the promise of population screening for genetic disease risk to inform care and enable diagnoses. It may be that there is particular utility in screening populations for RET variants and that additional interventions to encourage thyroidectomies among participants with a RET variant are indicated. Or, it may be that these cases represent lead time bias, over-diagnosis, and over-treatment. Resolving this question will require larger, longitudinal studies designed to elucidate the impact of population genetic screening on morbidity and mortality. Such studies can also investigate whether to alter risk management recommendations for individuals ascertained from population-based cohorts [47]. For example, if further studies support that thyroidectomy in patients with RET variants identified via genomic screening leads to over-diagnosis, perhaps a stepped approach to risk management in adults that begins with calcitonin screening and imaging to determine whether to proceed to thyroidectomy in asymptomatic individuals with RET variants could be considered. Biochemical screening may offer a less burdensome management behavior that can enable earlier diagnoses in those identified to have genomic ETS risk via broader screening. Further studies will be needed to resolve this question, explore additional factors that might be important when determining individualized risk management (e.g., age, specific variant detected), and generate evidence to determine the appropriateness of altering care in those receiving RET results in broader populations. Until such studies are completed, care based on current guidelines represents the conservative approach.

Two limitations of this study are the small sample size and short follow-up time, which limit our ability to determine factors associated with post-disclosure risk management through bivariate analyses, our capacity to use more sophisticated modeling, and our ability to evaluate long-term outcomes of disclosure and follow-up care. The small number of participants receiving MEN1 and VHL variants limits our capacity to draw conclusions about the disclosure of variants in those genes.

Furthermore, personal history, self-reported family history, and post-disclosure risk management relied on data available in the Geisinger EHR which has its limitations as a data source. Personal history of ETS based on the EHR might not include all such diagnoses if a patient receives care outside of the system. Self-reported family history might be incomplete for some patients since capture and recording of these data vary (e.g., 21% of participants had a pedigree while the remaining only had family history data in the EHR). As such, relying on EHR data might lead to underestimates of relevant personal and family history [29]. Additionally, post-disclosure risk management relied on documentation of management in the EHR which, again, fails to capture care that occurs outside of the system. This could lead to an underestimate of the impact of these results on care and diagnoses. Finally, while we attempted to determine if the risk management behaviors were clearly attributed to the disclosure of the genetic variant identified, the indication for surveillance was not always clearly documented in the EHR. Future studies that include EHR data as well as other data sources, including patient-reported care information could further determine the clinical impact of these genetic results.

Other genes (e.g., SDHA, MAX, TMEM127, PRKAR1A) associated with ETS risk were not included in this analysis since they were not being disclosed to MyCode participants at the time of the study. Additionally, genes that have clinical presentation and management that are not primarily endocrine tumor syndrome focused (e.g., PTEN with risks for thyroid cancer as well as risks for neurodevelopmental disorders, colorectal, breast, and endometrial cancers) were also excluded.

It is unclear if these findings will be generalizable to more racially and ethnically diverse cohorts, to other healthcare systems, or to the general population. Additional studies, including those with longer follow-up, those that compare rates of phenotypes to variant-negative and clinically ascertained individuals, and those that combine data from multiple, diverse cohorts, are needed to explore P/LP variant prevalence and penetrance in unselected cohorts, clinical utility and outcomes of genomic screening, cost-effectiveness of result disclosure, and care, and determine factors that are associated with post-disclosure risk management behaviors. Future studies will also be needed to examine the de novo rate and whether parent-of-origin effects are seen in genomic screening cohorts.

Despite the need for additional studies to fully understand the clinical utility and outcomes of genomic screening for ETS, this work demonstrates that screening healthcare populations can enable the detection of individuals at genetic risk for ETS, lead to uptake of risk management, and facilitate relevant clinical diagnoses.