Uterine serous carcinoma and germline genetic testing: patterns of referral, completion and pathogenic variant detection Authors: Lauren Tostrud, Simge Bagci Turkmen, Jiaqi Zhang, Rachel Hodan, Kerry Kingham, Oliver Dorigo, James M Ford, Allison W Kurian, Talayeh S Ghezelayagh Published: 2026 ABSTRACT AbstractBackgroundCurrent guidelines recommend consideration of germline genetic testing for patients with uterine serous carcinoma (USC) but real-world data are limited on completion rates of testing and pathogenic variant (PV) identification. This study aimed to evaluate the rate of genetic testing referral and completion in a cohort of patients with USC, determine the prevalence of clinically meaningful PVs found on testing and explore factors associated with genetics referral and testing completion.MethodsWe retrospectively examined the medical records of all individuals diagnosed with USC between 2019 and 2024 seen at a single academic cancer centre. Outcomes of interest included referral for germline genetic testing, completion of testing and testing results.ResultsOf 131 individuals included, 5 (3.8%) had prior genetic testing and only 45 (34.4%) were recommended to undergo genetic testing or referred to cancer genetics. Younger individuals and those with a personal history of cancer other than USC or family history of breast or ovarian cancer were more likely to be referred. Nine (26.5%) of 34 individuals who completed germline testing had a PV identified in a cancer-related gene, including BRCA1, BRCA2, BRIP1, CHEK2, MSH6, PMS2 and ATM. Only a personal history of cancer other than USC was independently associated with the discovery of a PV on germline genetic testing. In those without a prior personal history of cancer, the PV prevalence was 5.6%.ConclusionsGiven the high prevalence of PVs in this population, germline genetic testing for all patients diagnosed with USC can provide clinically meaningful benefit but is currently underused in practice.Keywords: Genetic Testing, Medical Oncology, Genetic Predisposition to Disease, Mutation Rate WHAT IS ALREADY KNOWN ON THIS TOPIC Both the incidence and mortality of uterine cancer are rising in the USA. There is an increasing understanding of associations between germline pathogenic variants and uterine serous carcinoma (USC) but data are limited. Both the incidence and mortality of uterine cancer are rising in the USA. There is an increasing understanding of associations between germline pathogenic variants and uterine serous carcinoma (USC) but data are limited. WHAT THIS STUDY ADDS The majority of patients with USC are not undergoing germline genetic testing, but 26.5% of patients who did complete genetic testing in this cohort carried a cancer-related pathogenic variant. This study found that younger patients and those with a personal and family history of cancer were significantly more likely to be referred to genetics, but only a personal history of cancer other than USC was found to be independently predictive for the presence of a germline pathogenic variant. The majority of patients with USC are not undergoing germline genetic testing, but 26.5% of patients who did complete genetic testing in this cohort carried a cancer-related pathogenic variant. This study found that younger patients and those with a personal and family history of cancer were significantly more likely to be referred to genetics, but only a personal history of cancer other than USC was found to be independently predictive for the presence of a germline pathogenic variant. HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Given the high prevalence of pathogenic variants found among this patient population, this study argues for strengthening guideline recommendations for germline genetic testing in patients with USC. Given the high prevalence of pathogenic variants found among this patient population, this study argues for strengthening guideline recommendations for germline genetic testing in patients with USC. Introduction In 2022, the most recent global estimates suggest that endometrial cancer accounted for 420 242 new cases worldwide and 97 704 deaths.1 Global burden analyses show that incident cases have risen substantially over recent decades (from ~1 87 191 in 1990 to ~4 35 041 in 2019).2 Uterine serous carcinoma (USC) is an aggressive subtype of endometrial cancer—representing only 10% of cases but responsible for 40% of endometrial cancer-related deaths.3 Accumulating data suggest an increased risk of USC in individuals with hereditary cancer syndromes, primarily in those with germline pathogenic variants (PVs) in BRCA1 but also potentially BRCA2.4–8 In 2023, the National Comprehensive Cancer Network (NCCN) Genetic/Familial Breast, Ovarian and Pancreatic guidelines updated their recommendations to consider genetic testing with appropriate pretest education and access to post-test management in patients with a personal history of USC.9 This recommendation was made given evolving evidence including findings of a potential 10-fold or higher increased risk of serous-like endometrial cancers, with 1%–3% absolute lifetime risk, in BRCA1/2 PV carriers.4–8 Furthermore, germline genetic testing in patients with unselected endometrial cancer cohorts including USC cases has revealed high rates of PVs in a broad range of cancer-related genes.10 11 However, as the population prevalence of germline variants other than BRCA1/2 is low, epidemiological cohort studies examining uterine cancer risk in these individual groups are lacking. This project was designed to determine contemporary rates of genetic testing referral and completion in patients with USC at a tertiary care centre, and to report and describe the prevalence of clinically meaningful PVs found on testing. Methods Study population We conducted a retrospective cohort analysis of patients treated for USC at a tertiary academic cancer centre. Individuals diagnosed with endometrial or uterine cancer between 1 January 2019 and 31 December 2024 and seen by physicians in the Gynecologic Oncology division were identified through search of STARR, ‘STAnford medicine Research data Repository’, a clinical data warehouse containing live Epic electronic medical record data from Stanford Health Care, the Stanford Children’s Hospital, the University Healthcare Alliance and Packard Children’s Health Alliance clinics.12 All individuals had at least 1 year of follow-up time after diagnosis of cancer available at the time of chart review. Manual review of electronic medical records was then conducted to confirm inclusion with review of pathology reports confirming uterine serous histology. Individuals with endometrioid endometrial adenocarcinomas with serous components and other endometrial and uterine sarcoma histologies were excluded. Medical record review Medical records were manually abstracted to collect data on demographics, cancer pathology, staging (per FIGO 2009 staging system), molecular classification and personal and family history of cancer (excluding cutaneous basal cell and squamous cell carcinomas). Outcomes of interest included referral for germline genetic testing, completion of germline genetic testing and germline genetic testing results. Genetic testing may have been ordered at both our institution or at outside institutions and all available records accessible through Epic were reviewed. The medical charts were reviewed by two trained investigators who independently abstracted data. Discrepancies in data extraction were resolved by a third trained investigator. Genetic variant classification In individuals who underwent germline genetic testing, results were obtained through medical record review. All testing was completed through commercial-based multigene panel tests. Variants classified as pathogenic or likely pathogenic in the germline test report in genes involved in oncologic germline conditions found in the Online Mendelian Inheritance in Man (OMIM) database were recorded and confirmed with review through ClinVar.13 14 All reported germline variants were validated for Human Genome Variation Society (HGVS) nomenclature compliance using Mutalyzer.15 For all but one individual, the primary result report or genetic counselling report was located and independently reviewed. Only one BRCA1 PV could not be confirmed through testing report review but was included given multiple secondary records confirming a PV. APC variant c.3920T>A (p.Ile1307Lys) was not considered a pathogenic cancer-related variant in our primary analysis given conflicting interpretations and international management strategies.16 It was included in secondary analyses as NCCN still recommends different colorectal screening for individuals carrying this variant.17 Statistical analysis Statistical comparisons between the individuals who were referred or were not referred for genetic testing were completed using Student’s t-test for continuous variables and χ2 or Fisher’s exact test (if cell counts <5) for categorical variables. Similar comparisons were also made between individuals who ultimately completed or did not complete genetic testing and those who were found to carry a germline PV or had no PVs identified on testing. Crude and multivariable logistic regression was conducted to determine if a priori defined clinical variables (age at diagnosis, body mass index (BMI), stage at diagnosis, family history of breast or ovarian cancer and personal history of cancer) could predict the identification of PVs on germline genetic testing. Of note, the five patients who completed genetic testing prior to diagnosis of USC were excluded from the analysis of referral outcomes but were included in the analysis of testing results. All tests were two-sided, with the significance level at 0.05. Analysis was completed using StataNow/SE V.18.5.18 Results 131 patients with pathology-report confirmed USC diagnosed between 2019 and 2024 were seen at our institution and included in the final study cohort (table 1). Only one tumour was mismatch repair deficient on immunohistochemistry testing. The median age was 70 (min–max 44–87) with median BMI of 29 (min–max 17–50). The cohort was primarily White (n=70, 53.4%) and non-Hispanic (n=107, 81.7%). The cohort was skewed towards patients diagnosed in 2019 and 2020 (42.0% in these years), secondary to institutional referral patterns. 80 (61.1%) individuals had stage III or stage IV disease at diagnosis, reflecting the more aggressive presentation of USC. At last follow-up, 33.6% of individuals had recurrent disease. 31 (23.7%) individuals had a personal history of other cancers (including 16 with breast cancer, 7 with lung cancer) and 46 (35.1%) had a family history of breast or ovarian cancer. Table 1Patient characteristics by genetics referral statusAll (n=131)Referred for testing (n=45)*Not referred (n=81)P valueAge at diagnosis0.018 Median (min, max)70 (44, 87)69 (44, 78)72 (53, 87)BMI0.382 Median (min, max)29 (17, 50)28 (18, 50)30 (17, 49)Race0.044 American Indian/Alaskan Native1 (0.8%)01 (1.2%) Asian24 (18.3%)13 (28.9%)10 (12.4%) Black or African American7 (5.3%)4 (8.9%)3 (3.7%) Native Hawaiian or Other Pacific Islander1 (0.8%)1 (2.2%)0 White70 (53.4%)20 (44.4%)47 (58.0%) Other/unknown28 (21.4%)7 (15.6%)20 (24.7%)Hispanic/Non-Hispanic0.745 Non-Hispanic107 (81.7%)38 (84.4%)65 (80.3%) Hispanic21 (16.0%)7 (15.6%)14 (17.3%) Unknown3 (2.3%)02 (2.5%)Language0.223 English-speaking110 (84.0%)35 (77.8%)70 (86.4%) Interpreter use21 (16.0%)10 (22.2%)11 (13.6%)Insurance status0.366 Private35 (26.7%)14 (31.1%)19 (23.5%) Public94 (71.8%)30 (66.7%)61 (75.3%) Self-pay1 (0.8%)01 (1.2%) Unknown1 (0.8%)1 (2.2%)0Personal history of cancer0.030 Yes31 (23.7%)14 (31.1%)12 (14.8%) No100 (76.3%)31 (68.9%)69 (85.2%)Family history of breast or ovarian cancer0.084 Yes46 (35.1%)19 (42.2%)22 (27.2%) No85 (64.9%)26 (57.8%)59 (72.8%)Stage at diagnosis (FIGO 2009)0.503 I41 (31.3%)14 (31.1%)24 (29.6%) II10 (7.6%)3 (6.7%)7 (8.6%) III30 (22.9%)7 (15.6%)21 (25.9%) IV50 (38.2%)21 (46.7%)29 (35.8%)Year of diagnosis0.019 201928 (21.4%)10 (22.2%)16 (19.8%) 202027 (20.6%)13 (28.9%)14 (17.3%) 202123 (17.6%)4 (8.9%)16 (19.8%) 202213 (9.9%)1 (2.2%)12 (14.8%) 202319 (14.5%)5 (11.1%)14 (17.3%) 202421 (16.0%)12 (26.7%)9 (11.1%)Disease status at last follow-up0.200 Primary87 (66.4%)26 (57.8%)56 (69.1%) Recurrent44 (33.6%)19 (42.2%)25 (30.9%)*Individuals who underwent genetic testing prior to uterine serous carcinoma diagnosis were excluded.BMI, body mass index; FIGO, The International Federation of Gynecology and Obstetrics. Patient characteristics by genetics referral status Individuals who underwent genetic testing prior to uterine serous carcinoma diagnosis were excluded. BMI, body mass index; FIGO, The International Federation of Gynecology and Obstetrics. Of the 131 patients with USC, 81 (61.8%) were not referred to genetics or recommended for genetic testing by their oncologist, 45 (34.4%) were recommended to undergo genetic testing/referred to genetic counselling and 5 (3.8%) had prior genetic testing (figure 1). Of the 45 patients recommended to undergo genetic testing, 16 (35.6%) did not complete genetic testing and 29 (64.4%) completed genetic testing. 22 (48.9%) of the 45 patients completed an appointment with a genetic counsellor who subsequently ordered testing. 11 patients did not see a genetic counsellor initially, but had testing ordered by their primary oncologist. Flow diagram for genetic testing referral and testing results of patients diagnosed with USC between 2019 and 2024. USC, uterine serous carcinoma; PV, pathogenic variant. Excluding individuals who had undergone genetic testing prior to USC diagnosis, we evaluated demographic and clinical factors that may have influenced the decision for genetics referral, finding statistically significant differences in age at diagnosis, personal and family history of cancer and year of diagnosis (table 1). Patients were more likely to be referred for genetic testing if they were younger at the time of diagnosis (median 69 years old in the referred group vs 72 years old in the not-referred group, p=0.018), had a personal history of cancer (prior to and other than their diagnosis of USC) (31.1% vs 14.8%, p=0.030) and were diagnosed in 2019 or 2020 (51.1% vs 37.1%, p=0.019). The referred group had a higher proportion of individuals of Asian and Black race (p=0.044), though interpretation must consider the smaller sample sizes. There were no statistically significant differences in ethnicity, insurance status, family history of breast or ovarian cancer or stage at diagnosis between the two groups. There was no significant difference in referral rate in those who were diagnosed in 2023 or 2024 (when NCCN testing guidelines were updated9) versus those diagnosed in 2019 through 2022 (42.5% vs. 32.6%, p=0.278). Among 45 patients referred for genetic testing, we analysed differences between those who did versus did not complete testing (online supplemental table 1). While this sample was small, our results did demonstrate a possible disparity in genetic testing completion based on race. Individuals who completed testing after referral were more likely to identify as White (58.6% vs 18.8%, p=0.008). The completion rate was highest in White patients at 85.0%, followed by Asian patients at 69.2%. Only one of four Black patients completed testing. There was a non-significant trend towards decreased testing completion in individuals identifying as Hispanic or Latino (71.0% of non-Hispanic patients completing testing vs 28.6% of Hispanic patients, p=0.079). There was no difference in testing by age, BMI, insurance status, personal history of cancer or family history of cancer. 81.8% of individuals who saw a genetic counsellor (18 out of 22) completed genetic testing. Among five individuals who completed genetic testing prior to USC diagnosis, 4 (80%) had a cancer-related PV identified (figure 1). Of the 29 patients who completed genetic testing following their diagnosis of USC, 5 (17.2%) had a PV identified. In total, 9 patients out of 34 total individuals who completed genetic testing (26.5%) were found to carry a cancer-related PV (figure 2). Seven (36.8%) out of 19 individuals with a family history of breast or ovarian cancer carried a PV, while 8 (50%) out of 16 individuals with a personal history of cancer other than USC carried a PV. In the 18 patients who underwent genetic testing without a personal history of cancer other than USC, the PV prevalence was 5.56%. Pathogenic variants found on germline genetic testing of 34 patients with USC. PV, pathogenic variant. PVs were identified in BRCA2 (n=2), MSH6 (n=2), ATM (n=1), BRCA1 (n=1), BRIP1 (n=1), CHEK2 (n=1) and PMS2 (n=1) (table 2). HGVS verification details are provided in table 2. In individuals with USC who underwent genetic testing, Lynch syndrome prevalence was 8.8% (3/34), familial breast and ovarian cancer syndrome prevalence was 8.8% (3/34) and 8.8% (3/34) of others carried another genetic susceptibility to breast cancer. Of the individuals with Lynch syndrome, one tumour demonstrated isolated MSH6 mismatch repair deficiency on immunohistochemistry which was concordant with the germline MSH6 PV. The other two tumours were mismatch repair intact and a single somatic PV in MSH6 or PMS2 was identified on tumour sequencing. Table 2Cancer-related pathogenic variants found on germline genetic testing in patients with uterine serous carcinomaGeneN (total 34)Variant (HGVS coding)Variant (HGVS protein)ClinVar designationGermline oncologic condition (phenotype MIM number)Age at diagnosisStage at diagnosisPrior history of cancer BRCA2 2 (5.9%)c.3217C>Tp.Gln1073TerPathogenicFamilial breast-ovarian cancer (612555)51IBreastc.470_474delp.Lys157fsPathogenicFamilial breast-ovarian cancer (612555)74IVBreast BRCA1 1 (2.9%)UnknownUnknown–Familial breast-ovarian cancer (604370)70IIIBreast, kidney MSH6 2 (5.9%)c.3984_3987dupp.Leu1330fsPathogenicLynch Syndrome (614350)69ILungc.3416dupp.Lys1140fsPathogenic/Likely PathogenicLynch Syndrome (614350)55ILung PMS2 1 (2.9%)c.2156delp.Gln719fsPathogenicLynch Syndrome (614337)66ILung ATM 1 (2.9%)c.3852delp.Asp1285fsPathogenic/likely pathogenicSusceptibility to breast cancer (114480)57IBreast BRIP1 1 (2.9%)c.2392C>Tp.Arg798TerPathogenicSusceptibility to breast cancer, early-onset (114480)66IVNone CHEK2 1 (2.9%)c.1263delp.Ser422fsPathogenic/likely pathogenicTumour predisposition syndrome, breast/prostate/colorectal (609265)74IIIBreastNo pathogenic variant identified25 (73.5%)–––––––HGVS, Human Genome Variation Society; MIM, Mendelian Inheritance in Man. Cancer-related pathogenic variants found on germline genetic testing in patients with uterine serous carcinoma HGVS, Human Genome Variation Society; MIM, Mendelian Inheritance in Man. The only significant difference between individuals who were found to carry a PV versus those who had no PVs on germline testing was a prior personal history of cancer other than USC (88.9% vs 32%, p=0.006, online supplemental table 2) and disease status at last follow-up (no recurrence 100% vs recurrence 56%, p=0.017). On both crude and adjusted logistic regression, only prior personal history of cancer was found to be significantly associated with discovery of a PV on germline genetic testing (online supplemental table 3, adjusted OR 42.9, 95% CI 1.28 to 1437.1, p=0.036). Age at diagnosis, BMI, stage III or IV at diagnosis, and family history of breast/ovarian cancer was not significantly associated with PV identification. Of the individuals with a PV, four had a prior history of breast cancer, three had a prior history of lung cancer, one had a history of kidney cancer and one had a history of both breast and kidney cancer (additional clinical details are available in online supplemental table 4). One individual had no personal history of cancer. Of the patients with breast cancer, three had undergone genetic testing after their breast cancer diagnosis (prior to USC diagnosis) and one had limited BRCA1/2 testing but expanded to multigene panel testing after USC diagnosis. The remaining patient would not have qualified for genetic testing per current NCCN guidelines.9 The patients with lung cancer (pathologies including one neuroendocrine carcinoma and two non-small cell lung cancers) had very remote or no tobacco exposure history. One individual who underwent genetic testing was found to carry a germline variant in APC (c.3920T>A, p.Ile1307Lys) that was not considered actionable in our primary analysis. Including it as a cancer-related PV increases the PV prevalence rate to 29.4% in our cohort but would not change overall significant associations seen with personal history of cancer. Discussion This retrospective cohort study analysed the referral patterns and germline genetic testing outcomes of patients with USC seen at an academic institution within a 5-year period. While a minority of individuals ultimately underwent genetic testing, there was a high rate of germline PV identification in this selected group. Although germline testing results do not impact uterine cancer treatment at this time, PV identification can significantly impact both patients and their families’ future cancer screenings and prevention strategies. This high prevalence suggests undertesting, and that PVs are being missed in patients who were not offered a genetic test. Among all patients who completed genetic testing in our cohort, 26.5% had a PV found in a cancer-related gene, including an 8.8% prevalence of familial breast and ovarian cancer syndrome and 8.8% prevalence of Lynch syndrome. In the unselected US population, germline PVs in BRCA1 and BRCA2 are only found in about 0.5% of individuals and Lynch syndrome in about 0.3% of individuals, suggesting that the findings are not due to chance.19–21 Previous studies looking at germline genetic testing for USC have reported lower but still elevated rates of PV ranging from 4.6% to 12.5%. Pennington et al assessed 151 subjects with USC for germline mutations in 30 tumour suppressor genes, finding 4.6% carried variants in BRCA1, TP53, CHEK2 and MSH6.22 Long and colleagues similarly tested a consecutively collected cohort of 135 patients with USC, finding PVs in 6.7%, including in BRCA1, BRIP1, CHEK2, MSH6, PMS2, ATM, NF1 and TP53.11 A retrospective cohort of individuals with USC at another tertiary care institution reported a PV prevalence of 12.5% (3/24) including PVs in BRCA1, BRCA2 and MSH6.23 Our PV prevalence is higher than in these prior reports, likely due to both smaller sample size and selection of more high-risk individuals for genetic testing, including younger individuals at time of diagnosis and those with a personal or family history of breast cancer. We conducted sensitivity analysis excluding patients with a prior personal history of cancer, finding a PV prevalence of 5.56%, which is more in line with previous reports but still reflects clinically meaningful benefit for identification. Importantly, many of the patients with PVs identified in this study would likely not have qualified for germline genetic testing based on their prior cancer diagnoses (eg, lung, kidney) or family history. Although four of the five patients with prior breast cancer had completed panel or limited genetic testing prior to USC diagnosis, one patient had not qualified for genetic testing based on prior criteria. In total, four additional individuals were found to carry actionable PVs due to testing triggered by their USC diagnosis who would otherwise not have qualified based on current guidelines. Given these findings and the evidence previously published, universal germline testing in patients with USC is likely to identify a significant number of individuals with actionable PVs. Based on testing after their USC diagnosis, patients in this cohort would be eligible for increased breast and colorectal cancer screenings and their families can benefit from cascade testing and subsequent cancer risk reduction strategies. Germline testing in this cohort identified variants conferring susceptibility to familial breast and ovarian cancer, susceptibility to breast cancer, Lynch syndrome and increased risk of colon cancer. Many prior studies have focused on categorising the risk of USC and other endometrial cancers in individuals carrying pathogenic germline variants in BRCA1 and BRCA2 mutations, finding significantly higher risks of endometrial cancer, with a particular association between BRCA1 mutations and USC,5 23 while other studies have questioned the link between these variants and USC.24 When research focuses on the inverse relationship (ie, chance of PVs in patients with USC), then testing reveals higher than expected rates of actionable PVs in a broader range of cancer-related genes. This highlights the need to use expanded gene panels for germline genetic testing in this population. Larger prospective cohort studies of individuals with hereditary cancer syndromes other than BRCA1/2 would more definitively support (or negate) any associations with USC and should be encouraged. Starting in 2023, NCCN guidelines for genetic/familial risk assessment for breast, ovarian, pancreatic and prostate cancer supported the consideration of germline genetic testing for those with a personal history of USC. We did not see an uptake in genetic testing referral after 2023. Given the updated recommendations and the high prevalence of PV found in our tested population, these referral practices can and should improve. Only 34.4% of individuals with USC were referred for genetic counselling or testing, despite a robust Cancer Genetics programme at our institution. This is similar to a 47.2% referral rate seen at another tertiary care institution over a 7-year interval.25 Although over 80% of individuals who saw a genetic counsellor completed testing, there was significant attrition of individuals between patient referral and a genetic counselling appointment. Both increased clinician education regarding current guidelines and mainstreaming of genetic testing (where non-genetic clinicians can perform initial patient pretest counselling, test ordering and results disclosure) can help improve these rates.26 27 Of the 45 patients who were recommended to undergo genetic testing or referred to cancer genetics at our institution, 35% did not complete testing. While this was a small sample, the results did suggest a possible disparity in genetic testing completion based on race and ethnicity. These findings align with other studies showing lower rates of germline genetic testing among Asian, Black and Hispanic patients with cancer.28 29 We also demonstrated a trend towards decreased testing completion in patients who did not speak English. It is imperative that continued work is done to identify and help address these disparities. Strengths of this study include that this is a contemporary comprehensive assessment of all cases of USC diagnosed at a leading academic cancer centre. There was sufficient follow-up time built in to capture genetic testing referrals and data were abstracted from internal and outside institutions as available in the electronic health record. The study is limited by the smaller sample size and limited diversity that comes with a single-institution cohort, limiting secondary statistical analyses. Furthermore, selection bias affecting genetic counselling/testing referrals likely contributes to the higher PV prevalence seen in our cohort. In conclusion, this study found that germline genetic testing of patients with USC has important clinical impact with 26.5% of tested patients carrying a cancer-related PV. These results carry meaningful implications on the care of patients and their families. Accordingly, germline genetic testing should be offered to all patients with USC and studies should further investigate PV prevalence at the population level. Supplementary material Acknowledgements We thank our colleagues from Stanford University for their support for this project and acknowledge our patients and the work of the genetic counsellors and oncologists. Footnotes Funding: The authors have not declared a specific grant for this research from any funding agency in the public, commercial or not-for-profit sectors. Provenance and peer review: Not commissioned; externally peer reviewed. Patient consent for publication: Not applicable. Data availability free text: As data are potentially identifiable, data can be made upon request with adequate data use and IRB agreements. Ethics approval: This study was reviewed and approved by the Stanford University Institutional Review Board, protocol number 75776. Data availability statement Data are available upon reasonable request. References Associated Data Supplementary Materials Data Availability Statement Data are available upon reasonable request.