Socioeconomic and urban-rural disparities in genome-matched treatment receipt and survival after genomic tumor testing

Abstract

Background

Emerging cancer treatments are often most available to socially advantaged individuals. This study examines the relationship of patient educational attainment, income level, and rurality to the receipt of genome-matched treatment and overall survival.

Methods

Survey and clinical data were collected from patients with cancer (n = 1258) enrolled in the Maine Cancer Genomics Initiative. Logistic regression models examined whether receipt of genome-matched treatment differed by patient education, income, and rurality. Kaplan–Meier curves and Cox regression were conducted to evaluate 12-month mortality. We completed additional exploratory analyses using Kaplan–Meier curves and Cox models stratified by receipt of genome-matched treatment. Logistic and Cox regression models were adjusted for age and gender.

Results

Educational attainment, income level, and rurality were not associated with genome-matched treatment receipt. Of 1258 patients, 462 (36.7%) died within 365 days of consent. Mortality risk was associated with lower educational attainment (hazard ratio [HR] = 1.30, 95% confidence interval [CI] = 1.06 to 1.59; P = .013). No statistically significant differences in mortality risk were observed for income level or rurality. Exploratory models suggest that patients who did not receive genome-matched treatment with lower educational attainment had higher mortality risk (HR = 1.36, 95% CI = 1.09 to 1.69; P = .006). For patients who did receive genome-matched treatment, there was no difference in mortality risk between the education groups (HR = 1.01, 95% CI = 0.56 to 1.81; P > .9).

Conclusion

Although there were no disparities in who received genome-matched treatment, we found a disparity in mortality associated with education level, which was more pronounced for patients who did not receive genome-matched treatment. Future research is warranted to investigate the intersectionality of social disadvantage with clinical outcomes to address survival disparities.

Historical and current structural inequities have produced disparities throughout the cancer care continuum, from detection to treatment to survival (1). Socioeconomic status (SES) refers to a person’s position within a power hierarchy and control over resources, and factors such as education level, income, and occupational prestige can serve as indicators of socioeconomic position (2). SES influences when people seek medical services for cancer, the care they receive, and the financial burden of care (3). Individuals with lower SES face barriers to treatment and poorer survival and have less access to information compared with those in higher SES groups (4-6). Additional drivers, such as where a person lives and the stigmatized status of a person’s racial identity, also have the potential to intersect with SES to drive cancer disparities. For example, Black patients who live in rural areas experience the greatest cancer burden (7). Issues of access are particularly important in the context of emerging cancer treatments that are not standard of care. Scientific advancements in emerging cancer treatments have the potential to improve care and survivorship (8). However, whether these treatments equitably improve population health is constrained by the degree to which they are accessible to all.

One emerging technology that is being increasingly used is large-panel genomic tumor testing. These tests analyze genomic variants and other biomarkers in cancer cells and can inform targeted, genome-matched treatments. Broadly, genome-matched treatments are drugs that target molecular pathways disrupted by specific mutations. Genomic technologies may create disparities because they are often provided in well-resourced cancer centers to socially advantaged individuals (9,10). Little is known about how social drivers, including educational attainment, income, and rurality, are associated with the likelihood of receiving genome-matched treatment and overall survival. Educational attainment is strongly related to health literacy, patient-provider relationships, and self-care (11), so it is plausibly related to how well patients might understand and advocate for emerging technology like genome-matched treatment.

There is some early evidence that people with lower SES are less likely to receive emerging cancer treatments. Patients with lung cancer were less likely to receive immunotherapy if they had lower educational attainment and income compared with high educational attainment and income patients (12). Treatment-related out-of-pocket costs were statistically significantly higher for lower-income patients receiving emerging hormonal therapy for advanced prostate cancer (13), which may result in patients with fewer resources forgoing these treatments. These SES disparities may be amplified by a patient’s geographic location. Rural oncology practices are less likely to have genomic testing resources (eg, protocols for biomarker testing, genomic tumor boards), which may affect access and use of genomic tumor testing (14). Additionally, few studies have examined the impact of SES on survival disparities for people who receive emerging treatments.

Given the limited literature examining the relationship between SES, receipt of emerging genome-matched cancer treatments, and clinical outcomes, our goal was to explore whether SES (educational attainment, household income) and geography (rural residence) were associated with the likelihood of a patient receiving genome-matched treatment and overall survival in a rural statewide genomic cancer initiative.

Methods

Study population and design

We used data from the Maine Cancer Genomics Initiative (MCGI), a statewide initiative that provided free large-panel genomic tumor testing to Maine oncology clinicians and their patients in predominantly rural communities (15). Clinicians were recruited to participate in the initiative by phone and site visits. Enrolled clinicians were given patient educational material about genomic tumor testing [see Anderson et al. (16) for informational brochures] and could offer their patients free large-panel genomic tumor testing. Patients who agreed to testing were offered the opportunity to participate in the observational study that consisted of completing surveys and having their electronic medical records abstracted.

Patient surveys were completed after enrollment and collected demographic information and patients’ knowledge and perceptions of genomic tumor testing. Treatments and overall survival were abstracted from medical records. This study cohort includes patients who enrolled in the initiative between July 2017 and October 2020. Patients with any cancer stage or World Health Organization grade (for primary brain cancers), any solid malignancy and treatment, and adequate functional status (Eastern Cooperative Oncology Group performance status 0-2) could enroll. Written informed consent was obtained from participants. The MCGI study protocol was reviewed and approved by the Western institutional review board. Participants completed surveys electronically using the online survey platform REDCap Cloud or by paper (later entered into REDCap Cloud by site coordinators).

Measures

Genome-matched treatment

Genomic tumor testing results were delivered to patients’ treating clinicians in a report. The reports identified potentially actionable tumor variants and biomarkers based on guidelines at the time (17) and the available US Food and Drug Administration–approved drugs or clinical trials that targeted those biomarkers [see previous publication for details (15)]. To understand whether patients received genome-matched treatment, we analyzed all treatments administered to patients after genomic tumor testing results were returned and within 1 year of enrollment. Genome-matched treatments were defined as patients receiving a drug that targeted a gene variant or biomarker identified on their genomic tumor testing [full details of definition in Anderson et al. (18)].

Mortality

We defined mortality as a patient death within 12 months of the study enrollment date. Electronic medical records were reviewed for indications that patients were alive by the study endpoint. For patients with an absence of active documentation after 12 months, site research coordinators searched obituaries or spoke to close relatives to confirm a date of death. If no documentation of death was found, then patients were treated as alive at the end of 12 months.

Sociodemographic and clinical factors

Participants answered survey questions about their educational attainment, income level, zip code, age, gender, race, ethnicity, health insurance, and physical quality of life (19). Educational attainment was reported as the following categories: less than high school, high school graduate or general educational development, some college or trade school, bachelor’s degree, and graduate degree. To visualize stratified Kaplan–Meier survival curves, we collapsed the education levels into 2 categories. Two classification groups for educational attainment were created based on the distribution observed in our sample: individuals who reported less than high school, high school graduate, or general educational development were grouped into lower educational attainment, and those who earned some college or trade school, bachelor’s, and graduate degree were grouped into a higher educational attainment category.

Income level was reported as 1 of 6 categories: less than $25 000, $25 000-$49 999, $50 000-$74 999, $75 000-$100 000, more than $100 000, and “I don’t know.” For this analysis, reports of “I don’t know” were treated as missing. We collapsed the remaining 5 income categories into 2 groups: higher income and lower income. We used Maine’s median income in 2021 of $63 182 (20) as the cut point for the 2 groups; those earning less than $49 999 were grouped into the lower income category and those earning above $50 000 were grouped into higher income. A sensitivity analysis using 3 level categories for educational attainment and income level was also conducted and is included in the Supplemental Materials (available online).

Home zip codes were used to identify rural-urban commuting area (RUCA) codes, which use measures of population density, urbanization, and daily commuting to assign zip codes to urban or rural categories (21). We then classified our RUCA codes according to the Washington, Wyoming, Alaska, Montana, and Idaho Rural Health Research Center’s Four Category classification (22); however, because of small sample size in some categories, we collapsed rurality into 3 groups using the 2010 secondary RUCA codes (21): small and isolated rural, large rural, or urban (22).

Cancer site and clinical disease status at time of enrollment were collected from participants’ electronic medical records. Cancer locations were grouped into lung, breast, colon, prostate, gynecologic, brain, and all others (see Supplementary Table 3, available online, for all other cancer types). Clinical disease status at time of enrollment (recorded as clinical cancer stage) were categorized as stage I, stage II, stage III, stage IV, and unknown.

Statistical analysis

Summary statistics were calculated for the sociodemographic variables and clinical variables (cancer site and clinical cancer stage) (Table 1). To examine sociodemographic differences in receipt of genome-matched treatment, we fit separate logistic regression models with educational attainment, income level, and rurality as predictor variables and receipt of genome-matched treatment as the outcome variable (Table 2).

Table 1.

Sample demographics

Characteristic	No. (%)
	(n = 1258)
Age
Median (IQR), ya	65 (14)
Missing	154
Gender
Female	662 (60)
Male	442 (40)
Missing	154
Race
African or African American	6 (0.5)
American Indian or Alaskan Native	6 (0.5)
Asian	4 (0.3)
Multiple	14 (1.1)
Not given or other	175 (14)
White	1053 (84)
Ethnicity
Hispanic	12 (1.1)
Non-Hispanic	1076 (99)
Missing	170
Insurance
Medicare and Medicaid	76 (7.9)
Medicare	504 (53)
Medicaid	46 (4.8)
Private	331 (35)
Missing	301
Clinical cancer stageb
Stage I	50 (4.1)
Stage II	51 (4.2)
Stage III	180 (15)
Stage IV	930 (77)
Missing	47
Cancer sitec
Lung	159 (13)
Breast	130 (10)
Gynecologic	255 (20)
Brain	132 (11)
Colon	137 (11)
Prostate	64 (5.1)
Other	380 (30)
Missing	1
Genome-matched treatment status
Received genome-matched treatment	206 (16)
No genome-matched treatment	1052 (84)
Education
Higher educational attainment	658 (62)
Lower educational attainment	408 (38)
Missing	192
Household income
Higher income	388 (40)
Lower income	590 (60)
Missing	280
Rurality
Urban	464 (44)
Large rural	216 (21)
Small and isolated rural	371 (35)
Missing	207

^aIQR = interquartile range.

^bBrain cancers are rated by grade (1, 2, 3, 4) and are grouped according to respective stage (I, II, III, IV).

^cSee Supplementary Table 3 (available online) for all cancers categorized as Other.

Table 2.

Results of age- and gender-adjusted associations of receipt of genome-matched treatment with socioeconomic and geographic measures^a

Characteristic	OR (95% CI)	P
Educationa		.8
Higher educational attainment	[Reference]
Lower educational attainment	1.05 (0.75 to 1.47)	.8
Incomea		.6
Higher income	[Reference]
Lower income	0.91 (0.64 to 1.29)	.6
Ruralitya		.2
Urban	[Reference]
Large rural	0.77 (0.49 to 1.18)	.2
Small and isolated rural	0.71 (0.49 to 1.03)	.070

^aAll models are adjusted for patient age and gender. CI = confidence interval; OR = odds ratio.

To assess socioeconomic and geographic differences in mortality, we first examined overall survival outcomes using Kaplan–Meier survival curves and then conducted separate Cox proportional hazards regression models for each predictor (education, income, rurality) with mortality as the outcome (Table 3). A previous study [Anderson et al. (18)] found receipt of genome-matched treatment had large main effects on mortality, so we also completed exploratory Kaplan–Meier survival curves stratified by receipt of genome-matched treatment and tested the interaction between the predictors and receipt of genome-matched treatment in Cox proportional hazards regression models stratified by receipt of genome-matched treatment (Table 4). Because of the low frequency of events in our sample, we did not model genome-matched treatment receipt as an interaction term with education, income, or rurality because we had insufficient statistical power to precisely estimate interactive associations. We did not adjust for physical quality of life, clinical cancer stage, and cancer site in the models because they are believed to be on the casual pathway from educational attainment, income level, and rurality to overall survival (23,24). That is, clinical cancer stage, cancer site, and quality of life are “downstream” from our exposures of interest and adjusting for them would mask the indirect effects of educational attainment, income level, and rurality on overall survival (24). Ideally these relationships would be evaluated as moderated mediation models, but such modeling is beyond the scope and sample size of this study (see Supplementary Figure 1, available online). All hypothesis tests were 2-sided, and all models were adjusted for age and gender. Gender was used as a proxy for sex assigned at birth to control for sex-specific factors. A P value less than .05 was considered statistically significant. Additionally, because previous research by our lab found that SES is associated with patient knowledge and perceptions of genomic tumor testing (25), we conducted an exploratory analysis of those variables (see Supplementary Materials, available online). Analyses were conducted using R version 4.2.1 (26).

Table 3.

Results of age- and gender-adjusted associations of mortality with socioeconomic measures

Characteristic	HR (95% CI)	P
Educationa		.013
Higher educational attainment	[Reference]
Lower educational attainment	1.30 (1.06 to 1.59)	.013
Incomea		.3
Higher income	[Reference]
Lower income	1.12 (0.90 to 1.39)	.3
Ruralitya		.5
Urban	[Reference]
Large rural	1.17 (0.90 to 1.52)	.3
Small and isolated rural	1.02 (0.81 to 1.28)	.9

^aAll models are adjusted for patient age and gender. CI = confidence interval; HR = hazard ratio.

Table 4.

Results of age- and gender-adjusted associations of mortality with socioeconomic measures, stratified by receipt of genome-matched treatment

Characteristic	Genome-matched treatment: No		Genome-matched treatment: Yes
	HR (95% CI)	P	HR (95% CI)	P
Educationa		.006		>.9
Higher educational attainment	[Reference]		[Reference]
Lower educational attainment	1.36 (1.09 to 1.69)	.006	1.01 (0.56 to 1.81)	>.9
Incomea		.14		.2
Higher income	[Reference]		[Reference]
Lower income	1.19 (0.94 to 1.51)	.15	0.70 (0.39 to 1.26)	.2
Ruralitya		.5		>.9
Urban	[Reference]		[Reference]
Large rural	1.18 (0.89 to 1.57)	.2	0.92 (0.43 to 1.96)	.8
Small and isolated rural	1.01 (0.78 to 1.29)	>.9	1.01 (0.54 to 1.87)	>.9

^aAll models are adjusted for patient age and gender. CI = confidence interval; HR = hazard ratio.

Results

Descriptive statistics

A total of 1603 adult patient participants were enrolled in the MCGI, 1502 (93.7%) patients had genomic tumor testing performed, and 1290 (85.9%) patients had genomic tumor testing results returned. For this analysis, we excluded patients with no actionable variants identified (2.5%). The final cohort of patients, 1258, had at least 1 potentially actionable variant identified and were eligible to receive genome-matched treatment, and 206 (16%) received at least 1 genome-matched treatment. Of the 1258 patients, 462 (36.7%) died within 365 days of consent.

The study population had a median age of 65 years (Interquartile range = 14), more than half (60%) were female, most (84%) were White, 38% reported lower educational attainment (no college), 60% had lower household income (less than $49 999), and 56% lived in rural areas. Most patients entered the study with clinical cancer stage IV (77%) with the most prevalent cancer sites being other (30%), gynecologic (20%), lung (13%), colon (11%), and brain (11%).

Associations of socioeconomic and geographic predictors with receipt of genome-matched treatment

We assessed whether the receipt of genome-matched treatment was associated with educational attainment, income level, or rurality controlling for age and gender (Table 2; see Supplementary Table 1, available online, for the full model with all independent variables). There were no statistically significant differences in receipt of genome-matched treatment by educational attainment (odds ratio [OR] = 1.05, 95% confidence interval [CI] = 0.75 to 1.47; P = .8), household income (OR = 0.91, 95% CI = 0.64 to 1.29; P = .6), or rurality (P = .2). Patients residing in small and isolated regions were slightly less likely to receive genome-matched treatment compared with urban patients (OR = 0.71, 95% CI = 0.49 to 1.03; P = .070); however, this difference was not statistically significant. Additionally, we did not observe any relationships between knowledge and perceptions of genomic tumor testing and receipt of genome-matched treatment in the exploratory analyses (see Supplementary Materials, available online).

Associations of education, income, and rurality with 12-month overall survival

Patients with lower educational attainment had worse 12-month overall survival compared with patients with higher educational attainment (P = .002) (Figure 1A). There were no statistically significant differences in 12-month overall survival by income level (P = .1) (Figure 1B) or rurality (P = .4) (Figure 1C). After including age and gender in the models, lower educational attainment remained associated with higher hazard of mortality compared with higher educational attainment (HR = 1.30, 95% CI = 1.06 to 1.59; P = .013), and there remained no differences in hazard of 12-month mortality by patient income level (P = .3) or rurality (P = .5) (Table 3). See Supplementary Table 2 (available online) for the full model with all independent variables.

Figure 1.

One-year overall survival in 1258 cancer patients in relation to (A) educational attainment, (B) income level, and (C) rurality. Error bars represent 95% confidence intervals. ^*Statistically significant difference in overall survival (P < .05). iso = isolated; NS = not statistically significant.

Exploratory analysis of genome-matched treatment, education, income, and rurality with 12-month overall survival

Exploratory Kaplan–Maier survival curves stratified by receipt of genome-matched treatment showed that for patients who did not receive genome-matched treatment, those with lower educational attainment and lower income had worse overall survival than higher educational attainment and higher income patients (P = .001 and P = .04) (Figure 2, Aa and Bc). However, for patients who did receive genome-matched treatment, there were no differences in overall survival by educational attainment or income level (P = .9 and P = .2) (Figure 2, Ab and Bd). No statistically significant differences in overall survival were observed for rurality groups in either genome-matched treatment category (P = .3 and P = 1.0) (Figure 2, Ce and Cf).

Figure 2.

One-year overall survival in 1258 cancer patients in relation to (A) educational attainment, (B) income level, and (C) rurality stratified by receipt of genome-matched treatment. Error bars represent 95% confidence intervals. ^*Statistically significant difference in overall survival (P < .05). GMT = genome-matched treatment; iso = isolated; NS = not statistically significant.

Similar effects were observed after adjusting for age and gender (Table 4; see Supplementary Table 4, available online, for the full model with all independent variables). Lower educational attainment was associated with higher hazard of 12-month mortality in the group who did not receive genome-matched treatment (HR= 1.36, 95% CI = 1.09 to 1.69; P = .006), and there was no educational attainment group difference in mortality for those who received genome-matched treatment (HR = 1.01, 95% CI = 0.56 to 1.81; P > .9). There were no statistically significant differences in mortality by income level or rurality in either group that received genome-matched treatment. Lower income patients and those living in large rural areas who did not receive genome-matched treatment had higher hazard of mortality compared with patients earning higher income and living in urban areas, but these differences were not statistically significant (HR = 1.19, 95% CI = 0.94 to 1.51; P = .15; HR = 1.18, 95% CI = 0.89 to 1.57; P = .2, respectively).

Discussion

This study investigated SES and geographic differences in clinical outcomes from a large statewide genomics initiative. Our main finding is that patients with lower educational attainment had worse 12-month overall survival compared with those with higher educational attainment. Further, this disparity was most pronounced among patients who did not receive genome-matched treatment.

Contrary to our initial hypotheses, lower SES was not associated with the likelihood of receiving genome-matched treatment. We originally hypothesized that socially disadvantaged patients may be less likely to receive genome-matched treatment because of the structural barriers they face. For instance, people with lower income might be less likely to utilize genome-matched treatments because of high out-of-pocket costs or travel to enroll in clinical trials. This null finding contributes to a mixed body of literature. Several studies in a recent systematic review (27) found that lower SES was associated with worse access to treatment (eg, stem cell transplantation); however, there were multiple studies that did not observe this relationship. Another study found that patients with higher SES were more likely to receive CART-cell therapy (an emerging cancer therapy) with higher disease burden (28). This suggests higher SES parents might have the ability to advocate for emerging treatment or travel to treatment centers in the context of children with more severe disease. In the present study, it is possible we observed no effect of SES because testing was offered at no cost through a statewide genomics initiative. Genomic tumor boards were available through the initiative to support clinical decision making, and initiative staff assisted with drug acquisition through clinical trials and compassionate use programs. Therefore, the impact of SES without the support of a state-based initiative is unknown. Additionally, the SES disparities examined in this study focus on one part of the cancer care continuum (receipt of emerging treatment), but SES impacts all parts of the care continuum (eg, screening, diagnosis), and future work should prioritize studying SES disparities across the continuum.

We found that patients with lower educational attainment had worse 12-month overall survival compared with those with higher educational attainment, consistent with large cancer and noncancer population studies (27,29). Further, exploratory analyses stratified by receipt of genome-matched treatment found patients who did not receive genome-matched treatment with lower educational attainment had worse overall survival compared with higher educational attainment patients. However, for patients who did receive genome-matched treatment, there was no educational attainment difference in overall survival. This finding is similar to a recent study that observed neither household poverty nor low-neighborhood opportunity was associated with survival in a cohort of children with acute lymphoblastic leukemia who received CAR T-cell therapy (28).

The finding that patients who received genome-matched treatment did not show a mortality disparity by educational attainment level should be interpreted with caution. This comparison was between small groups (n = 68 for lower and n = 106 for higher educational attainment who received genome-matched treatment), so we had limited power to detect a difference. Moreover, patients who received genome-matched treatment, and those who did not, could have differed in other ways such as functional ability, overall health, or access to other resources, such as palliative care services or family caregivers. For these reasons, we do not believe that giving all patients genome-matched treatment would necessarily eliminate all cancer survival disparities. Other work shows that even when access to emerging treatments is equitable, racial differences in overall survival remain (30,31), likely because of inequities in social drivers of health outside the cancer care context (housing, transportation, etc) (32).

This study had several limitations that qualify the findings and call for further research. First, it was not possible to model quality of life, clinical cancer stage, or cancer site as mediators and moderators on the casual pathway because of the low frequency of events in our sample. However, such an analysis should be prioritized in the future (see Supplementary Figure 1, available online, for more details). Second, this study took place in the context of a unique genomics initiative that limits generalizability to other contexts. Genomic tumor testing was available to all Maine cancer patients at no cost, outside the traditional insurance model. Genomic tumor boards and other supports were available through the initiative that are not always available. These factors may have equalized the probability of all patients receiving genome-matched treatments. The study sample also lacked racial and ethnic diversity, though the sample is regionally representative because the study population was 99% non-Hispanic and 84% White, which is fairly similar to Maine’s 2020 population that is 98% non-Hispanic and 90.8% White (33). A future direction could be enhanced outreach to diverse communities to better understand the receipt of genome-matched treatment and overall survival across all communities. Additionally, interventions are needed to reduce cancer disparities for people who do not receive genome-matched treatment or other emerging treatments.

One feature of the MCGI study was broad inclusion criteria. Clinicians could enroll patients with any solid tumor at any stage. Clinicians frequently enrolled patients with later stage cancers (77% were stage IV or grade 4) but also enrolled patients with earlier stages of cancer (23%). Sometimes clinicians wanted to explore genomic tumor testing options for patients who failed other standard-of-care treatments. Other times, patients with earlier cancer stages were tested to have information available for later. There was also a range of different cancer types represented with higher frequencies of brain and gynecological cancer.

We found that lower SES did not influence the likelihood of receiving genome-matched treatment but that there was a relationship between lower educational attainment and 12-month overall survival. This difference was more pronounced for patients who did not receive genome-matched treatment. Because of the unique nature of this genomics initiative, subgroup sample sizes, and unmeasured confounders, more research is warranted to fully understand disparities related to SES and to ensure the equitable use and benefit from emerging cancer technologies.

(1)

Appendix 1

Additional members of the Maine Cancer Genomics Initiative are as follows:

Maine Cancer Genomics Initiative Steering Committee Members

Central Maine Medical Center, Lewiston: Nicholette Erickson; Dahl-Case Pathology Associates: Mayur Movalia, Marek Skacel; Jefferson Cary Cancer Center, Caribou: Allan Espinosa; MaineGeneral Medical Center, Augusta: Ridhi Gupta, Rachit Kumar, Richard Polkinghorn; Maine Medical Center, Portland: Christopher Darus, Scot Remick; Maine Medical Center, Spectrum Medical Group, Portland: Robert Christman, Karen Rasmussen; New England Cancer Specialists, Scarborough: Christian Thomas; Northern Light Eastern Maine Medical Center, Bangor: Philip Brooks, Catherine Chodkiewicz, Antoine Harb, Sarah Sinclair; Southern Maine Health Care Biddeford: Peter Rubin: Waldo County General Hospital, Belfast: Elizabeth Connelly; York Hospital, York: Peter Georges; the Jackson Laboratory: Jennifer Bourne, Linda Choquette, Ken Fasman, Cristen Flewellen, Emily Edelman, Lory Guerrette, Petra Helbig, Susan Mockus, Kate Reed, Jens Rueter, Kunal Sanghavi.

Center for Interdisciplinary Population and Health Research, MaineHealth Institute for Research

Eric Anderson, Sumayo Awale, Jessica DiBiase, John DiPalazzo, Anny Fenton, Cara Frankenfeld, Caitlin Gutheil, Paul Han, Ally Hinton, Michael Kohut, Susan Leeds, Lee Lucas, Elizabeth Scharnetzki, Leo Waterston, Lisbeth Wierda.

Data availability

The data underlying this article are from the patient participants who opted in to have their data included in the MCGI registry (>95% of all patient participants in this manuscript) and are available for future cancer-related research through the MCGI Registry. Qualified researchers can apply for access to the datasets via the MCGI Registry by contacting mcgi@jax.org. If a request is approved, the datasets will be made available via data use agreements with the Jackson Laboratory.

Author contributions

Jessica DiBiase, MPH (Conceptualization; Data curation; Formal analysis; Investigation; Methodology; Project administration; Visualization; Writing—original draft; Writing—review & editing), Elizabeth Scharnetzki, PhD (Conceptualization; Formal analysis; Methodology; Visualization; Writing—review & editing), Emily Edelman, MS (Investigation; Writing—review & editing), E. Kate Reed, MPH, ScM (Investigation; Writing—review & editing), Petra Helbig, MSc (Data curation; Investigation; Project administration; Resources; Validation; Writing—review & editing), Jens Rueter, MD (Conceptualization; Data curation; Funding acquisition; Investigation; Project administration; Resources; Writing—review & editing), Susan Miesfeldt, MD (Writing—review & editing), Cara L. Frankenfeld, PhD (Conceptualization; Formal analysis; Methodology; Supervision; Validation; Writing—review & editing), Paul K. J. Han, MD, MA, MPH (Conceptualization; Data curation; Formal analysis; Funding acquisition; Methodology; Project administration; Resources; Supervision; Validation; Writing—review & editing), Elizabeth A. Jacobs, MD, MAPP (Conceptualization; Formal analysis; Funding acquisition; Methodology; Project administration; Writing—review & editing), and Eric C. Anderson, PhD (Conceptualization; Data curation; Formal analysis; Funding acquisition; Investigation; Methodology; Project administration; Supervision; Validation; Visualization; Writing—original draft; Writing—review & editing).

Funding

This work was supported by funding from Harold Alfond Foundation; the Jackson Laboratory; and the National Center for Advancing Translational Sciences, National Institutes of Health (Grant Number KL2TR002545 to ECA). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Conflicts of interest

Jens Rueter reports financial support was provided by the Harold Alfond Foundation. Jens Rueter reports financial support was provided by the Jackson Laboratory. Jessica F. DiBiase reports a relationship with National Society of Genetic Counselors that includes travel reimbursement. Jens Rueter reports a relationship with Massachusetts Association of Practicing Urologists that includes speaking and lecture fees. Jens Rueter reports a relationship with Advisory Board for American Society of Clinical Oncology that includes board membership. Elizabeth A. Jacobs reports a relationship with Association for Accreditation of Human Research Protection Programs that includes travel reimbursement. Elizabeth A. Jacobs reports a relationship with :-elect, Society for General Internal Medicine that includes board membership. Elizabeth A. Jacobs reports a relationship with Association for Accreditation of Human Research Protection Programs that includes board membership. Cara L. Frankenfeld reports a relationship with EpidStrategies, LLC, that includes consulting or advisory. Cara L. Frankenfeld reports a relationship with editorial board member for the Annals of Epidemiology that includes speaking and lecture fees. Cara L. Frankenfeld reports a relationship with American College of Epidemiology Board of Directors that includes board membership. Emily Edelman reports a relationship with Northwestern University Genetic Counseling Program Advisory Board that includes board membership. Emily Edelman reports a relationship with University of Maryland Genetic Counseling Program Advisory Board that includes board membership. Emily Edelman reports a relationship with National Society of Genetic Counselors Annual Conference Committee that includes board membership. If there are other authors, they declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.