Geographic Variation in Overscreening for Colorectal, Cervical, and Breast Cancer Among Older Adults

Key Points

Question  What are the prevalence and demographic characteristics associated with overscreening for colorectal, cervical, and breast cancers among older adults in the US?

Findings  In a large, nationally representative, cross-sectional telephone survey with 176 348 participants, more than 45% of older adults in the US reported being screened for colorectal, cervical, or breast cancer after the recommended upper age limit, and overscreening for cervical and breast cancer was more common in metropolitan areas. Life expectancy was not associated with overscreening.

Meaning  The findings of this study suggest that interventions to reduce overscreening among older adults are needed at the levels of patients, health care professionals, communities, and systems to reduce patient harms and costs and to increase health care efficiency.

Importance  National guidelines balance risks and benefits of population-level cancer screening among adults with average risk. Older adults are not recommended to receive routine screening, but many continue to be screened (ie, are overscreened).

Objective  To assess the prevalence of overscreening for colorectal, cervical, and breast cancers among older adults as well as differences in overscreening by metropolitan status.

Design, Setting, and Participants  The cross-sectional study examined responses to a telephone survey of 176 348 community-dwelling adults. Participants were included if they met age and sex criteria, and they were excluded from each cancer-specific subsample if they had a history of that cancer. Data came from the 2018 Behavioral Risk Factor Surveillance System, administered by the US Centers for Disease Control and Prevention.

Exposures  Metropolitan status, according to whether participants lived in a metropolitan statistical area.

Main Outcomes and Measures  Overscreening was assessed using US Preventive Services Task Force definitions, ie, whether participants self-reported having a screening after the recommended upper age limit for colorectal (75 years), cervical (65 years), or breast (74 years) cancer.

Results  Of 176 348 participants (155 411 [88.1%] women; mean [SE] age, 75.0 [0.04] years; 150 871 [85.6%] non-Hispanic white; 60 456 [34.3%] with nonmetropolitan residence) the cancer-specific subsamples contained 20 937 [11.9%] men and 34 244 [19.4%] women for colorectal cancer, 82 811 [47.0%] women for cervical cancer, and 38 356 [21.8%] women for breast cancer. Overall, 9461 men (59.3%; 95% CI, 57.6%-61.1%) were overscreened for colorectal cancer; 14 463 women (56.2%; 95% CI, 54.7%-57.6%), for colorectal cancer; 31 988 women (45.8%; 95% CI, 44.9%-46.7%), for cervical cancer; and 26 198 women (74.1%; 95% CI, 73.0%-75.3%), for breast cancer. Overscreening was more common in metropolitan than nonmetropolitan areas for colorectal cancer among women (adjusted odds ratio [aOR], 1.23; 95% CI, 1.08-1.39), cervical cancer (aOR, 1.20; 95% CI, 1.11-1.29), and breast cancer (aOR, 1.36; 95% CI, 1.17-1.57). Overscreening for cervical and breast cancers was also associated with having a usual source of care compared with not (eg, cervical cancer: aOR, 1.87; 95% CI, 1.56-2.25; breast cancer: aOR, 2.08; 95% CI, 1.58-2.76), good, very good, or excellent self-reported health compared with fair or poor self-reported health (eg, cervical cancer: aOR, 1.21; 95% CI, 1.11-1.32; breast cancer: aOR, 1.47; 95% CI, 1.28-1.69), an educational attainment greater than a high school diploma compared with a high school diploma or less (eg, cervical cancer: aOR, 1.14; 95% CI, 1.06-1.23; breast cancer: aOR, 1.30; 95% CI, 1.16-1.46), and being married or living as married compared with other marital status (eg, cervical cancer: OR, 1.36; 95% CI, 1.26-1.46; breast cancer: OR, 1.54; 95% CI, 1.34-1.77).

Conclusions and Relevance  In this study, overscreening for cancer among older adults was high, particularly for women living in metropolitan areas. Overscreening could be associated with health care access and patient-clinician relationships. Additional research on why overscreening persists and how to reduce overscreening is needed to minimize risks associated with cancer screening among older adults.

The US Preventive Services Task Force (USPSTF) recommends routine screening of average-risk, asymptomatic patients for cancers of the colon and rectum,¹ cervix,² and (female) breast.³ However, recommendations are discontinued after patients reach an upper age limit or develop a disqualifying condition (eg, limited life expectancy). These discontinuations are due to (1) lack of data, because older adults are excluded from screening trials; (2) lower expected benefits of screening because of more serious, noncancer comorbidities among older adults⁴; and (3) greater expected risks of screening among older adults (eg, gastrointestinal perforation from colonoscopy).⁵ Other organizations have similar recommendations for cancer screening, although details (eg, eligible ages) vary.^6,7

Overscreening refers to routine screening of patients older than the recommended upper age limit or with limited life expectancy.⁸ Preventing overscreening is challenging for several reasons. First, estimates of life expectancy can be imprecise.⁹ Second, patients and clinicians may prefer not to discuss life expectancy or factor it into clinical decision-making.¹⁰ Third, health care system incentives are frequently linked to high screening rates, encouraging screening regardless of whether it is appropriate for a given patient.¹¹ Fourth, public health campaigns often do not indicate upper age limits for screening, limiting awareness that older adults age out of routine screening.

National prevalence estimates for overscreening have not been reported, and it is unknown how overscreening varies among subgroups. Disparities in appropriate screening among subgroups are well documented; for example, adults in nonmetropolitan areas are less likely to be screened for colorectal cancer (CRC)¹² and breast cancer^13,14 than adults in metropolitan areas, potentially because of differences in access to screening.^15,16

This study fills these gaps by estimating the prevalence of overscreening for colorectal, cervical, and breast cancers and identifying differences in overscreening among subgroups, ie, based on patient characteristics, such as residency in a metropolitan community. An exploratory aim was to examine the association between life expectancy and overscreening.

Data came from the 2018 Behavioral Risk Factor Surveillance System (BRFSS),¹⁷ an annual, nationally representative data set managed by the US Centers for Disease Control and Prevention (CDC). As part of its core survey, BRFSS monitors health behaviors, chronic health conditions, and preventive health care in the US, including items about cancer screening, approximately every 2 years. Data collection procedures vary, but generally, state health departments contact community-dwelling adults 7 days a week each month using a sampling frame of landline and cellular telephone numbers. The 2018 state-specific response rates (using American Association of Public Opinion Research response rate 4¹⁸) ranged from 38.8% to 67.2% (mean [SD], 49.8% [6.9%]; median [interquartile range], 49.9% [44.2%-55.0%]).¹⁹ Complete details on BRFSS methodology are available online.¹⁷ Participants provided verbal informed consent for BRFSS participation. Data collection was approved by the institutional review board at the CDC and review boards of state health departments.¹⁷ The present study was exempt from additional institutional review because it involved analysis of publicly available, deidentified data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.

All variables were constructed from self-reported responses to BRFSS survey items. We examined overscreening for colorectal, cervical, and breast cancers using USPSTF recommendations.^1-3 Specific definitions of overscreening variables were constrained by data availability.¹⁷ For each cancer type, we assessed participant sex, age, receipt of screening (ie, colorectal: sigmoidoscopy, colonoscopy, or blood stool test; cervical: Papanicolaou test or human papillomavirus test; breast: mammography), and time since most recent screening.

Participants 76 years or older who were screened for CRC after age 75 years were coded as overscreened and all others as not overscreened (Table 1).¹ Women 66 years or older who self-reported screening for cervical cancer after age 65 years were coded as overscreened and all others as not overscreened.² Women 75 years or older who self-reported being screened for breast cancer after age 74 years were coded as overscreened and all others as not overscreened.³

Our primary independent variable was metropolitan residency, assessed with metropolitan statistical area (MSA) codes.¹⁷ We coded participants with valid MSA data as metropolitan if they lived in the center city of an MSA, inside the county containing the center city of an MSA, or inside a suburban county of an MSA; all others were coded as nonmetropolitan. We excluded participants from Guam, Puerto Rico, and the US Virgin Islands because they did not have valid MSA data.

Control variables included having a usual source of care, self-reported health, educational attainment, marital status, and race/ethnicity. In addition, in the analysis of cervical cancer overscreening, we controlled for history of hysterectomy.

Finally, we estimated 10-year mortality risk (ie, the converse of 10-year life expectancy) using a modified version of an index developed and validated in older adults by Lee et al.^20,21 The index is a weighted sum of mortality risk factors (range, 2.5-24.5) (eTable 1 in the Supplement). The items and scoring system for our mortality risk index was modified based on data availability in BRFSS.¹⁷

Analyses examined overscreening separately for colorectal, cervical, and breast cancers. Screening for CRC is recommended for men and women, but cervical and breast cancer screenings are only recommended for women; therefore, we stratified analyses of CRC overscreening by sex. Any participant with a history of colorectal, cervical, or breast cancer was excluded from analysis of overscreening for that cancer.

We generated descriptive statistics to summarize the independent variables for each cancer-specific subsample. Next, we calculated the prevalence and 95% CIs of each overscreening outcome. We used bivariate logistic regression to examine the association between metropolitan status and each outcome. Finally, we used multivariable logistic regression to adjust for control variables to quantify the independent association between metropolitan status and overscreening.

We performed 4 supplementary analyses. First, we repeated the multivariable analysis for cervical cancer overscreening, stratifying by participants’ hysterectomy history, given that most women with a total hysterectomy should not receive routine cervical cancer screening.² Second, to evaluate the extent of overscreening among the oldest participants, we repeated the multivariable analyses stratified by participant age (ie, <80 years vs ≥80 years at the time of BRFSS participation). Third, to evaluate the extent of overscreening among participants with the lowest estimated life expectancy (ie, converse of predicted mortality risk), we repeated the multivariable analyses stratified by scores on the modified mortality risk score (ie, <8 vs ≥8 score, given that Cruz et al²¹ demonstrated that estimated 10-year risk of mortality surpasses 50% for older adults with scores ≥8). Fourth, to assess the continuous association between life expectancy and overscreening, we used the multivariable models to calculate (1) the association between estimated mortality risk and overscreening and (2) the adjusted probability of overscreening for each cancer across scores on the modified mortality risk index. We plotted these adjusted probabilities, stratified by metropolitan status, to illustrate the association between mortality risk and overscreening.

Analyses were conducted in SAS version 9.4 (SAS Institute). All models incorporated sampling weights to account for complex survey design of BRFSS.¹⁷ Analyses used a 2-sided P < .05 to determine significance.

Of 176 348 participants (155 411 [88.1%] women; mean [SE] age, 75.0 [0.04] years; 150 871 [85.6%] non-Hispanic white; 60 456 [34.3%] with nonmetropolitan residence), we assessed overscreening among 20 937 (11.9%) men and 34 244 (19.4%) women for CRC, 82 811 (47.0%) women for cervical cancer, and 38 356 (21.8%) women for breast cancer (Table 2). More than 80% of participants lived in a metropolitan area in all 4 subsamples (eg, CRC sample of men: 13 894 participants [82.2%]; cervical cancer sample of women: 54 553 participants [82.2%]). Most participants had a usual source of care (>93% in all subsamples), reported good, very good, or excellent health (>72% in all subsamples), and were non-Hispanic white (>77% in all subsamples). Mean (SE) scores on the modified mortality risk index ranged from 7.03 (0.02) for the cervical cancer subsample to 10.54 (0.04) for the CRC in men subsample.

In this subsample, 9461 participants (59.3%; 95% CI, 57.6%-61.1%) were overscreened for CRC. Overscreening was more common in metropolitan (6392 of 11 163 participants [weighted percentage, 60.1%; 95% CI, 58.1%-62.1%]) vs nonmetropolitan areas (3069 of 5432 participants [weighted percentage, 55.8%; 95% CI, 52.5%-59.1%]) (P = .03), but this was not statistically significant in multivariable analysis (Table 3). Compared with other participants, overscreening was more common for Hispanic participants (eg, aOR for Hispanic vs white participants, 1.91; 95% CI, 1.24-2.95). Among men who were overscreened for CRC, the most commonly reported test received was colonoscopy (7146 of 9461 participants [weighted percentage, 74.7%; 95% CI, 72.7%-76.7%]) (eTable 2 in the Supplement).

In this subsample, 14 463 participants (56.2%; 95% CI, 54.7%-57.6%) were overscreened for CRC. Overscreening was more common in metropolitan areas (9731 of 17 750 participants [weighted percentage, 57.0%; 95% CI, 55.3%-58.8%]) vs nonmetropolitan areas (4732 of 9062 participants [weighted percentage, 51.9%; 95% CI, 49.4%-54.4%]) (P < .001), an association that held true in multivariable analysis (aOR, 1.23; 95% CI, 1.08-1.39) (Table 3). Among women who were overscreened for CRC, the most commonly reported test received was colonoscopy (10 543 of 14 463 participants [weighted percentage, 70.3%; 95% CI, 68.5%-72.2%]) (eTable 2 in the Supplement).

In this subsample, 31 988 participants (45.8%; 95% CI, 44.9%-46.7%) were overscreened for cervical cancer. Overscreening was more common in metropolitan areas (21 752 of 49 156 participants [weighted percentage, 47.1%; 95% CI, 46.1%-48.1%]) vs nonmetropolitan areas (10 236 of 25 514 participants [weighted percentage, 40.0%; 95% CI, 38.5%-41.4%]) (P < .001), an association that held true in multivariable analysis (aOR, 1.20; 95% CI, 1.11-1.29) (Table 3). Overscreening was also more common for participants with a usual source of care compared with those without (aOR, 1.87; 95% CI, 1.56-2.25); for participants with good, very good, or excellent self-reported health compared with those with fair or poor self-reported health (aOR, 1.21; 95% CI, 1.11-1.32); for participants who have obtained education beyond a high school degree compared with who had a high school degree or less (aOR, 1.14; 95% CI, 1.06-1.23); for participants who were married or living as married compared with those who were not (aOR, 1.36; 95% CI, 1.26-1.46); and for non-Hispanic black or Hispanic patients compared with white patients (non-Hispanic black: aOR, 2.05; 95% CI, 1.77-2.38; Hispanic: aOR, 2.41; 95% CI, 1.94-3.00). Among women who were overscreened for cervical cancer, almost all participants reported receiving a Papanicolaou test (31 593 of 31 988 participants [weighted percentage, 98.7%; 95% CI, 98.5%-99.0%]) (eTable 2 in the Supplement).

Overscreening was higher among women without a hysterectomy (20 752 of 41 487 participants [weighted percentage, 52.9%; 95% CI, 51.8%-54.1%]) than with a hysterectomy (11 181 of 33 062 participants [weighted percentage, 37.2%; 95% CI, 35.9%-38.5%]) (P < .001) (eTable 3 in the Supplement). For both subsamples, overscreening was more prevalent in metropolitan compared with nonmetropolitan areas, and the association between metropolitan residency and overscreening did not vary by hysterectomy history (overscreening for metropolitan vs nonmetropolitan residents with a hysterectomy: aOR, 1.17; 95% CI, 1.05-1.31; without a hysterectomy: aOR, 1.22; 95% CI, 1.10-1.34; P for interaction = .23).

In this subsample, 26 198 participants (74.1%; 95% CI, 73.0%-75.3%) were overscreened for breast cancer. Overscreening was more common in metropolitan areas (17 397 of 23 050 participants [weighted percentage, 75.4%; 95% CI, 74.1%-76.7%]) vs nonmetropolitan areas (8801 of 12 127 participants [weighted percentage, 68.4%; 95% CI, 65.9%-70.9%]) (P < .001), an association that held true in multivariable analysis (aOR, 1.36; 95% CI, 1.17-1.57) (Table 3). Compared with other participants, overscreening was more common for participants with a usual source of care compared with those without (aOR, 2.08; 95% CI, 1.58-2.76); for participants with good, very good, or excellent self-reported health compared with those with fair or poor self-reported health (aOR, 1.47; 95% CI, 1.28-1.69); for participants who have obtained education beyond a high school degree compared with those who had a high school degree or less (aOR, 1.30; 95% CI, 1.16-1.46); for participants who were married or living as married compared with those who were not (aOR, 1.54; 95% CI, 1.34-1.77); and for non-Hispanic black patients compared with white patients (non-Hispanic black: aOR, 1.48; 95% CI, 1.09-2.01). We found high prevalence of overscreening among older adults (>45% for each cancer); in fact, 5408 of 34 244 women aged 76 or older (weighted percentage, 17.1%; 95% CI, 16.1%-18.1%) were overscreened for all 3 cancers.

When comparing participants younger than 80 years with those 80 years or older, overscreening for CRC was higher among older than younger participants (eg, overscreening in men aged <80 years old, 3154 of 6914 participants [weighted percentage, 48.9%; 95% CI, 46.3%-51.5%]; overscreening in men aged ≥80 years, 6307 of 9681 participants [weighted percentage, 67.0%; 95% CI, 64.6%-69.3%]; P < .001), while overscreening for cervical and breast cancers was lower among older participants (eg, overscreening for breast cancer in women aged <80 years, 11 800 of 15 584 participants [weighted percentage, 76.5%; 95% CI, 74.9%-78.0%]; overscreening in women aged ≥80 years, 14 398 of 19 593 participants [weighted percentage, 72.2%; 95% CI, 70.6%-73.9%]; P < .001) (eTable 4 in the Supplement). Overscreening was generally higher in metropolitan than nonmetropolitan areas; however, the association between metropolitan status and overscreening did not vary by age category (eg, overscreening for CRC in men aged <80 years in metropolitan vs nonmetropolitan areas: aOR, 1.12 [95% CI, 0.91-1.39] vs overscreening for CRC in men aged ≥80 years in metropolitan vs nonmetropolitan areas: aOR, 1.18 [95% CI, 0.92-1.50]; P for interaction = .81).

Overscreening for CRC was higher among participants with high vs low estimated mortality risk (eg, among women, 7714 of 13 538 participants [weighted percentage, 59.7%; 95% CI, 57.7%-61.7%] vs 6749 of 13 274 participants [weighted percentage, 52.6%; 95% CI, 50.4%-54.8%]; P < .001), while overscreening for cervical and breast cancers was lower among participants with high mortality risk vs low mortality risk (8149 of 22 983 participants [weighted percentage, 39.4%; 95% CI, 37.9%-40.9%] vs 23 839 of 51 687 participants [weighted percentage, 48.8%; 95% CI, 47.7%-49.8%]; P < .001 for cervical cancer) (eTable 5 in the Supplement). Overscreening was generally higher in metropolitan than nonmetropolitan areas; however, the association between metropolitan status and overscreening did not vary by predicted mortality risk (eg, overscreening for cervical cancer among women with low estimated 10-year mortality risk in metropolitan vs nonmetropolitan areas; aOR, 1.19 [95% CI, 1.08-1.30] vs women with high estimated 10-year mortality risk; aOR, 1.21 [95% CI, 1.07-1.38]; P for interaction = .61).

The Figure depicts the association between continuous mortality risk scores and overscreening in each subsample. Overscreening for CRC among men was more common as mortality risk increased (ie, as life expectancy decreased) (metropolitan: parameter estimate [est], 0.08; SE, 0.02; P < .001; nonmetropolitan: est, 0.13; SE, 0.02; P < .001); similar findings emerged for overscreening among women (metropolitan: est, 0.05; SE, 0.02; P < .001; nonmetropolitan: est, 0.05; SE, 0.02; P = .01). Mortality risk was negatively associated with overscreening for cervical cancer (metropolitan: est, −0.04; SE, 0.01; P < .001; nonmetropolitan: est, −0.04; SE, 0.01; P < .001) and breast cancer (metropolitan: est, −0.06; SE, 0.01; P < .001; nonmetropolitan: est, −0.05; SE, 0.02; P < .001). None of these effect estimates differed by metropolitan status.

In this observational study using BRFSS data, we examined overscreening for colorectal, cervical, and breast cancers. To our knowledge, this is the first study demonstrating the high prevalence of older adults receiving cancer screenings that are not recommended by USPSTF or supported by randomized clinical trials. This pattern emphasizes the need for additional research to identify risks and benefits of screening in older adults and to determine who may benefit from screening after the recommended upper age limits. In the meantime, interventions to reduce overscreening are needed to improve preventive care, reduce patient harms and health care costs, and increase efficiency of health systems and public health programs.

Overscreening for colorectal, cervical, and breast cancers was higher for women in metropolitan vs nonmetropolitan areas. At least 4 factors may explain these findings. First, clinicians in nonmetropolitan areas may have longer, more trusting relationships with patients,²² which may support conversations about stopping cancer screening. Second, women in nonmetropolitan areas have challenges accessing screening facilities,^23,24 which likely limits overscreening and appropriate screening. Third, clinicians in metropolitan areas are more likely to have technology to remind patients to receive screenings²⁵; patients at these clinics may continue to receive screening reminders after it is no longer appropriate. Fourth, people in metropolitan areas have less fatalistic beliefs about cancer,²⁶ which might make them more receptive to screening. Why these factors do not extend to overscreening for CRC among men is unclear.

Geographic maldistribution of health care resources is among the most persistent characteristics of the US health care system.²⁷ For example, family medicine is the most common specialty in nonmetropolitan areas,²⁸ but women in metropolitan areas are more likely to see an obstetrician or gynecologist for cervical and breast cancer screening.²⁹ These women tend to have higher perceived risk for cervical and breast cancer compared with women seen by family physicians,^30,31 which could influence screening behaviors.

Guidelines for cancer screening must balance risks (eg, emotional anxiety, unnecessary follow-up procedures) with benefits (eg, population mortality benefit).³² Individuals with limited life expectancy can anticipate fewer benefits from cancer screening, particularly for colonoscopy.^1,33-35 We assessed life expectancy using a modified version of a previously validated index^20,21 to estimate 10-year mortality risk. Cruz et al²¹ reported 10-year mortality risk surpassed 50% when participants scored 8 points or more on the index; we detected no meaningful reduction in overscreening when patients passed this threshold. Mortality risk was positively associated with CRC overscreening (although this association was very small), perhaps because patients and clinicians overestimate the benefits of screening.³⁶ In contrast, mortality risk was negatively associated with overscreening for cervical and breast cancers, although this association was very small. End-of-life care and quality of life may be improved by appropriate cessation of cancer screening.

At the population level, we need better data regarding when cancer screening tests become more risky than beneficial for older adults. Based on current scientific understanding, continuing to screen patients who are older and/or have limited life expectancy may cause more harms than benefits. The development of successful interventions to address this problem is thus essential. Previous studies have evaluated electronic health record alerts,³⁷ decision aids,³⁸ and social marketing^36,39-41 to reduce overscreening, with mixed results. Another potential approach is to intervene on patient-clinician discussions about screening.³⁹ Patient-clinician conversations that involve shared decision-making are encouraged before prostate cancer screening^42,43 and lung cancer screening⁴⁴; similar conversations may help reduce overscreening with other types of cancer. In addition, public health campaigns should include statements about the upper age limits for cancer screening.

This study has several strengths. It leveraged a large, national health survey, increasing the representativeness of the findings. In addition, we examined participants’ mortality risk and its association with overscreening, which represents a novel contribution. In terms of study limitations, cancer screening variables are subject to recall bias, which could be especially problematic for CRC screening, which has a long time horizon. We cannot distinguish between cancer screening and surveillance tests. Furthermore, we have no clinical information on participants, including whether they were at above-average risk for cancer, and therefore, screening may still have been appropriate. Also, it is likely that participants in this analysis are healthier than the average US adult older than 65 years, given that institutionalized adults were excluded, and participants had to be healthy enough to complete a lengthy phone survey. As a result of these limitations, the prevalence of overscreening is likely overestimated; future research should validate self-reported screening with health records or claims data to determine the true prevalence of overscreening. Finally, we used current USPSTF recommendations to define overscreening; however, emerging evidence suggests that it may be appropriate to screen women after the current upper age limits for breast cancer^45,46 and cervical cancer.^47,48 It will be important to monitor the prevalence and trends in the demographic characteristics associated with overscreening as recommendations change.

In this study, overscreening for colorectal, cervical, and breast cancers was high among older adults in the US, particularly for women in metropolitan areas. Importantly, these patterns are not strongly associated with predicted life expectancy. The results of this study can be used by public health, primary care, and health care systems to prevent overscreening. Interventions to reduce overscreening for cancer among older adults are needed to improve preventive care and reduce health care burden, costs, and harms for this population.

Accepted for Publication: May 14, 2020.

Published: July 27, 2020. doi:10.1001/jamanetworkopen.2020.11645

Open Access: This is an open access article distributed under the terms of the CC-BY License. © 2020 Moss JL et al. JAMA Network Open.

Corresponding Author: Jennifer L. Moss, PhD, Department of Family and Community Medicine, Department of Public Health Sciences, Penn State College of Medicine, The Pennsylvania State University, 134 Sipe Ave, PO Box 850, MC HS72, #205, Hershey, PA 17033 (jmoss1@pennstatehealth.psu.edu).

Author Contributions: Dr Moss had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Moss, Roy, Lengerich, Adelman, Curry, Ruffin.

Acquisition, analysis, or interpretation of data: Moss, Shen, Cooper, Lennon, Lengerich, Adelman, Curry, Ruffin.

Drafting of the manuscript: Moss, Roy, Shen, Cooper, Lennon, Lengerich, Curry, Ruffin.

Critical revision of the manuscript for important intellectual content: Moss, Roy, Lennon, Lengerich, Adelman, Curry, Ruffin.

Statistical analysis: Moss, Shen, Lengerich, Adelman, Ruffin.

Obtained funding: Ruffin.

Administrative, technical, or material support: Roy, Cooper, Lengerich, Adelman, Ruffin.

Supervision: Lengerich, Adelman, Ruffin.

Conflict of Interest Disclosures: None reported.

Funding/Support: This study was supported by National Cancer Institute grant K22 CA225705 (Dr Moss); by an Institutional Research Grant, number IRG 17-175-04, from the American Cancer Society (Dr Roy); by National Center for Advancing Translational Sciences, National Institutes of Health grant UL1 TR002014 (Dr Lengerich); and by the Hershey Company Professorship and Dr and Mrs Forney George Fellowship (Dr Ruffin).

Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Disclaimer: The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the funding sources.