Publications

Prostate cancer (PCa) incidence is reportedly lower in men with HIV compared to men without HIV for unknown reasons. We describe PCa incidence by HIV status in Medicaid beneficiaries, allowing for comparison of men with and without HIV who are similar with respect to socioeconomic characteristics and access to healthcare. Men (N = 15,167,636) aged 18-64 with ≥7 months of continuous enrollment during 2001-2015 in 14 US states were retained for analysis. Diagnoses of HIV and PCa were identified using non-drug claims. We estimated cause-specific (csHR) comparing incidence of PCa by HIV status, adjusted for age, race-ethnicity, state of residence, year of enrollment, and comorbid conditions, and stratified by age and race-ethnicity. Hazard of PCa was lower in men with HIV than men without HIV (csHR = 0.89; 95% CI: 0.80, 0.99), but varied by race-ethnicity, with similar observations among non-Hispanic Black (csHR = 0.79; 95% CI: 0.69, 0.91) and Hispanic (csHR = 0.85; 95% CI: 0.67, 1.09), but not non-Hispanic white men (csHR = 1.17; 95% CI: 0.91, 1.50). Findings were similar in models restricted to men aged 50-64 and 40-49, but not in men aged 18-39. Reported deficits in PCa incidence by HIV status may be restricted to specific groups defined by age and race ethnicity.

Disparities in HIV care by socioeconomic status, place of residence, and race/ethnicity prevent progress toward epidemic control. No study has comprehensively characterized the HIV care cascade among people with HIV enrolled in Medicaid - an insurance source for low-income individuals in the US. We analyzed data from 246,127 people with HIV enrolled in Medicaid 2001-2015, aged 18-64, living in 14 US states. We estimated the monthly prevalence of four steps of the care cascade: retained in care/adherent to ART; retained/not adherent; not retained/adherent; not retained/not adherent. Beneficiaries were retained in care if they had an outpatient care encounter every six months. Adherence was based on medication possession ratio. We estimated prevalence using a non-parametric multi-state approach, accounting for death as a competing event and for Medicaid disenrollment using inverse probability of censoring weights. Across 2001-2015, the proportion of beneficiaries with HIV who were retained/ART adherent increased, overall and in all subgroups. By 2015, approximately half of beneficiaries were retained in care, and 42% of beneficiaries were ART adherent. We saw meaningful differences by race/ethnicity and region. Our work highlights an important disparity in the HIV care cascade by insurance status during this time period.

Prostate cancer is projected to account for the greatest proportion of cancer-related burden among men with HIV. However, incidence is reportedly lower than in men without HIV, potentially due to differences in screening. Factors influencing receipt of screening in men with HIV are unknown. We described receipt of prostate-specific antigen (PSA) testing and assessed factors for association with receipt of PSA test. Demographics, measures of HIV and related care, and non-HIV care were assessed for association with receipt of first PSA test in men ≥40 years old each calendar year in 2000-2020 using univariable and multivariable Poisson regression. Models were additionally stratified by calendar period to identify changes in determinants of PSA test as prostate cancer screening guidelines changed. Men (n = 2,063) 72% Non-Hispanic Black, median age of 47 (IQR: 41, 53), contributed median of 4.7 years (IQR: 2.3, 10.0) of follow-up. Receipt of antiretroviral therapy (aIRR = 1.33; 95% CI: 1.14, 1.55), engagement in HIV care (aIRR = 2.09; 95% CI: 1.66, 2.62), history of testosterone-replacement therapy (aIRR = 1.34; 95% CI: 1.19, 1.50), urologist evaluation (aIRR = 1.66; 95% CI: 1.35, 2.05), and receipt of PSA test in preceding two years (no elevated PSA aIRR = 2.37; 95% CI: 2.16, 2.61; elevated PSA aIRR = 4.35; 95% CI: 3.24, 5.84) were associated with PSA testing in men aged 50 or older. Associations varied across calendar time. Findings suggest men with greater interaction with healthcare are more likely to receive PSA test. Measures of control of HIV did not appear to influence the decision to screen.

Studies have reported lower incidence of prostate cancer in men living with HIV compared with men without HIV for reasons that remain unclear. Lower prostate cancer screening in men living with HIV could explain these findings. We describe receipt of prostate-specific antigen (PSA) test each calendar year by HIV status in Medicaid beneficiaries enrolled in 14 U.S. states, 2001-2015. A total of 15,299,991 Medicaid beneficiaries aged 18-64 with ≥7 months of continuous enrollment were included in analyses. HIV diagnosis and PSA tests were identified using non-drug claims. Incidence rate ratios comparing receipt of PSA test by HIV status adjusted for age, race/ethnicity, state of residence, calendar year, comorbid conditions, benign prostatic conditions, and receipt of testosterone-replacement therapy were estimated using Poisson regression. Models were also stratified by state, and estimates were pooled using random-effects meta-analysis to account for heterogeneity by state. Models were additionally stratified by age and race/ethnicity. There were 42,503 PSA tests over 314,273 person-years and 1,669,835 PSA tests over 22,023,530 person-years observed in beneficiaries with and without HIV, respectively. The incidence of PSA test was slightly lower in men living with HIV than men without HIV (incidence rate ratio [IRR] = 0.98; 95% confidence interval [CI]: 0.97, 0.99) when adjusting for state. In the pooled estimate, the rate was higher among men living with HIV (IRR = 1.11; 95% CI: 0.97, 1.27). Pooled estimates indicated approximately equal or higher rates of PSA test in men living with HIV compared with men without HIV across models stratified by age and race/ethnicity groups. Findings do not support the hypothesis that differences in prostate cancer screening explain differences in incidence by HIV status.

People with human immunodeficiency virus (HIV; PWH) in the United States have a lower incidence of colon cancer than the general population. The lower incidence may be explained by differences in receipt of screening. Thus, we sought to estimate colon cancer incidence under scenarios in which Medicaid beneficiaries, with or without HIV, followed the same screening protocols. We used data from 1.5 million Medicaid beneficiaries who were enrolled in 14 US states in 2001-2015 and aged 50-64 years; 72 747 beneficiaries had HIV. We estimated risks of colon cancer and death by age, censoring beneficiaries when they deviated from 3 screening protocols, which were based on Medicaid's coverage policy for endoscopies during the time period, with endoscopy once every 2, 4, or 10 years. We used inverse probability weights to control for baseline and time-varying confounding and informative loss to follow-up. Analyses were performed overall, by sex, and by race/ethnicity. PWH had a lower incidence of colon cancer than beneficiaries without HIV. Compared with beneficiaries without HIV, the risk difference at age 65 years was -1.6% lower (95% confidence interval, -2.3% to -.7%) among PWH with the 2-year protocol and -0.8% lower (-1.3% to -.3%) with the 10-year protocol. Results were consistent across subgroup and sensitivity analyses. Our findings suggest that the lower risk of colon cancer that has been observed among PWH aged 50-64 years compared with those without HIV is not due to differences in receipt of lower endoscopy. Keywords: colon cancer, colorectal cancer screening, endoscopy, Medicaid, human immunodeficiency virus.

10-year outcomes in patients living with HIV who are diagnosed with prostate cancer are unknown. 52 patients living with HIV were diagnosed with prostate cancer. Disease-free survival stratified by clinical, pathologic, and HIV characteristics were examined. No difference in disease-free survival was observed based on prostate cancer treatment modality, CD4 count, or HIV viral load. Prostate cancer outcomes in patients living with HIV are favorable irrespective of treatment modality.

Studies suggest a lower colorectal cancer (CRC) risk and lower or similar CRC screening among people with HIV (PWH) compared with the general population. We evaluated the incidence of lower endoscopy and average-onset (diagnosed at ≥50) and early-onset (diagnosed at <50) colon cancer by HIV status among Medicaid beneficiaries with comparable sociodemographic factors and access to care. We obtained Medicaid Analytic eXtract (MAX) data from 2001 to 2015 for 14 states. We included 41 727 243 and 42 062 552 unique individuals with at least 7 months of continuous eligibility for the endoscopy and colon cancer analysis, respectively. HIV and colon cancer diagnoses and endoscopy procedures were identified from inpatient and other nondrug claims. We used Cox proportional hazards regression models to assess endoscopy and colon cancer incidence, controlling for age, sex, race/ethnicity, calendar year and state of enrollment, and comorbidities conditions. Endoscopy and colon cancer incidence increased with age in both groups. Compared with beneficiaries without HIV, PWH had an increased hazard of endoscopy; this association was strongest among those 18-39 years [hazard ratio: 1.85, 95% confidence interval (95% CI) 1.77-1.92] and attenuated with age. PWH 18-39 years also had increased hazard of early-onset colon cancer (hazard ratio: 1.66, 95% CI:1.05-2.62); this association was attenuated after comorbidity adjustment. Hazard ratios were null among all beneficiaries less than 50 years of age. PWH had a lower hazard of average-onset colon cancer compared with those without HIV (hazard ratio: 0.79, 95% CI: 0.66-0.94). PWH had a higher hazard of endoscopy, particularly at younger ages. PWH had a lower hazard of average-onset colon cancer. Early-onset colon cancer was higher among the youngest PWH but not associated with HIV overall.

Life expectancy among people with HIV (PWH) is increasing, making chronic conditions-including cancer-increasingly relevant. Among PWH, cancer burden has shifted from AIDS-defining cancers (ADCs) toward non-AIDS-defining cancers (NADCs). We described incidence of cancer in a claims-based cohort of Medicaid beneficiaries. We included 43,426,043 Medicaid beneficiaries (180,058 with HIV) from 14 US states, aged 18-64, with >6 months of enrollment (with no dual enrollment in another insurance) and no evidence of a previous cancer. We estimated cumulative incidence of site-specific cancers, NADCs, and ADCs, by baseline HIV status, using age as the time scale and accounting for death as a competing risk. We compared cumulative incidence across HIV status to estimate risk differences. We examined cancer incidence overall and by sex, race/ethnicity, and calendar period. PWH had a higher incidence of ADCs, infection-related NADCs, and death. For NADCs such as breast, prostate, and colon cancer, incidence was similar or higher among PWH below age 50, but higher among those without HIV by age 65. Incidence of lung and head and neck cancer was always higher for female beneficiaries with HIV, whereas the curves crossed for male beneficiaries. We saw only small differences in incidence trends by race/ethnicity. Our findings suggest an increased risk of certain NADCs at younger ages among PWH, even when compared against other Medicaid beneficiaries, and highlight the importance of monitoring PWH for ADCs and NADCs. Future work should explore possible mechanisms explaining the differences in incidence for specific cancer types.

Medical and population health science researchers frequently make ambiguous statements about whether they believe their study sample or results are representative of some (implicit or explicit) target population. This article provides a comprehensive definition of representativeness, with the goal of capturing the different ways in which a study can be representative of a target population. It is proposed that a study is representative if the estimate obtained in the study sample is generalisable to the target population (owing to representative sampling, estimation of stratum specific effects, or quantitative methods to generalise or transport estimates) or the interpretation of the results is generalisable to the target population (based on fundamental scientific premises and substantive background knowledge). This definition is explored in the context of four covid-19 studies, ranging from laboratory science to descriptive epidemiology. All statements regarding representativeness should make clear the way in which the study results generalise, the target population the results are being generalised to, and the assumptions that must hold for that generalisation to be scientifically or statistically justifiable.

Inverse probability weighting (IPW) and g-computation are commonly used in time-varying analyses. To inform decisions on which to use, we compared these methods using a plasmode simulation based on data from the Effects of Aspirin in Gestation and Reproduction (EAGeR) Trial (June 15, 2007-July 15, 2011). In our main analysis, we simulated a cohort study of 1,226 individuals followed for up to 10 weeks. The exposure was weekly exercise, and the outcome was time to pregnancy. We controlled for 6 confounding factors: 4 baseline confounders (race, ever smoking, age, and body mass index) and 2 time-varying confounders (compliance with assigned treatment and nausea). We sought to estimate the average causal risk difference by 10 weeks, using IPW and g-computation implemented using a Monte Carlo estimator and iterated conditional expectations (ICE). Across 500 simulations, we compared the bias, empirical standard error (ESE), average standard error, standard error ratio, and 95% confidence interval coverage of each approach. IPW (bias = 0.02; ESE = 0.04; coverage = 92.6%) and Monte Carlo g-computation (bias = -0.01; ESE = 0.03; coverage = 94.2%) performed similarly. ICE g-computation was the least biased but least precise estimator (bias = 0.01; ESE = 0.06; coverage = 93.4%). When choosing an estimator, one should consider factors like the research question, the prevalences of the exposure and outcome, and the number of time points being analyzed.

In many research settings, the intervention implied by the average causal effect of a time-varying exposure is impractical or unrealistic, and we might instead prefer a more realistic target estimand. Instead of requiring all individuals to be always exposed versus unexposed, incremental effects quantify the impact of merely shifting each individual's probability of being exposed. We demonstrate the estimation of incremental effects in the time-varying setting, using data from the Effects of Aspirin in Gestation and Reproduction trial, which assessed the effect of preconception low-dose aspirin on pregnancy outcomes. Compliance to aspirin or placebo was summarized weekly and was affected by time-varying confounders such as bleeding or nausea. We sought to estimate what the incidence of pregnancy by 26 weeks postrandomization would have been if we shifted each participant's probability of taking aspirin or placebo each week by odds ratios (OR) between 0.30 and 3.00. Under no intervention (OR = 1), the incidence of pregnancy was 77% (95% CI: 74%, 80%). Decreasing women's probability of complying with aspirin had little estimated effect on pregnancy incidence. When we increased women's probability of taking aspirin, estimated incidence of pregnancy increased, from 83% (95% confidence interval [CI] = 79%, 87%) for OR = 2 to 89% (95% CI = 84%, 93%) for OR=3. We observed similar results when we shifted women's probability of complying with a placebo. These results estimated that realistic interventions to increase women's probability of taking aspirin would have yielded little to no impact on the incidence of pregnancy, relative to similar interventions on placebo.

Integrative analysis of datasets generated by multiple cohorts is a widely-used approach for increasing sample size, precision of population estimators, and generalizability of analysis results in epidemiological studies. However, often each individual cohort dataset does not have all variables of interest for an integrative analysis collected as a part of an original study. Such cohort-level missingness poses methodological challenges to the integrative analysis since missing variables have traditionally: (1) been removed from the data for complete case analysis; or (2) been completed by missing data interpolation techniques using data with the same covariate distribution from other studies. In most integrative-analysis studies, neither approach is optimal as it leads to either loosing the majority of study covariates or challenges in specifying the cohorts following the same distributions. We propose a novel approach to identify the studies with same distributions that could be used for completing the cohort-level missing information. Our methodology relies on (1) identifying sub-groups of cohorts with similar covariate distributions using cohort identity random forest prediction models followed by clustering; and then (2) applying a recursive pairwise distribution test for high dimensional data to these sub-groups. Extensive simulation studies show that cohorts with the same distribution are correctly grouped together in almost all simulation settings. Our methods' application to two ECHO-wide Cohort Studies reveals that the cohorts grouped together reflect the similarities in study design. The methods are implemented in R software package relate.

There are important challenges to the estimation and identification of average causal effects in longitudinal data with time-varying exposures. Here, we discuss the difficulty in meeting the positivity condition. Our motivating example is the per-protocol analysis of the Effects of Aspirin in Gestation and Reproduction (EAGeR) Trial. We estimated the average causal effect comparing the incidence of pregnancy by 26 weeks that would have occurred if all women had been assigned to aspirin and complied versus the incidence if all women had been assigned to placebo and complied. Using flexible targeted minimum loss-based estimation, we estimated a risk difference of 1.27% (95% CI: -9.83, 12.38). Using a less flexible inverse probability weighting approach, the risk difference was 5.77% (95% CI: -1.13, 13.05). However, the cumulative probability of compliance conditional on covariates approached 0 as follow-up accrued, indicating a practical violation of the positivity assumption, which limited our ability to make causal interpretations. The effects of nonpositivity were more apparent when using a more flexible estimator, as indicated by the greater imprecision. When faced with nonpositivity, one can use a flexible approach and be transparent about the uncertainty, use a parametric approach and smooth over gaps in the data, or target a different estimand that will be less vulnerable to positivity violations.

Cohort collaborations often require meta-analysis of exposure-outcome association estimates across cohorts as an alternative to pooling individual-level data that requires a laborious process of data harmonization on individual-level data. However, it is likely that important confounders are not all measured uniformly across the cohorts due to differences in study protocols. This imbalance in measurement of confounders leads to association estimates that are not comparable across cohorts and impedes the meta-analysis of results. In this article, we empirically show some asymptotic relations between fully adjusted and unadjusted exposure-outcome effect estimates, and provide theoretical justification for the same. We leverage these results to obtain fully adjusted estimates for the cohorts with no information on confounders by borrowing information from cohorts with complete measurement on confounders. We implement this novel method in CIMBAL (confounder imbalance), which additionally provides a meta-analyzed estimate that appropriately accounts for the dependence between estimates arising due to borrowing of information across cohorts. We perform extensive simulation experiments to study CIMBAL's statistical properties. We illustrate CIMBAL using National Children's Study (NCS) data to estimate association of maternal education and low birth weight in infants, adjusting for maternal age at delivery, race/ethnicity, marital status, and income. Our simulation studies indicate that estimates of exposure-outcome association from CIMBAL are closer to the truth than those from commonly-used approaches for meta-analyzing cohorts with disparate confounder measurements. CIMBAL is not too sensitive to heterogeneity in underlying joint distributions of exposure, outcome and confounders but is very sensitive to heterogeneity of confounding bias across cohorts. Application of CIMBAL to NCS data for a proof-of-concept analysis further illustrates the utility and advantages of CIMBAL. CIMBAL provides a practical approach for meta-analyzing cohorts with imbalance in measurement of confounders under a weak assumption that the cohorts are independently sampled from populations with the same confounding bias.

The incidence rate of breast cancer has been increasing over time across race/ethnicity in the United States. It is unclear whether these trends differ among stage, poverty, and geography subgroups. Using data from the North American Association of Central Cancer Registries, this study estimated trends in age-adjusted breast cancer incidence rates among women aged 50 to 84 years from 1999 to 2017 by race/ethnicity (non-Hispanic Black, non-Hispanic White, and Hispanic) and across subgroups (stage, county-level poverty, county urban/rural status, and geographic region [West, Midwest, South, and Northeast]). From 2004 to 2017, breast cancer incidence rates increased across race/ethnicity and subgroups, with the greatest average annual percent increases observed for non-Hispanic Black women, overall (0.9%) and those living in lower poverty areas (0.8%), rural areas (1.2%), and all regions except the West (0.8%-1.0%). Stronger increases among non-Hispanic Black women were observed for local-stage disease and for some subgroups of distant-stage disease. Non-Hispanic Black women had the smallest decrease in regional-stage disease across most subgroups. Similarly, Hispanic women had the strongest increases in some subgroups, including areas with higher poverty (0.6%-1.2%) and in the West (0.8%), for local- and distant-stage disease. These trends highlight concerns for an increasing burden of breast cancer among subpopulations, with some already experiencing disparate breast cancer mortality rates, and they highlight the need for targeted breast cancer prevention and efforts to reduce mortality disparities in areas with increasing incidence.

Characterizing estradiol among women with HIV may have implications for breast cancer and cardiovascular disease risk but has not been adequately explored. We quantified differences in total (E2), free (FE2) estradiol, and sex hormone binding globulin (SHBG) by HIV and viral suppression status. Women from a substudy (2003-2006) within the Women's Interagency HIV Study (IRB approved at each participating site) were included if they reported: a period in the last six months, were not pregnant/breastfeeding, no oophorectomy, and no exogenous hormone use in the prior year. Serum was collected on days 2-4 of the menstrual cycle. We assessed differences in biomarkers at 25th, 50th, and 75th percentiles by HIV and viral suppression status using weighted quantile regression. Among 643 women (68% with HIV) median age was 37 years. All E2 percentiles were significantly (p < 0.05) lower in women with suppressed viral load versus women without HIV (4-10 pg/mL). The 25th and 50th percentile of E2 were 4-5 pg/mL lower in women with unsuppressed viral load compared to women without HIV (p < 0.05). The 25th and 50th percentile of SHBG was significantly higher in women with unsuppressed viral load compared to women without HIV (10 and 12 nmol/L, respectively). There were no consistent differences in estradiol or SHBG by suppression status. There were no differences in FE2 but significantly lower E2 and higher SHBG among women with HIV versus without HIV. Further research is merited in a large contemporary sample to clarify the clinical implications of these findings.

Measurement error, although ubiquitous, is uncommonly acknowledged and rarely assessed or corrected in epidemiologic studies. This review offers a straightforward guide to common problems caused by measurement error in research studies and a review of several accessible bias-correction methods for epidemiologists and data analysts. Although most correction methods require criterion validation including a gold standard, there are also ways to evaluate the impact of measurement error and potentially correct for it without such data. Technical difficulty ranges from simple algebra to more complex algorithms that require expertise, fine tuning, and computational power. However, at all skill levels, software packages and methods are available and can be used to understand the threat to inferences that arises from imperfect measurements.

Studies suggest lower risk of breast cancer in women with HIV versus without HIV. These estimates may be biased by lower life expectancy and younger age distribution of women with HIV. Our analysis evaluated this bias and characterized secular trends in breast cancer among women with HIV initiating antiretroviral therapy. We hypothesized breast cancer risk would increase over time as mortality decreased. Women with HIV prescribed antiretroviral therapy in the North American AIDS Cohort Collaboration on Research and Design (NA-ACCORD) from 1997 through 2016. We estimated breast cancer hazard (cause-specific hazard ratios) and cumulative incidence accounting for competing risks (subdistribution hazard ratios) to assess changes in breast cancer risk over time. This was assessed overall (1997-2016) and within/across calendar periods. Analyses were adjusted for race/ethnicity and inverse probability weighted for cohort. Cumulative incidence was graphically assessed by calendar period and race/ethnicity. We observed 11,587 women during 1997-2016, contributing 63 incident breast cancer diagnoses and 1,353 deaths [73,445 person-years (median follow-up = 4.5 years)]. Breast cancer cumulative incidence was 3.2% for 1997-2016. We observed no secular trends in breast cancer hazard or cumulative incidence. There were annual declines in the hazard and cumulative incidence of death (cause-specific hazard ratios and subdistribution hazard ratios: 0.89, 95% confidence interval: 0.87 to 0.91) which remained within and across calendar periods. These findings contradict the hypothesis of increasing breast cancer risk with declining mortality over time among women with HIV, suggesting limited impact of changing mortality on breast cancer risk. Additional inquiry is merited as survival improves among women with HIV.