Timothy L Lash’s research while affiliated with Emory University and other places

Background Improvements in breast cancer therapy since the randomized controlled trials of mammography screening might have reduced the screening benefit. Most observational studies of mammography effectiveness would be confounded by these improvements and other factors. Using a design resistant to this confounding, we evaluated whether mammography in asymptomatic women reduces breast cancer mortality during the treatment era succeeding the trials. Methods We designed a quasi-experimental cohort study in regions of Denmark without organized screening. We predicted the number of expected mammograms for each general practice based on observed numbers of mammograms and individual risk factors. Regardless of a woman’s individual exposure to mammography, we assigned her the ratio of observed to expected mammograms of her general practice as her instrumental variable. We employed this potential instrumental variable as mammography exposure status and followed women from January 1, 2006 until death, emigration, or December 31, 2014, whichever came first. Results We included 169,197 women aged 50–66 from 738 general practices and without previous breast cancer as of January 1, 2006. Women affiliated with a practice referring more women than expected, compared with less, had a lower hazard of breast cancer death (hazard ratio 0.80, 95% confidence interval 0.68, 0.95). Negative control associations were near null, suggesting no confounding bias. Conclusions This quasi-experimental study estimated a continued protective effect of mammography in women where most were presumably asymptomatic. In contrast to conventional observational studies, the use of practice referral ratio as an instrumental variable may avoid bias from uncontrolled confounding.

Importance No large randomized clinical trial has directly compared empagliflozin with dapagliflozin, leaving their comparative effectiveness regarding kidney outcomes unknown. Objective To compare kidney outcomes between initiation of empagliflozin vs dapagliflozin in adults with type 2 diabetes who were receiving antihyperglycemic treatment. Design, Setting, and Participants This target trial emulation used nationwide, population-based routinely collected Danish health care data to compare initiation of empagliflozin vs dapagliflozin in adults with type 2 diabetes who received antihyperglycemic treatment between June 1, 2014, and October 31, 2020. Data were analyzed from October 2023 to August 2024. Persons were followed up until an outcome, emigration, death, 6 years, or December 31, 2021, whichever occurred first. Exposure Initiation of empagliflozin vs dapagliflozin. Main Outcomes and Measures Outcomes included acute kidney injury, incident chronic kidney disease (stages G3 to G5 or stage A2 or A3), and progression of chronic kidney disease (≥40% decrease in estimated glomerular filtration rate from baseline). Risks of kidney outcomes were estimated in intention-to-treat and per-protocol analyses using an Aalen-Johansen estimator that adjusted for 56 potential confounders and considered death as a competing event. Results A total of 32 819 individuals who initiated treatment with empagliflozin and 17 464 with dapagliflozin were included (median [IQR] age, 63 [54-71] years; 18 872 female individuals [37.5%]; median [IQR] estimated glomerular filtration rate, 88 [73-104] mL/min/1.73 m ² ). After weighting, all measured covariates were well balanced between the groups. In intention-to-treat analyses, people who initiated treatment with empagliflozin and dapagliflozin exhibited comparable 6-year risks of acute kidney injury (18.2% vs 18.5%; risk ratio, 0.98; 95% CI, 0.91-1.06), chronic kidney disease stages G3 to G5 (11.8% vs 12.1%; risk ratio, 0.97; 95% CI, 0.89-1.05), chronic kidney disease stage A2 or A3 (14.8% vs 14.3%; risk ratio, 1.04; 95% CI, 0.93-1.15), and progression of chronic kidney disease (5.3% vs 5.7%; risk ratio, 0.94; 95% CI, 0.56-1.58). The primary analyses were supported by corresponding per-protocol analyses. Conclusions and Relevance The results of this cohort study suggest that people with type 2 diabetes who initiated treatment with empagliflozin and dapagliflozin had comparable long-term kidney outcomes.

A compilation of factors over the past decade—including the availability of increasingly large and rich healthcare datasets, advanced technologies to extract unstructured information from health records and digital sources, advancement of principled study design and analytic methods to emulate clinical trials, and frameworks to support transparent study conduct—has ushered in a new era of real‐world evidence (RWE). This review article describes the evolution of the RWE era, including pharmacoepidemiologic methods designed to support causal inferences regarding treatment effects, the role of regulators and other health authorities in establishing distributed real‐world data networks enabling analytics at scale, and the many global guidance documents on principled methods of producing RWE. This article also highlights the growing opportunity for RWE to support decision making by regulators, health technology assessment groups, clinicians, patients, and other stakeholders and provides examples of influential RWE studies. RWE holds promise to address important questions that clinical trials typically do not answer about treatment benefits and risks, and to ultimately impact public health by helping to guide decision making across the healthcare ecosystem.

Despite similar incidence rates, nationwide breast cancer mortality is 40% higher among non-Hispanic Black (NHB) than non-Hispanic White (NHW) women. The racial disparity persists even among women with early-stage disease, prognostically favorable subtypes, and indicators of high socioeconomic status and is not evenly distributed throughout the US. Understanding geographic differences may provide additional insight into the drivers of the disparity. However, current data are geographically limited, based primarily on death certificate information, do not incorporate incidence, and often do not provide estimates or account for areas with small populations or sparse case data. Using a Bayesian framework, we estimated the local racial disparity in 5-year mortality for non-metastatic breast cancer diagnosed during 2005–2013 across counties in Georgia, a racially and geographically diverse state. Overall, during the study period, 5-year breast cancer mortality was 43% higher among NHB than NHW women. The racial disparity varied across Georgia with more pronounced disparity observed in the central and southeast and less pronounced disparity in the southwest. Rurality increased and the proportion of owner-occupied housing decreased the magnitude of the disparity, but only after accounting for other area-level covariates. This approach can help guide decisions and resource allocation at the local level.

Rotavirus vaccine appears to perform sub-optimally in countries with higher rotavirus burden. We hypothesized that differences in the magnitude of rotavirus exposures may bias vaccine efficacy (VE) estimates, so true differences in country-specific rotavirus VE would be exaggerated without accommodating differences in exposure. We estimated VE against any-severity and severe rotavirus gastroenteritis (RVGE) using Poisson regression models fit to pooled individual-level data from Phase II and III monovalent rotavirus vaccine trials conducted between 2000 and 2012. The standard approach model included terms for vaccination, country, and a vaccination-country interaction. Other models used proxies for exposure magnitude like severe RVGE rate or age at severe RVGE instead of country. Country-specific proxies were calculated from placebo group data or extracted from an external meta-analysis. Analyses included 83,592 infants from 23 countries in the Americas, Europe, Africa, and Asia. Using the standard approach, VE against severe RVGE substantially varied (10-100%). Using the severe RVGE rate proxy brought VE from all but two countries between 80% and 86%. Heterogeneity for VE against any-severity RVGE was similarly attenuated. Adjusting for exposure proxies reduced heterogeneity in country-specific rotavirus VE estimates. This phenomenon may extend to other vaccines against partially immunizing pathogens with global disparities in burden.

Context Although the increased burden of mental health problems among patients with classic 46,XX congenital adrenal hyperplasia (CAH) is well-documented, it remains unclear if this comorbidity is attributable to the burden of living with a chronic medical condition or the potential psychosocial/sexual consequences of being born with a difference of sex development (DSD) and its associated clinical management. Objective To compare the prevalence of psychiatric diagnoses among patients with CAH and 3 reference groups: matched males and females from the general population, and females with type 1 diabetes mellitus (T1DM). Methods This was a retrospective cross-sectional study using 3 large integrated health systems. Participants included patients with CAH (n = 115), general population male and female referents (n = 1150 per group), and 66 002 female T1DM referents. Results The prevalence of depression, anxiety, personality disorders, and suicidal ideation was higher among CAH participants than in males and females from the general population, but similar to or lower than in T1DM referents. Patients with CAH were more likely to be diagnosed with neurodevelopmental disorders than both female reference groups, whereas the prevalence of elimination disorders (predominantly enuresis) and gender dysphoria was higher in the CAH cohort than in all reference groups. Conclusion Females with classic CAH experience a higher burden of psychiatric illness, including anxiety and depression, than demographically similar men and women in the general population. The similar psychiatric burden between females with CAH and T1DM suggests that morbidity may be influenced by the challenges associated with managing a chronic condition. Investigating long-term mental health trajectories in this population will require longitudinal studies.

To account for misclassification of dichotomous variables using probabilistic bias analysis, beta distributions are often assigned to bias parameters ( e.g., PPV and NPV) based on data from an internal validation substudy. Due to the small sample size of validation substudies, zero cell frequencies can occur. In these scenarios, it may be helpful to assign prior distributions or apply continuity corrections to the predictive value estimates. We simulated cohort studies of varying sizes, with a binary exposure and outcome and a true risk ratio (RR) = 2.0, as well as internal validation substudies to account for exposure misclassification. We conducted bias adjustment under five approaches assigning prior distributions to the NPV and PPV parameters: (1) conventional method ( i.e. , no prior), (2) uniform prior beta (α = 1, β = 1), (3) Jeffreys prior beta (α = 0.5, β = 0.5), (4) using Jeffreys prior as a continuity correction only when zero cells occurred, and (5) using the uniform prior as a continuity correction only when zero cells occurred. We evaluated performance by measuring coverage probability, bias, and mean squared error. For sparse validation data, methods (2)–(5) all had better coverage and lower MSE than the conventional method, with the uniform prior (2) yielding the best performance. However, little difference between methods was observed when the validation substudy did not contain zero cells. If sparse data are expected in a validation substudy, using a uniform prior for the beta distribution of bias parameters can improve the validity of bias-adjusted measures.

Background A 2013 meta-analysis observed a protective association between overweight BMI (versus normal BMI) and all-cause mortality that was particularly strong in people aged ≥65. Estimates informing this meta-analysis were highly heterogeneous, and critics raised insufficient or inappropriate confounder adjustment in many studies as an explanation for the protective summary association. Using this topic as an example, we demonstrate a novel approach for external adjustment of individual studies for a uniform and sufficient confounder set before meta-analysis. Methods We abstracted summary data on the 33 associations comprising the age ≥65 stratum of the 2013 meta-analysis. Using an external dataset (NHANES III), we derived covariates used in each study’s multivariable model of the overweight–mortality association. We then calculated a bias factor to quantify the direction and magnitude of displacement of the ratio measure of association after changing from the original adjustment set to a sufficient adjustment set. After applying bias factors to adjust original associations, we compared summary results from random effects meta-analyses with and without such adjustment. Results We reproduced the original meta-analysis of overweight–mortality estimates among older participants and found a protective association similar to that reported in 2013 (summary RR=0.88, 95% CI: 0.84, 0.92, I ² =38.4%). After we simulated uniform adjustment of all 33 associations for a minimally sufficient confounder set (age, sex, and smoking status), the meta-analysis showed a similar summary association (summary RR=0.90, 95% CI: 0.86, 0.94), but with reduced heterogeneity (I ² =34.6%). Conclusion Simulated uniform adjustment for a sufficient confounder set may improve rigor and promote consensus in meta-analysis.

Background We evaluated the impact of systematic bias due to loss of heterozygosity (LOH) and incomplete phenotyping in studies examining the relationship between CYP2D6 variants and breast cancer recurrence among women treated with tamoxifen. Methods We performed a systematic review of the literature on tamoxifen, CYP2D6 variants, and breast cancer recurrence. A quantitative bias analysis was performed to adjust for LOH and incomplete phenotyping. Bias-adjusted results were then combined in a meta-analysis. Results Thirty-three studies informed the bias analysis and meta-analysis on CYP2D6 variants and breast cancer recurrence and/or mortality. An unadjusted meta-analysis suggested increased risk of recurrence and/or mortality for poor relative to normal metabolizers [RR = 1.28; 95% simulation interval (SI), 1.04–1.58] with substantial heterogeneity (I2 = 27%; P for heterogeneity = 0.07). Adjusting for LOH and incomplete genotyping resulted in a slight change in the effect estimate and a decrease in heterogeneity (RR = 1.34; 95% SI, 1.10–1.63; I2 = 0%; P for heterogeneity = 0.17). Intermediate metabolizers had a slightly increased risk of recurrence and/or mortality relative to normal metabolizers (RR = 1.15; 95% SI, 1.00–1.34; I2 = 0%; P for heterogeneity = 0.89). Conclusions Adjusting for biases such as LOH and incomplete genotyping reduced observed heterogeneity between studies. Individuals with poor CYP2D6 phenotypes were at increased risk for breast cancer outcomes compared with those with normal phenotypes. Impact Reduction in CYP2D6 activity was associated with an increased risk of breast cancer recurrence and/or mortality, and results underscore the importance of quantitatively adjusting for biases when aggregating study results.