International cohort study indicates no association between alpha-1 blockers and susceptibility to COVID-19 in benign prostatic hyperplasia patients

Purpose: Alpha-1 blockers, often used to treat benign prostatic hyperplasia (BPH), have been hypothesized to prevent COVID-19 complications by minimising cytokine storm release. The proposed treatment based on this hypothesis currently lacks support from reliable real-world evidence, however. We leverage an international network of large-scale healthcare databases to generate comprehensive evidence in a transparent and reproducible manner. Methods: In this international cohort study, we deployed electronic health records from Spain (SIDIAP) and the United States (Department of Veterans Affairs, Columbia University Irving Medical Center, IQVIA OpenClaims, Optum DOD, Optum EHR). We assessed association between alpha-1 blocker use and risks of three COVID-19 outcomes—diagnosis, hospitalization, and hospitalization requiring intensive services—using a prevalent-user active-comparator design. We estimated hazard ratios using state-of-the-art techniques to minimize potential confounding, including large-scale propensity score matching/stratification and negative control calibration. We pooled database-specific estimates through random effects meta-analysis. Results: Our study overall included 2.6 and 0.46 million users of alpha-1 blockers and of alternative BPH medications. We observed no significant difference in their risks for any of the COVID-19 outcomes, with our meta-analytic HR estimates being 1.02 (95% CI: 0.92–1.13) for diagnosis, 1.00 (95% CI: 0.89–1.13) for hospitalization, and 1.15 (95% CI: 0.71–1.88) for hospitalization requiring intensive services. Conclusion: We found no evidence of the hypothesized reduction in risks of the COVID-19 outcomes from the prevalent-use of alpha-1 blockers—further research is needed to identify effective therapies for this novel disease.

: Kaplan-Meier plots for time until COVID-19 diagnosis in the three databases with lesser statistical power in terms of estimated MDRRs (Optum DOD, Optum EHR, and CUIMC). Shown beneath each plot are the numbers of at-risk patients in each cohort at various points in time. Supplementary Table 1: Baseline patient characteristics for Alpha-1 blocker (T) and 5ARI/PDE5 (C) prevalent-use in the SIDIAP, VA, CUIMC, OpenClaims, Optum DOD, and Optum EHR data sources. We report the proportion of use satisfying selected based-line characteristics and the standardized difference of population proportions (SDf) before and after stratification. Less extreme SDf through stratification suggest improved balance between patient cohorts through propensity score adjustment.

Negative control calibration: estimates before and after
Negative control experiment uses a large number of "true negatives," outcomes to which the exposures have no known relation, to detect potential biases induced by residual confoundings. The experiment also diagnoses potential under-or over-coverage of the null effect by the estimated CIs. These results can in turn be used to calibrate our estimates and confidence intervals for the actual clinical questions of interest to correct for potential biases and miscalibrated uncertainty estimates. In practice, negative control calibration typically results in more conservative conclusions with fewer statistically significant findings.
To demonstrate the process of negative control calibration, we first consider what our analysis would have produced had we not adequately addressed confounding. Supplementary Figure 2(a) below shows the results of negative control experiments in comparing the alpha-1 blocker and 5ARI/PDE5 cohorts if no covariates were to be adjusted. Residual confounding is evident. The uncalibrated estimates show that the alpha-1 blocker users tend to have higher hazard ratios on average across the variety of health outcomes. This could indicate, for example, that alpha-1 blocker users have worse baseline health. The calibration debiases the original estimates, in this case by lowering hazard ratios; graphically, the process manifests itself as the calibrated blue dots being shifted leftward compared to the uncalibrated orange ones. The calibration also widens the confidence intervals, with the blue dots shifted upward compared to the orange dots.
The presence of substantial confounding is no surprise in the toy example above. Even when we control for measured confounders, however, there remains substantial concern for residual confounding. This is where negative control experiments become essential. Supplementary Figure 2(b) shows the results of negative control experiments in comparing the alpha-1 blocker and 5ARI/PDE5 cohorts using our large-scale PS stratification analysis. The negative control experiments here reveal at most minor residual confoundings, with the estimates and CIs demonstrating resonable performance without calibration. Correspondingly, the calibration has rather small effects on the estimates. We also provide numeric values of the uncalibrated estimates in Supplementary Table 3, following the format of Table 3 in the main text.

Supplementary Figure 2(a):
Estimated hazard ratios and standard errors for negative control outcomes when no covariates are adjusted. Orange dots represent original uncalibrated estimates, while blue dots represent calibrated estimates. Shown here are the results for the three larger databases in terms of study cohort sizes (OpenClaims, VA, and Optum DOD).

Supplementary Figure 2(b):
Estimated hazard ratios and standard errors for negative control outcomes when PS stratification is deployed. Compared to the unadjusted case in Figure 1(a), the actual null coverage by the CIs is close to the nominal 95% value even before calibration. This suggests that the stratification by our large-scale PS model has successfully removed most of confoundings. [COVID ID133 V1] Persons with a COVID-19 diagnosis or a SARS-CoV-2 positive test with no required prior observation Initial Event Cohort People having any of the following:

Supplementary
• a measurement of SARS-CoV-2 positive test measurement pre-coordinated 2 ○ occurrence start is after 2019-12-01 • a measurement of SARS-CoV-2 test measurement 3 ○ occurrence start is after 2019-12-01 ○ value as concept is any of: Detected, Detected, Positive, Positive, Present, Present • an observation of SARS-CoV-2 test measurement 3 ○ occurrence start is after 2019-12-01 ○ value as concept is any of: Detected, Detected, Positive, Positive, Present, Present • a condition occurrence of COVID-19 conditions 1 ○ occurrence start is after 2019-12-01 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: earliest event per person.
Limit qualifying cohort to: earliest event per person.

End Date Strategy
Date Offset Exit Criteria This cohort defintion end date will be the index event's start date plus 1 days Cohort Collapse Strategy: Collapse cohort by era with a gap size of 90 days. People having any of the following:

COVID-19 conditions
• a visit occurrence of Inpatient Visit 2 ○ occurrence start is after 2019-12-01 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having any of the following criteria: • at least 1 occurrences of a measurement of SARS-CoV-2 positive test measurement pre-coordinated 3 • where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date • or at least 1 occurrences of a measurement of SARS-CoV-2 test measurement 4 ○ value as concept is any of: Detected, Detected, Positive, Positive, Present, Present • where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date • or at least 1 occurrences of an observation of SARS-CoV-2 test measurement 4 ○ value as concept is any of: Detected, Detected, Positive, Positive, Present, Present • where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date • or at least 1 occurrences of a condition occurrence of COVID-19 conditions 1 • where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date ○ occurrence start is after 2019-12-01 • a procedure of tracheostomy 7 ○ occurrence start is after 2019-12-01 • a procedure of Extracorporeal membrane oxygenation (ECMO) procedure 2 ○ occurrence start is after 2019-12-01 with continuous observation of at least 0 days prior and 0 days after event index date, and limit initial events to: all events per person.
For people matching the Primary Events, include: Having all of the following criteria: • at least 1 occurrences of a visit occurrence of Inpatient Visit 3 ○ Having any of the following criteria: ■ at least 1 occurrences of a measurement of SARS-CoV-2 positive test measurement pre-coordinated 5 ■ where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date ■ or at least 1 occurrences of a measurement of SARS-CoV-2 test measurement 6 ■ value as concept is any of: Detected, Detected, Positive, Positive, Present, Present ■ where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date ■ or at least 1 occurrences of an observation of SARS-CoV-2 test measurement 6 ■ value as concept is any of: Detected, Detected, Positive, Positive, Present, Present ■ where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date ■ or at least 1 occurrences of a condition occurrence of COVID-19 conditions 1 ■ where event starts between 21 days Before and all days After index start date and event starts between all days Before and 0 days After index end date • where event starts between all days Before and 0 days After index start date and event ends between 0 days Before and all days After index start date

Covariates included in large-scale PS model
There are substantial variations in the range, resolution, and coding of clinical covariates captured in different healthcare databases. This makes it difficult to define a set of covariates to control for that is consistent across databases. OHDSI's Common Data Model (CDM) nonetheless allows us to define a consistent criteria broad enough to include most potential confounders, from which we can select a relevant subset in a data-driven manner using a sparse regression technique. The table below describes our selection criteria in more detail.