Controlling Time-Dependent Confounding by Health Status and Frailty: Restriction Versus Statistical Adjustment

The State of Use and Utility of Negative Controls in Pharmacoepidemiologic Studies

Uses of real-world data in drug safety and effectiveness studies are often challenged by various sources of bias. We undertook a systematic search of the published literature through September 2020 to evaluate the state of use and utility of negative controls to address bias in pharmacoepidemiologic studies. Two reviewers independently evaluated study eligibility and abstracted data. Our search identified 184 eligible studies for inclusion. Cohort studies (115, 63%) and administrative data (114, 62%) were, respectively, the most common study design and data type used. Most studies used negative control outcomes (91, 50%), and for most studies the target source of bias was unmeasured confounding (93, 51%). We identified 4 utility domains of negative controls: 1) bias detection (149, 81%), 2) bias correction (16, 9%), 3) P-value calibration (8, 4%), and 4) performance assessment of different methods used in drug safety studies (31, 17%). The most popular methodologies used were the 95% confidence interval and P-value calibration. In addition, we identified 2 reference sets with structured steps to check the causality assumption of the negative control. While negative controls are powerful tools in bias detection, we found many studies lacked checking the underlying assumptions. This article is part of a Special Collection on Pharmacoepidemiology.

Outcomes of COVID-19 Infection in People Previously Vaccinated Against Influenza: Population-Based Cohort Study Using Primary Health Care Electronic Records

BACKGROUND

A possible link between influenza immunization and susceptibility to the complications of COVID-19 infection has been previously suggested owing to a boost in the immunity against SARS-CoV-2.

OBJECTIVE

This study aimed to investigate whether individuals with COVID-19 could have benefited from vaccination against influenza. We hypothesized that the immunity resulting from the previous influenza vaccination would boost part of the immunity against SARS-CoV-2.

METHODS

We performed a population-based cohort study including all patients with COVID-19 with registered entries in the primary health care (PHC) electronic records during the first wave of the COVID-19 pandemic (March 1 to June 30, 2020) in Catalonia, Spain. We compared individuals who took an influenza vaccine before being infected with COVID-19, with those who had not taken one. Data were obtained from Information System for Research in Primary Care, capturing PHC information of 5.8 million people from Catalonia. The main outcomes assessed during follow-up were a diagnosis of pneumonia, hospital admission, and mortality.

RESULTS

We included 309,039 individuals with COVID-19 and compared them on the basis of their influenza immunization status, with 114,181 (36.9%) having been vaccinated at least once and 194,858 (63.1%) having never been vaccinated. In total, 21,721 (19%) vaccinated individuals and 11,000 (5.7%) unvaccinated individuals had at least one of their outcomes assessed. Those vaccinated against influenza at any time (odds ratio [OR] 1.14, 95% CI 1.10-1.19), recently (OR 1.13, 95% CI 1.10-1.18), or recurrently (OR 1.10, 95% CI 1.05-1.15) before being infected with COVID-19 had a higher risk of presenting at least one of the outcomes than did unvaccinated individuals. When we excluded people living in long-term care facilities, the results were similar.

CONCLUSIONS

We could not establish a protective role of the immunity conferred by the influenza vaccine on the outcomes of COVID-19 infection, as the risk of COVID-19 complications was higher in vaccinated than in unvaccinated individuals. Our results correspond to the first wave of the COVID-19 pandemic, where more complications and mortalities due to COVID-19 had occurred. Despite that, our study adds more evidence for the analysis of a possible link between the quality of immunity and COVID-19 outcomes, particularly in the PHC setting.

Diabetes acts on mortality in hemodialysis patients predicted by asymmetric dimethylarginine and inflammation

BACKGROUND

The mortality rate of diabetic patients on dialysis is higher than that of non-diabetic patients. Asymmetric dimethylarginine and inflammation are strong predictors of death in hemodialysis. This study aimed to evaluate asymmetric dimethylarginine and C-reactive protein interaction in predicting mortality in hemodialysis according to the presence or absence of diabetes.

METHODS

Asymmetric dimethylarginine and C-reactive protein were measured in 202 patients in maintenance hemodialysis assembled from 2011 to 2012 and followed for four years. Effect modification of C-reactive protein on the relationship between asymmetric dimethylarginine and all-cause mortality was investigated dividing the population into four categories according to the median of asymmetric dimethylarginine and C-reactive protein.

RESULTS

Asymmetric dimethylarginine and C-reactive protein levels were similar between diabetics and non-diabetics. Asymmetric dimethylarginine - median IQR μM - (1.95 1.75-2.54 versus 1.03 0.81-1.55 P=0.000) differed in non-diabetics with or without evolution to death (HR 2379 CI 1.36-3.68 P=0.000) and was similar in diabetics without or with evolution to death. Among non-diabetics, the category with higher asymmetric dimethylarginine and C-reactive protein levels exhibited the highest mortality (69.0% P=0.000). No differences in mortality were seen in diabetics. A joint effect was found between asymmetric dimethylarginine and C-reactive protein, explaining all-cause mortality (HR 15.21 CI 3.50-66.12 P=0.000).

CONCLUSIONS

Asymmetric dimethylarginine is an independent predictor of all-cause mortality in non-diabetic patients in hemodialysis. Other risk factors may overlap asymmetric dimethylarginine in people with diabetes. Inflammation dramatically increases the risk of death associated with high plasma asymmetric dimethylarginine in hemodialysis.

Diabetes acts on mortality in hemodialysis patients predicted by asymmetric dimethylarginine and inflammation

BACKGROUND

The mortality rate of diabetic patients on dialysis is higher than that of non-diabetic patients. Asymmetric dimethylarginine and inflammation are strong predictors of death in hemodialysis. This study aimed to evaluate asymmetric dimethylarginine and C-reactive protein interaction in predicting mortality in hemodialysis according to the presence or absence of diabetes.

METHODS

Asymmetric dimethylarginine and C-reactive protein were measured in 202 patients in maintenance hemodialysis assembled from 2011 to 2012 and followed for four years. Effect modification of C-reactive protein on the relationship between asymmetric dimethylarginine and all-cause mortality was investigated dividing the population into four categories according to the median of asymmetric dimethylarginine and C-reactive protein.

RESULTS

Asymmetric dimethylarginine and C-reactive protein levels were similar between diabetics and non-diabetics. Asymmetric dimethylarginine - median IQR μM - (1.95 1.75-2.54 versus 1.03 0.81-1.55 P=0.000) differed in non-diabetics with or without evolution to death (HR 2379 CI 1.36-3.68 P=0.000) and was similar in diabetics without or with evolution to death. Among non-diabetics, the category with higher asymmetric dimethylarginine and C-reactive protein levels exhibited the highest mortality (69.0% P=0.000). No differences in mortality were seen in diabetics. A joint effect was found between asymmetric dimethylarginine and C-reactive protein, explaining all-cause mortality (HR 15.21 CI 3.50-66.12 P=0.000).

CONCLUSIONS

Asymmetric dimethylarginine is an independent predictor of all-cause mortality in non-diabetic patients in hemodialysis. Other risk factors may overlap asymmetric dimethylarginine in people with diabetes. Inflammation dramatically increases the risk of death associated with high plasma asymmetric dimethylarginine in hemodialysis.

The Role of the Contextual Cohort to Resolve Some Challenges and Limitations of Comparisons in Pharmacoepidemiology

In pharmacoepidemiology, comparison studies can provide a useful estimate of the level of increased or decreased risk of specific events with a medication (through a measure of effect). A key focus of pharmacoepidemiological studies is the safety and effectiveness of medicines in their real-world use, and adequate comparisons of effect estimates are critical. However, consideration of guidelines, pharmacoeconomic assessments, and policies for reimbursement have made comparisons in pharmacoepidemiological studies far more difficult to conduct in recent years. Where certain subject characteristics influence the probability of being exposed to a treatment, this can introduce issues of selection bias and confounding. Methodologies are available to minimise selection bias (through case-only and randomised study designs) and deal with confounding (such as regression modelling or propensity score matching methods), however these each have their own limitations. Where prescribing guidelines are present, conducting comparisons in pharmacoepidemiology produces many challenges and not all of these can be easily overcome. Patient channelling can be more frequent with adherence to clinical guidelines compared with when prescribing decisions by doctors are based predominantly on their clinical judgement. Use of a contextual cohort could be considered as an option to characterise the adoption of new medications into clinical practice and describe the prevalence of clinical characteristics and risk factors in the two cohorts, rather than compare event rates and produce an estimate of effect.

Comparative safety of high-dose versus standard-dose influenza vaccination in patients with end-stage renal disease

BACKGROUND

High-dose influenza vaccine (HDV) is an alternative vaccination strategy in patients with end-stage renal disease (ESRD), though the safety of HDV has not been evaluated in this population. The objective of this study was to estimate the relative occurrence of adverse vaccine reactions in patients with ESRD following vaccination with HDV compared with standard-dose influenza vaccine (SDV).

METHODS

Using data from the United States Renal Data System, we identified patients with ESRD aged ≥ 65 years at influenza vaccination during yearly influenza seasons from 2010 through 2016. Patients were followed after vaccination to observe serious (anaphylaxis, angioedema, seizure, encephalopathy, Guillain-Barré syndrome [GBS], and short-term, all-cause mortality) and milder (urticaria/hives, rash, pain in limb, cellulitis, myalgia/myositis, fever, nausea and vomiting, diarrhea, and syncope) adverse events. Propensity score-weighted hazard ratios (HRs) and 95% confidence intervals (CIs) for HDV versus SDV were estimated with Cox proportional hazards models.

RESULTS

Of 520,876 vaccinations observed (mean age = 74.7 years at vaccination; 63% white race), 7.4% were HDV. For serious events, the weighted HRs were null for seizure, encephalopathy, and mortality and inestimable due to too few cases for anaphylaxis, angioedema, and GBS. For milder vaccine reactions, the weighted HRs demonstrated generally increased risks in the HDV group, including rash (HR = 1.86; 95% CI, 1.34-2.57), diarrhea (HR = 1.26; 95% CI, 1.07-1.50), pain in limb (HR = 1.23; 95% CI, 1.12-1.34), and myalgia/myositis (HR = 1.16; 95% CI, 1.04-1.30).

CONCLUSIONS

The risks of serious adverse events were low and similar between treatment groups; however, HDV recipients had increased risks of several milder adverse events compared with SDV recipients, consistent with clinical trial findings in the general population of older adults. These results add important information to inform the risk-benefit tradeoff of the use of HDV versus SDV in patients with ESRD.

Comparative Effectiveness of High-Dose Versus Standard-Dose Influenza Vaccine Among Patients Receiving Maintenance Hemodialysis

RATIONALE & OBJECTIVE

Studies of patients on maintenance dialysis therapy suggest that standard-dose influenza vaccine (SDV) may not prevent influenza-related outcomes. Little is known about the comparative effectiveness of SDV versus high-dose influenza vaccine (HDV) in this population.

STUDY DESIGN

Cohort study using data from the US Renal Data System.

SETTING & PARTICIPANTS

507,552 adults undergoing in-center maintenance hemodialysis between the 2010 to 2011 and 2014 to 2015 influenza seasons.

EXPOSURES

SDV and HDV.

OUTCOMES

All-cause mortality, hospitalization due to influenza or pneumonia, and influenza-like illness during the influenza season.

ANALYTIC APPROACH

Patients were eligible for inclusion in multiple yearly cohorts; thus, our unit of analysis was the influenza patient-season. To examine the relationship between vaccine dose and effectiveness outcomes, we estimated risk differences and risk ratios using propensity score weighting of Kaplan-Meier functions, accounting for a wide range of patient- and facility-level characteristics. For nonmortality outcomes, we used competing-risk methods to account for the high mortality rate in the dialysis population.

RESULTS

Within 225,215 influenza patient-seasons among adults 65 years and older, 97.4% received SDV and 2.6% received HDV. We observed similar risk estimates for HDV and SDV recipients for mortality (risk difference, -0.08%; 95% CI, -0.85% to 0.80%), hospitalization due to influenza or pneumonia (risk difference, 0.15%; 95% CI, -0.69% to 0.93%), and influenza-like illness (risk difference, 0.00%; 95% CI, -1.50% to 1.08%). Our findings were similar among adults younger than 65 years, as well as within other subgroups defined by influenza season, age group, dialysis vintage, month of influenza vaccination, and vaccine valence.

LIMITATIONS

Residual confounding and outcome misclassification.

CONCLUSIONS

The HDV does not appear to provide additional protection beyond the SDV against all-cause mortality or influenza-related outcomes for adults undergoing hemodialysis. The additional cost and side effects associated with HDV should be considered when offering this vaccine. Future studies of HDV and other influenza vaccine strategies are warranted.

Supervised injection facility use and all-cause mortality among people who inject drugs in Vancouver, Canada: A cohort study

BACKGROUND

People who inject drugs (PWID) experience elevated rates of premature mortality. Although previous studies have demonstrated the role of supervised injection facilities (SIFs) in reducing various harms associated with injection drug use, including accidental overdose death, the possible impact of SIF use on all-cause mortality is unknown. Therefore, we examined the relationship between frequent SIF use and all-cause mortality among PWID in Vancouver, Canada.

METHODS AND FINDINGS

Data were derived from 2 prospective cohort studies of PWID in Vancouver, Canada, between December 2006 and June 2017. Every 6 months, participants completed questionnaires that elicited information regarding sociodemographic characteristics, substance use patterns, social-structural exposures, and use of health services including SIFs. These data were confidentially linked to the provincial vital statistics database to ascertain mortality rates and causes of death. We used multivariable extended Cox regression analyses to estimate the independent association between frequent (i.e., at least weekly) SIF use and all-cause mortality. Of 811 participants, 278 (34.3%) were women, and the median age was 39 years (IQR 33-46) at baseline. In total, 432 (53.3%) participants reported frequent SIF use at baseline, and 379 (46.7%) did not. At baseline, frequent SIF users were on average younger than nonfrequent users, and a higher proportion of frequent SIF users than nonfrequent users were unstably housed, resided in the Downtown Eastside neighbourhood, injected in public, had a recent non-fatal overdose, used prescription opioids at least daily, injected heroin at least daily, injected cocaine at least daily, and injected crystal methamphetamine at least daily. A lower proportion of frequent SIF users than nonfrequent users were HIV positive and enrolled in addiction treatment at baseline. The median duration of follow-up among study participants was 72 months (IQR 24-123). In total, 112 participants (13.8%) died during the study period, yielding a crude mortality rate of 22.7 (95% CI 18.7-27.4) deaths per 1,000 person-years. The median years of potential life lost per death was 34 (IQR 27-42) years. In a time-updated multivariable model, frequent SIF use was inversely associated with risk of all-cause mortality after adjusting for potential confounders, including age, sex, HIV seropositivity, unstable housing, at least daily cocaine injection, public injection, incarceration, enrolment in addiction treatment, and calendar year of interview (adjusted hazard ratio 0.46, 95% CI 0.26-0.80, p = 0.006). The main study limitations are the limited generalizability of findings due to non-random sampling, the potential for reporting biases due to reliance on some self-reported information, and the possibility that residual confounding influenced findings.

CONCLUSIONS

We observed a high burden of premature mortality among a community-recruited cohort of PWID. Frequent SIF use was associated with a lower risk of death, independent of relevant confounders. These findings support efforts to enhance access to SIFs as a strategy to reduce mortality among PWID. Further analyses of individual-level data are needed to determine estimates of, and potential causal pathways underlying, associations between SIF use and specific causes of death.

Restriction of Pharmacoepidemiologic Cohorts to Initiators of Medications in Unrelated Preventive Drug Classes to Reduce Confounding by Frailty in Older Adults

Nonexperimental studies of the effectiveness of seasonal influenza vaccine in older adults have found 40%-60% reductions in all-cause mortality associated with vaccination, potentially due to confounding by frailty. We restricted our cohort to initiators of medications in preventive drug classes (statins, antiglaucoma drugs, and β blockers) as an approach to reducing confounding by frailty by excluding frail older adults who would not initiate use of these drugs. Using a random 20% sample of US Medicare beneficiaries, we framed our study as a series of nonrandomized "trials" comparing vaccinated beneficiaries with unvaccinated beneficiaries who had an outpatient health-care visit during the 5 influenza seasons occurring in 2010-2015. We pooled data across trials and used standardized-mortality-ratio-weighted Cox proportional hazards models to estimate the association between influenza vaccination and all-cause mortality before influenza season, expecting a null association. Weighted hazard ratios among preventive drug initiators were generally closer to the null than those in the nonrestricted cohort. Restriction of the study population to statin initiators with an uncensored approach resulted in a weighted hazard ratio of 1.00 (95% confidence interval: 0.84, 1.19), and several other hazard ratios were above 0.95. Restricting the cohort to initiators of medications in preventive drug classes can reduce confounding by frailty in this setting, but further work is required to determine the most appropriate criteria to use.

Assessing Residual Bias in Estimating Influenza Vaccine Effectiveness: Comparison of High-dose Versus Standard-dose Vaccines

BACKGROUND

Estimating influenza vaccine effectiveness using an unvaccinated comparison group may result in biased effect estimates.

OBJECTIVES

To explore the reduction of confounding bias in an active comparison of high-dose versus standard-dose influenza vaccines, as compared with vaccinated versus unvaccinated comparisons.

METHODS

Using Medicare data from the United States end-stage renal disease program (2009-2013), we compared the risk of all-cause mortality among recipients of high-dose vaccine (HDV) versus standard-dose vaccine (SDV), HDV versus no vaccine, and SDV versus no vaccine. To quantify confounding bias, analyses were restricted to the preinfluenza season, when the protective effect of vaccination should not yet be observed. We estimated the standardized mortality ratio-weighted cumulative incidence functions using Kaplan-Meier methods and calculated risk ratios (RRs) and risk differences between groups.

RESULTS

Among 350,921 eligible patients contributing 825,642 unique patient preinfluenza seasons, 0.8% received HDV, 70.5% received SDV, and 28.7% remained unvaccinated. Comparisons with unvaccinated patients yielded spurious decreases in mortality risk during the preinfluenza period, for HDV versus none [RR, 0.60; 95% confidence interval (CI), 0.51-0.70)] and SDV versus none (RR, 0.72; 95% CI, 0.70-0.75). The effect estimate was attenuated in the HDV versus SDV comparison (RR, 0.89; 95% CI, 0.77-1.03). Estimates on the absolute scale followed a similar pattern.

CONCLUSIONS

The HDV versus SDV comparison yielded less-biased estimates of the all-cause mortality before influenza season compared to those with nonuser comparison groups. Vaccine effectiveness and safety researchers should consider the active comparator design to reduce bias due to differences in underlying health status between vaccinated and unvaccinated individuals.

Approaches to Address Premature Death of Patients When Assessing Patterns of Use of Health Care Services After an Index Event

BACKGROUND

Studies of the use of health care after the onset of disease are important for assessing quality of care, treatment disparities, and guideline compliance. Cohort definition and analysis method are important considerations for the generalizability and validity of study results. We compared different approaches for cohort definition (restriction by survival time vs. comorbidity score) and analysis method [Kaplan-Meier (KM) vs. competing risk] when assessing patterns of guideline adoption in elderly patients.

METHODS

Medicare beneficiaries aged 65-95 years old who had an acute myocardial infarction (AMI) in 2008 were eligible for this study. Beneficiaries with substantial frailty or an AMI in the prior year were excluded. We compared KM with competing risk estimates of guideline adoption during the first year post-AMI.

RESULTS

At 1-year post-AMI, 14.2% [95% confidence interval (CI), 14.0%-14.5%) of beneficiaries overall initiated cardiac rehabilitation when using competing risk analysis and 15.1% (95% CI, 14.8%-15.3%) from the KM analysis. Guideline medication adoption was estimated as 52.3% (95% CI, 52.0%-52.7%) and 53.4% (95% CI, 53.1%-53.8%) for competing risk and KM methods, respectively. Mortality was 17.0% (95%CI, 16.8%-17.3%) at 1 year post-AMI. The difference in cardiac rehabilitation initiation at 1-year post-AMI from the overall population was 0.1%, 1.7%, and 1.9% compared with 30-day survivor, 1-year survivor, and comorbidity-score restricted populations, respectively.

CONCLUSIONS

In this study, the KM method consistently overestimated the competing risk method. Competing risk approaches avoid unrealistic mortality assumptions and lead to interpretations of estimates that are more meaningful.

Statins for primary prevention of cardiovascular events and mortality in old and very old adults with and without type 2 diabetes: retrospective cohort study

OBJECTIVE

To assess whether statin treatment is associated with a reduction in atherosclerotic cardiovascular disease (CVD) and mortality in old and very old adults with and without diabetes.

DESIGN

Retrospective cohort study.

SETTING

Database of the Catalan primary care system (SIDIAP), Spain, 2006-15.

PARTICIPANTS

46 864 people aged 75 years or more without clinically recognised atherosclerotic CVD. Participants were stratified by presence of type 2 diabetes mellitus and as statin non-users or new users.

MAIN OUTCOME MEASURES

Incidences of atherosclerotic CVD and all cause mortality compared using Cox proportional hazards modelling, adjusted by the propensity score of statin treatment. The relation of age with the effect of statins was assessed using both a categorical approach, stratifying the analysis by old (75-84 years) and very old (≥85 years) age groups, and a continuous analysis, using an additive Cox proportional hazard model.

RESULTS

The cohort included 46 864 participants (mean age 77 years; 63% women; median follow-up 5.6 years). In participants without diabetes, the hazard ratios for statin use in 75-84 year olds were 0.94 (95% confidence interval 0.86 to 1.04) for atherosclerotic CVD and 0.98 (0.91 to 1.05) for all cause mortality, and in those aged 85 and older were 0.93 (0.82 to 1.06) and 0.97 (0.90 to 1.05), respectively. In participants with diabetes, the hazard ratio of statin use in 75-84 year olds was 0.76 (0.65 to 0.89) for atherosclerotic CVD and 0.84 (0.75 to 0.94) for all cause mortality, and in those aged 85 and older were 0.82 (0.53 to 1.26) and 1.05 (0.86 to 1.28), respectively. Similarly, effect analysis of age in a continuous scale, using splines, corroborated the lack of beneficial statins effect for atherosclerotic CVD and all cause mortality in participants without diabetes older than 74 years. In participants with diabetes, statins showed a protective effect against atherosclerotic CVD and all cause mortality; this effect was substantially reduced beyond the age of 85 years and disappeared in nonagenarians.

CONCLUSIONS

In participants older than 74 years without type 2 diabetes, statin treatment was not associated with a reduction in atherosclerotic CVD or in all cause mortality, even when the incidence of atherosclerotic CVD was statistically significantly higher than the risk thresholds proposed for statin use. In the presence of diabetes, statin use was statistically significantly associated with reductions in the incidence of atherosclerotic CVD and in all cause mortality. This effect decreased after age 85 years and disappeared in nonagenarians.

Comparing methods to address bias in observational data: statin use and cardiovascular events in a US cohort

Background

The theoretical conditions under which causal estimates can be derived from observational data are challenging to achieve in the real world. Applied examples can help elucidate the practical limitations of methods to estimate randomized-controlled trial effects from observational data.

Methods

We used six methods with varying design and analytic features to compare the 5-year risk of incident myocardial infarction among statin users and non-users, and used non-cardiovascular mortality as a negative control outcome. Design features included restriction to a statin-eligible population and new users only; analytic features included multivariable adjustment and propensity score matching.

Results

We used data from 5294 participants in the Cardiovascular Health Study from 1989 to 2004. For non-cardiovascular mortality, most methods produced protective estimates with confidence intervals that crossed the null. The hazard ratio (HR) was 0.92, 95% confidence interval: 0.58, 1.46 using propensity score matching among eligible new users. For myocardial infarction, all estimates were strongly protective; the propensity score-matched analysis among eligible new users resulted in a HR of 0.55 (0.29, 1.05)-a much stronger association than observed in randomized controlled trials.

Conclusions

In designs that compare active treatment with non-treated participants to evaluate effectiveness, methods to address bias in observational data may be limited in real-world settings by residual bias.

Statins and new-onset atrial fibrillation in a cohort of patients with hypertension. Analysis of electronic health records, 2006-2015

Hypertension is the most prevalent risk factor for new-onset atrial fibrillation (AF). But few studies have addressed the effect of statins on the incidence of this arrhythmia in patients with hypertension. This study aimed to evaluate the effect of statins on new-onset of this arrhythmia in a hypertensive population, accounting for AF risk. Data from the Information System for the Development of Research in Primary Care was used to recruit a retrospective cohort of ≥55-year-old hypertensive individuals with no ischemic vascular disease, in 2006-2007, followed up through 2015. The effect of initiating statin treatment on new-onset atrial fibrillation was assessed with Cox proportional hazards models adjusted by the propensity score of receiving statin treatment, in the overall study population and stratified by AF risk. Of 100 276 included participants, 9814 initiated statin treatment. The AF incidence per 1000 person-years (95% confidence interval) was 12.5 (12.3-12.8). Statin use associated with a significant (9%) reduction in AF incidence. Differences in absolute AF incidence were higher in the highest AF risk subgroup, and the estimated number needed to treat to avoid one case was 720. The relative effect was poor, similar across groups, and non-significant, as was the association of statins with adverse effects. We found a limited protective effect of statins over new-onset AF in this hypertensive population with no ischemic vascular disease. If there is no further indication, hypertensive patients would not benefit from statin use solely for AF primary prevention.

Controlling confounding by frailty when estimating influenza vaccine effectiveness using predictors of dependency in activities of daily living

PURPOSE

To improve control of confounding by frailty when estimating the effect of influenza vaccination on all-cause mortality by controlling for a published set of claims-based predictors of dependency in activities of daily living (ADL).

METHODS

Using Medicare claims data, a cohort of beneficiaries >65 years of age was followed from September 1, 2007, to April 12, 2008, with covariates assessed in the 6 months before follow-up. We estimated Cox proportional hazards models of all-cause mortality, with influenza vaccination as a time-varying exposure. We controlled for common demographics, comorbidities, and health care utilization variables and then added 20 ADL dependency predictors. To gauge residual confounding, we estimated pre-influenza season hazard ratios (HRs) between September 1, 2007 and January 5, 2008, which should be 1.0 in the absence of bias.

RESULTS

A cohort of 2 235 140 beneficiaries was created, with a median follow-up of 224 days. Overall, 52% were vaccinated and 4% died during follow-up. During the pre-influenza season period, controlling for demographics, comorbidities, and health care use resulted in a HR of 0.66 (0.64, 0.67). Adding the ADL dependency predictors moved the HR to 0.68 (0.67, 0.70). Controlling for demographics and ADL dependency predictors alone resulted in a HR of 0.68 (0.66, 0.70).

CONCLUSIONS

Results were consistent with those in the literature, with significant uncontrolled confounding after adjustment for demographics, comorbidities, and health care use. Adding ADL dependency predictors moved HRs slightly closer to the null. Of the comorbidities, health care use variables, and ADL dependency predictors, the last set reduced confounding most. However, substantial uncontrolled confounding remained.

Comparative Effectiveness of High-Dose Versus Standard-Dose Influenza Vaccines Among US Medicare Beneficiaries in Preventing Postinfluenza Deaths During 2012-2013 and 2013-2014

Background

Recipients of high-dose vs standard-dose influenza vaccines have fewer influenza illnesses. We evaluated the comparative effectiveness of high-dose vaccine in preventing postinfluenza deaths during 2012-2013 and 2013-2014, when influenza viruses and vaccines were similar.

Methods

We identified Medicare beneficiaries aged ≥65 years who received high-dose or standard-dose vaccines in community-located pharmacies offering both vaccines. The primary outcome was death in the 30 days following an inpatient or emergency department encounter listing an influenza International of Classification of Diseases, Ninth Revision, Clinical Modification code. Effectiveness was estimated by using multivariate Poisson regression models; effectiveness was allowed to vary by season.

Results

We studied 1039645 recipients of high-dose and 1683264 recipients of standard-dose vaccines during 2012-2013, and 1508176 high-dose and 1877327 standard-dose recipients during 2013-2014. Vaccinees were well-balanced for medical conditions and indicators of frail health. Rates of postinfluenza death were 0.028 and 0.038/10000 person-weeks in high-dose and standard-dose recipients, respectively. Comparative effectiveness was 24.0% (95% confidence interval [CI], .6%-42%); there was evidence of variation by season (P = .12). In 2012-2013, high-dose was 36.4% (95% CI, 9.0%-56%) more effective in reducing mortality; in 2013-2014, it was 2.5% (95% CI, -47% to 35%).

Conclusions

High-dose vaccine was significantly more effective in preventing postinfluenza deaths in 2012-2013, when A(H3N2) circulation was common, but not in 2013-2014.

Adverse Events Associated with Antipsychotic Use in Hospitalized Older Adults After Cardiac Surgery

OBJECTIVES

To evaluate in-hospital adverse events associated with typical and atypical antipsychotic medications (APMs) after cardiac surgery.

DESIGN

Retrospective cohort study.

SETTING

Nationwide inpatient database, 2003 to 14.

PARTICIPANTS

Individuals (mean age 70) newly treated with oral atypical (n = 2,580) or typical (n = 1,126 APMs) after coronary artery bypass grafting or valve surgery (N = 3,706).

MEASUREMENTS

In-hospital mortality, arrhythmia, pneumonia, use of brain imaging (surrogate for oversedation and neurological events), and length of stay after drug initiation RESULTS: In the propensity score-matched cohort, median treatment duration was 3 days (interquartile range (IQR) 1-6 days) for atypical APMs and 2 days (IQR 1-3 days) for typical APMs. There were no large differences in in-hospital mortality (atypical 5.4%, typical 5.3%; risk difference (RD) = 0.1%, 95% confidence interval (CI) = -2.1 to 2.3%), arrhythmia (2.0% vs 2.2%; RD = 0.0%; 95% CI = -1.4 to 1.4%), pneumonia (16.1% vs 14.5%; RD = 1.6%, 95% CI = -1.9 to 5.0%), and length of stay (9.9 days vs 9.3 days; mean difference = 0.5 days, 95% CI = -1.2 to 2.2). Use of brain imaging was more common after initiating atypical APMs (17.3%) than after typical APMs (12.4%; RD = 4.9%, 95% CI = 1.4-8.4).

CONCLUSION

In hospitalized individuals who underwent cardiac surgery, short-term use of typical APMs was associated with risks of adverse events similar to those with atypical APMs. Moreover, greater use of brain imaging associated with atypical APMs suggests that these drugs may cause oversedation or adverse neurological events. Because of the low event rates, the analysis could not exclude modest differences in adverse events between atypical and typical APMs.

Use and Interpretation of Propensity Scores in Aging Research: A Guide for Clinical Researchers

Observational studies are an important source of evidence for evaluating treatment benefits and harms in older adults, but lack of comparability in the outcome risk factors between the treatment groups leads to confounding. Propensity score (PS) analysis is widely used in aging research to reduce confounding. Understanding the assumptions and pitfalls of common PS analysis methods is fundamental to applying and interpreting PS analysis. This review was developed based on a symposium of the American Geriatrics Society Annual Meeting on the use and interpretation of PS analysis in May 2014. PS analysis involves two steps: estimation of PS and estimation of the treatment effect using PS. Typically estimated from a logistic model, PS reflects the probability of receiving a treatment given observed characteristics of an individual. PS can be viewed as a summary score that contains information on multiple confounders and is used in matching, weighting, or stratification to achieve confounder balance between the treatment groups to estimate the treatment effect. Of these methods, matching and weighting generally reduce confounding more effectively than stratification. Although PS is often included as a covariate in the outcome regression model, this is no longer a best practice because of its sensitivity to modeling assumption. None of these methods reduce confounding by unmeasured variables. The rationale, best practices, and caveats in conducting PS analysis are explained in this review using a case study that examined the effective of angiotensin-converting enzyme inhibitors on mortality and hospitalization in older adults with heart failure.

Comparative effectiveness of postoperative chemotherapy among older patients with non-metastatic rectal cancer treated with preoperative chemoradiotherapy

OBJECTIVE

Postoperative chemotherapy is standard following preoperative chemoradiation therapy (CRT) and curative resection for clinically staged II/III rectal cancer. Recent trials have questioned whether postoperative chemotherapy improves overall survival. The objective of the study was to evaluate the comparative effectiveness of postoperative chemotherapy following CRT or radiation therapy (RT) with specific attention to the impact of age on postoperative chemotherapy effectiveness.

MATERIALS AND METHODS

Patients treated with CRT or RT then resection of pathologically staged 0-III rectal cancer diagnosed from 2004 to 2009 were identified from the Surveillance, Epidemiology and End Results program-Medicare database. Propensity score weighted Cox proportional hazards models and Kaplan Meier methods were used to compare the effectiveness of 1) postoperative 5-fluorouracil (5-FU) or capecitabine to no treatment and 2) postoperative oxaliplatin+5-FU/capecitabine to 5-FU/capecitabine alone on mortality. Results were stratified by age.

RESULTS

We identified 1316 patients; 49% received postoperative chemotherapy, 341 (52%) included oxaliplatin. After weighting, postoperative 5-FU/capecitabine alone was associated with decreased mortality in patients aged 66-74 (adjusted hazard ratio (aHR)=0.46, 95% CI: 0.30, 0.72), corresponding to a 5-year risk difference of -0.23, (95% CI: -0.33, -0.12). No further mortality reduction from adding oxaliplatin to 5-FU/capecitabine was seen in patients aged 66-74 (aHR=1.57, 95% CI: 0.93, 2.65). No mortality reduction for 5-FU/capecitabine alone was observed among patients aged 75+ (aHR=1.11, 95% CI: 0.76, 1.63).

CONCLUSIONS

Among patients <75years, postoperative 5-FU/capecitabine was associated with reduced mortality after preoperative CRT/RT and surgical resection; however, the addition of oxaliplatin was not associated with further mortality reduction. Decisions regarding postoperative chemotherapy after age 75 warrant consideration of individual patient risks and preferences, as benefits may be limited.