Presence Of Effect Modification Research Articles

Estimands and Estimation of COVID-19 Vaccine Effectiveness Under the Test-Negative Design: Connections to Causal Inference.

The test-negative design is routinely used for the monitoring of seasonal flu vaccine effectiveness. More recently, it has become integral to the estimation of COVID-19 vaccine effectiveness, in particular for more severe disease outcomes. Because the design has many important advantages and is becoming a mainstay for monitoring postlicensure vaccine effectiveness, epidemiologists and biostatisticians may be interested in further understanding the effect measures being estimated in these studies and connections to causal effects. Logistic regression is typically applied to estimate the conditional risk ratio but relies on correct outcome model specification and may be biased in the presence of effect modification by a confounder. We give and justify an inverse probability of treatment weighting (IPTW) estimator for the marginal risk ratio, which is valid under effect modification. We use causal directed acyclic graphs, and counterfactual arguments under assumptions about no interference and partial interference to illustrate the connection between these statistical estimands and causal quantities. We conduct a simulation study to illustrate and confirm our derivations and to evaluate the performance of the estimators. We find that if the effectiveness of the vaccine varies across patient subgroups, the logistic regression can lead to misleading estimates, but the IPTW estimator can produce unbiased estimates. We also find that in the presence of partial interference both estimators can produce misleading estimates.

S409 Gender-Specific Susceptibility to SARS-CoV-2 Infection Among Patients With Gastroesophageal Reflux Disease (GERD) Using Proton Pump Inhibitors (PPI): A Nationwide Study

Introduction: Gender is profoundly understudied in research especially when related to infectious diseases. Our aim was to examine gender-specific differences in COVID-19 diagnosis among GERD patients using PPIs. Methods: Data was analyzed from a large database (Explorys, IBM) that provides electronic health records of patients from 26 major integrated US health care systems. Only adults with diagnosis of GERD from January 1999-May 2021 and COVID-19 from December 2019-May 2021 were included. Gender, PPI use, age, heart disease (HD), chronic kidney disease (CKD), diabetes mellitus (DM), chronic obstructive pulmonary disease (COPD) were covariates of interest. Univariate and multivariate analyses were used for statistical analysis. The presence of effect modification by gender or PPI use was examined. Results: Among 6229810 GERD patients, 4067200 (65.3%) used PPIs. COVID-19 was diagnosed in 4790 (0.12%)(34% males vs 66% females). In univariate analysis, PPI users had highest odds of infection OR 2.74 [2.55-2.94, P< 0.001]. Among co-morbidities, odds of COVID-19 was highest for CKD OR 1.82 [1.70-1.93, P=0.003], DM OR 1.65 [1.55-1.73, P< 0.001], HTN OR 1.61 [1.52-1.70, P< 0.001], HD OR 1.38 [1.3-1.45, P=0.001] but lower for COPD OR 1.09 [1.02-1.17, P< 0.007]. In multivariate model, PPI use alone had the highest association with COVID-19 adjusted odds (aOR) 2.5 [2.25-2.8, P< 0.001]. In CKD aOR 2.23 [1.8-2.7, P< 0.001] but risk was modified if they used PPI aOR 0.71 [0.6-0.9, P< 0.001]. In HTN aOR 1.5 [1.2-1.6, P< 0.001] and DM aOR 1.3 [1.2-1.4, P< 0.001]. GERD patients with COPD had aOR 1.1 [0.9-1.4, P=0.21] but risk was reduced to 0.7 [0.6-0.91, P=0.20] if they used PPIs. With HD aOR 0.93 [0.8-1.1, P=0.4] but PPI use modified their risk to aOR 1.32 [1.1-1.6, P=0.001]. If model was stratified by gender, female PPI users had the highest susceptibility to COVID-19 aOR 2.68 [2.4-3.0, P< 0.001] vs male aOR 1.8 [1.51-2.1, P< 0.001]. PPI use was protective for those with COPD: males aOR 0.54 [0.39-0.73, P=0.001] vs females aOR 0.95 [0.70-1.28, p< 0.001]. Males with GERD and CKD had aOR 3.1 [2.3-4.1, P< 0.001] vs female aOR 1.8 [1.2-2.3, P< 0.001]. Susceptibility to COVID-19 was modified in patients with HD if they used PPIs: males OR 1.32 [1.1-1.6, P=0.001] vs female OR 1.8 [1.4-2.4, P< 0.001]. Conclusion: In GERD patients, females PPI users had higher odds of COVID-19 diagnosis compared to males if they had no comorbidities. GERD patients with CKD had the highest risk of COVID-19, especially if they were males.Figure 1.: Multivariable regression model. A forest plot showing odds ratio and 95% confidence interval for COVID-19 diagnosis among patients with gastroesophageal reflux disease (GERD) with select co-morbidities. Model stratified by gender. Effect modification by proton pump inhibitors (PPI) use was also examined. Abbreviations: CKD, chronic kidney disease; HTN, hypertension; DM, diabetes mellitus.Table 1.: Baseline Characteristics of Patients with Gastroesophageal Reflux Disease Included in the Study. Abbreviations: PPI, proton pump inhibitor; COPD, chronic obstructive pulmonary disease; CKD, chronic kidney disease; CI, confidence interval.

S411 Effect of Histamine-2 Receptor Blockers Use on COVID-19 Susceptibility in Patients With Gastroesophageal Reflux Disease (GERD): A Population-Based Study

Introduction: Studies have demonstrated that GERD is among the comorbidities that correlate with poorer outcomes in hospitalized patients with COVID-19. Proton pump inhibitors (PPI) increase the risk for enteric infections likely due to PPI effect on intra-gastric pH. However, evidence regarding the link between PPI use and COVID-19 is still conflicting. On the other hand, famotidine has been shown to correlate with improved clinical outcomes in hospitalized COVID-19 patients. We aimed to examine the relationship between COVID-19 risk and use of H2RB or PPIs in GERD patients. Methods: Data was extracted from a large commercial database (Explorys, IBM) that aggregates electronic health records from 26 large nationwide healthcare systems. Only adult patients with diagnosis of GERD from January 1999-May 2021 and COVID-19 from December 2019-May 2021 were included. Baseline characteristics including race, gender, H2RB or PPI use and health insurance type and status were collected. Incidence of COVID-19 among risk groups were examined. Univariable and multivariable logistic regression analyses were performed.The presence of effect modification by H2RB use was examined. Multivariate model was stratified by H2RB use. Results: We identified 6215670 GERD patients of which 4057825 (65%) were exclusive PPI users. A total of 1626345 (40%) used both H2RBs and PPIs and only 546930 (9%) used H2RBs alone. COVID-19 was diagnosed in 1140 (0.03%) patients using PPIs alone (62% females vs 38% males), 1300 (0.08%) users of both H2RB and PPI (68% females vs 32% males) and 190 (0.03%) patients using H2RBs alone (63% females vs 37% males). In univariate model, H2RB use was significantly associated with COVID-19 diagnosis OR 2.47 [2.29-2.66, P< 0.001]. COVID-19 risk was significantly modified if patient used both PPI and H2RB OR 1.92 [1.72-2.15, P< 0.001]. In multivariate model, after adjusting for covariates, PPI use was independently associated with COVID-19 diagnosis OR 3.4 [3.1-3.76, P< 0.001]. H2RB use alone was significantly associated with COVID-19 diagnosis OR 2.8 [2.35-3.2, P< 0.001]. But the use of both PPI and H2RB significantly reduced the risk of infection OR 0.63 [0.55-0.74, P< 0.001]. If the model was stratified by H2RB use, COVID-19 risk was reduced in PPI users but unchanged in patients regardless of their baseline demographics (Figure 1). Conclusion: In GERD patients, exclusive use of PPIs or H2RBs was significantly associated with COVID-19 diagnosis. Concurrent use may modify susceptibility to COVID-19.Figure 1.: Multivariable regression analysis stratified by histamine-2 receptor blockers (H2RB) use. A forest plot showing odds ratio (OR) and 95% confidence interval (CI) for the association between COVID-19 and select baseline demographic data in patients with gastroesophageal reflux disease (GERD).

Abstract P195: Systolic Blood Pressure and 8-Year Mortality in Community-Dwelling Older Adults. Does Frailty Modify This Effect?

Introduction: The current literature does not provide sufficient evidence to support different treatment targets for hypertension in older adults. Hypothesis: The effect of systolic blood pressure (SBP) on 8-years mortality may vary by frailty status in community-dwelling older adults. Methods: A frailty index (FI) was constructed according to the model of deficit accumulation with 38 items from the baseline exam at the Activity and Function in the Elderly Study. Each item had a score from 0 (no deficit) to 1 (deficit). FI represents the sum of all scores divided by 38. Cox-proportional hazards models adjusted for age, sex, education, smoking, diastolic blood pressure (DBP) and antihypertensive medications analyzed the association between SBP and 8-year mortality, evaluating the presence of effect modification by frailty. Results: A total of 940 participants were included in the analysis [median age 74, interquartile range (IQR 70.1, 81.0), 42.4% women]. The prevalence of hypertension was 54%, with median SBP 144.5 mmHg (IQR 135.0, 150.0), and median DBP 78.5 mmHg (IQR 71.0, 86.5). Median follow-up time was 8.1 years. A total of 201 deaths were observed. The median FI was 0.113 (IQR 0.069, 0.113), with 19.9% (n=187, 79 deaths) being identified as frail (FI 0.2). We detected effect modification by frailty (p-value interaction term &lt;0.05). Table 1 summarizes the hazard ratios for the multivariable adjusted models stratified by frailty. Conclusions: Our results suggest the presence of effect modification by frailty with a possible protective effect of elevated SBP in frail subjects with respect to 8-year mortality even after adjustment for DBP and antihypertensive treatment.

A comparison of different methods to handle missing data in the context of propensity score analysis

Propensity score analysis is a popular method to control for confounding in observational studies. A challenge in propensity methods is missing values in confounders. Several strategies for handling missing values exist, but guidance in choosing the best method is needed. In this simulation study, we compared four strategies of handling missing covariate values in propensity matching and propensity weighting. These methods include: complete case analysis, missing indicator method, multiple imputation and combining multiple imputation and missing indicator method. Concurrently, we aimed to provide guidance in choosing the optimal strategy. Simulated scenarios varied regarding missing mechanism, presence of effect modification or unmeasured confounding. Additionally, we demonstrated how missingness graphs help clarifying the missing structure. When no effect modification existed, complete case analysis yielded valid causal treatment effects even when data were missing not at random. In some situations, complete case analysis was also able to partially correct for unmeasured confounding. Multiple imputation worked well if the data were missing (completely) at random, and if the imputation model was correctly specified. In the presence of effect modification, more complex imputation models than default options of commonly used statistical software were required. Multiple imputation may fail when data are missing not at random. Here, combining multiple imputation and the missing indicator method reduced the bias as the missing indicator variable can be a proxy for unobserved confounding. The optimal way to handle missing values in covariates of propensity score models depends on the missing data structure and the presence of effect modification. When effect modification is present, default settings of imputation methods may yield biased results even if data are missing at random.

Quantifying heterogeneity in individual participant data meta-analysis with binary outcomes

BackgroundIn meta-analyses (MA), effect estimates that are pooled together will often be heterogeneous. Determining how substantial heterogeneity is is an important aspect of MA.MethodWe consider how best to quantify heterogeneity in the context of individual participant data meta-analysis (IPD-MA) of binary data. Both two- and one-stage approaches are evaluated via simulation study. We consider conventional I2 and R2 statistics estimated via a two-stage approach and R2 estimated via a one-stage approach. We propose a simulation-based intraclass correlation coefficient (ICC) adapted from Goldstein et al. to estimate the I2, from the one-stage approach.ResultsResults show that when there is no effect modification, the estimated I2 from the two-stage model is underestimated, while in the one-stage model, it is overestimated. In the presence of effect modification, the estimated I2 from the one-stage model has better performance than that from the two-stage model when the prevalence of the outcome is high. The I2 from the two-stage model is less sensitive to the strength of effect modification when the number of studies is large and prevalence is low.ConclusionsThe simulation-based I2 based on a one-stage approach has better performance than the conventional I2 based on a two-stage approach when there is strong effect modification with high prevalence.

When evidence of heat-related vulnerability depends on the contrast measure.

Many studies assessing which population subgroups are more vulnerable to heat-related mortality and morbidity have been conducted in recent years. However, given the non-linear (U or J shaped) relationship of temperature with mortality and morbidity, they generally consider only a single contrast measure to report evidence of heat-related vulnerability, despite the possibility that vulnerability depends on the selected contrast measure. In this manuscript, we highlight the importance of considering such issue in further studies by providing evidence for and against heat-related vulnerability using two different temperature contrast measures. We conducted time series analyses to characterize the association between mortality and mean daily temperature in Montreal, Canada (1990-2010). We used age (≥65 vs. 0-64years) as the effect modifier in stratified analyses. We assessed the presence of effect modification using Cochran Q tests. As contrast measures, we used (1) the percentage change in the outcome above 25°C, obtained through spline functions showing a linear relationship after this threshold and (2) a comparison of two percentiles (26 vs. 20°C) of the temperature. We found that evidence of effect modification depended on the contrast measure used. We encourage researchers aiming to identify populations more vulnerable to heat to perform sensitivity analyses using different contrast measures.

Susceptibility factors and long term effects of air pollution: mortality among 3 sub-cohorts of the Italian Longitudinal Study. Results of the LIFE MED HISS project (LIFE12 ENV/IT/000834).

Introduction There are few studies assessing the role played by susceptibility factors on mortality related to long term exposure to air pollution. The objective of this study is to estimate the risks of mortality related to major pollutants among people with diabetes, CVD, respiratory diseases or aging more than 65 years at the start of follow up. Methods The study is based on 13 year of mortality follow-up in three susceptible sub-cohorts (individuals with pre-existing diabetes, cardiovascular or respiratory diseases) resident in 1449 municipalities belonging to the Italian Longitudinal Study 1999-2012. The exposure to PM2.5, achieved by annual gridded data at spatial resolution of 4x4Km2 integrated with observations and up-scaled at municipality level, was assigned to each cohort member. Self-reported diseases, collected in the Survey, were used to recruit the sub-cohorts. Natural and cause-specific mortality risks were computed using a multivariate time-dependent Cox model, controlling for individual confounders, testing the risks proportionality assumption and the presence of effect modification. As sensitivity analysis, other susceptible cohorts were investigated. Results 4599 and 5807 individuals, respectively, had diabetes and CVD at baseline. For 10 μg/m3 increase of PM2.5 the adjusted hazard ratios and their 95% confidence intervals (HR, CI95%) for natural mortality were 1.05(1.00;1.10) among people with diabetes and CVD, 1.07(1.03;1.12) among people with chronic respiratory disease. Among those aged 65 years or more, 23069 individuals, the health risks for natural and lung cancer mortality were 1.04(1.02;1.06) and 1.22(1.12;1.32), respectively. Conclusions Pre-existing cardiovascular and respiratory diseases or age over 65 years seem to be risks modifiers leading higher impact of air pollution. It’s still not clear, however, if it is a true effect modification or it is associated with other health characteristics of the subjects.

Effect Modification and Interaction Terms: It Takes Two to Tango

In this Investigators' Corner I look more deeply into the previously discussed phenomenon of effect modification. I revisit an explanation and examples of the phenomenon and then examine how to account for it statistically. Specifically, I show, in detail, how to write a regression equation that includes interaction terms that account for the effect modification. Finally, I look at interpretation of regression coefficients both with and without the presence of effect modification, and the associated interaction terms.

Abstract A01: Prism regression: A new statistical tool for understanding determinants of cancer health disparities

Abstract As part of our ongoing collaboration with community stakeholders throughout the Miami metropolitan area, we have become increasingly aware of the limitations imposed by current measurements of race. Our stakeholders uphold that the categorical variable (black vs. white) most commonly used to approximate this social construct lacks necessary dimension for understanding its significance and contribution to persistent health disparity. In response to this input, we introduce a class of generalized regression models for cancer outcomes, which allow the coefficient of race to vary as a function of other variables. These other variables involve factors that typically interact with the race focus variable. Existing varying coefficient models (VCMs) have been limited in a number of ways: they do not easily extend to allow multiple effect modifiers at once, and they do not allow interactions between the effect modifier variables. We developed “prism regression” to overcome these limitations in a data-adaptive, non-parametric manner. Joint modeling of the effect modifiers can be viewed as a prism through which the overall race coefficient is “refracted” in different directions (the variables defining the prism are what we term prism variables). This allows one to delineate patterns of racial differences in cancer outcomes that were previously hidden or masked. Our particular implementation of prism regression uses a rule-based decision tree structure, which is extremely intuitive to understand. Importantly, these tree-like structures can be extended in elegant ways to allow for much more complex effect modification. One such example, which we term hierarchical prism regression, allows the effect of race to vary as a function of layered, directional effect modifiers (i.e. prism variables and hierarchical variables together). This cannot be accomplished by existing methods such as multiway interactions, mediation or moderation. We detail the prism regression methodology and demonstrate some optimality properties based on a new constrained parameter estimation algorithm called “prism fusion” that provides not only more accurate predictions due to shrinkage, but also evidence for or against the presence of effect modification. We compare prism regression to existing methods using a series of simulation studies and show significant gains in predictive accuracy. In addition, we analyzed data from the cancer registry of Florida to search for complex effect modification of racial disparity. The Florida Cancer Data System is Florida's legislatively mandated, population-based, statewide cancer registry. It is the second largest population based, cancer incidence registry in the nation processing over 180,000 new cases are from patient medical records annually, corresponding to 1155,000 newly diagnosed tumors since 1981. Cancer cases are submitted by hospitals, freestanding ambulatory surgical facilities, radiation therapy facilities, private physicians and death certificates. Information includes routine personal and demographic data, includes diagnosis, stage of disease, tissue pathology, and first course of medical treatment as well as passive death certificate linkage information. We illustrate the usefulness of prism regression focusing on breast cancer patients from the FCDS registry to investigate the individual and contextual level factors that lead to differential effects of race in relation to stage of disease at diagnosis. Individual level variables obtained from the FCDS registry are linked by geocode to obtain contextual variables from census data. We consider the individual level information as prism variables, and the geocode linked contextual level variables as hierarchical variables. Citation Format: Shari Messinger, Erin Kobetz, J Sunil Rao. Prism regression: A new statistical tool for understanding determinants of cancer health disparities. [abstract]. In: Proceedings of the Seventh AACR Conference on The Science of Health Disparities in Racial/Ethnic Minorities and the Medically Underserved; Nov 9-12, 2014; San Antonio, TX. Philadelphia (PA): AACR; Cancer Epidemiol Biomarkers Prev 2015;24(10 Suppl):Abstract nr A01.

Abstract P430: Spatial Proximity and Membership to Gyms or Related Facilities Jointly Predict Objectively Measured Physical Activity among Adults in New York City

Background: Over half of the adult population in the United States fails to meet public health recommendations for physical activity. Neighborhood physical activity facilities may help encourage moderate and vigorous physical activity. We hypothesize that commercial physical activity facilities near the home would predict both membership in a gym or other recreation facility in the past 12 months and current physical activity. We further hypothesize the presence of effect modification, such that physical activity facilities would have a stronger association with activity among individuals with a facility membership. Methods: Data were from the New York City Physical Activity and Transit Monitoring Study, a sample of 679 New York City adults aged 18 years and older with physical activity measured by accelerometer. Participants were excluded for incomplete data, extreme values for height, weight or BMI, or if their home address could not be geocoded. The final analytic sample was 625. Counts of commercial physical activity facilities within 1 km of each participant’s home address were generated from the National Establishment Time-Series data for year 2010. Linear and logistic regression models incorporated robust standard errors, sample weights, and adjustment for individual- and neighborhood-level characteristics. Results: Individuals living near more commercial physical activity facilities were more likely to report membership in a gym or other facility (adjusted odds ratio for top versus bottom quartile of facility count: 3.80; 95% CI: 1.60-9.02). The count of facilities was also associated with more physical activity as measured by accelerometer, particularly for those individuals reporting membership in a gym or other recreation facility (see figure). Conclusion: The evaluation of opportunities for physical activity should include accessibility of commercial physical activity facilities, including financial or social barriers to membership.

Avoidance of representativeness in presence of effect modification

Avoidance of representativeness in presence of effect modification

A cohort study investigating aspirin use and survival in men with prostate cancer

Aspirin use has been associated with reduced mortality from cancer including prostate cancer in some studies. A number of anti-cancer mechanisms of aspirin have been proposed, including the inhibition of the cyclooxygenase enzymes, through which aspirin mediates both anti-platelet and anti-inflammatory activities. This cohort study examines associations between pre-diagnostic aspirin use (overall and by dose and dosing intensity) and mortality in men with localised prostate cancer. Men with stage I-III prostate cancer were identified from Irish National Cancer Registry records, which have been linked to national prescribing data from the Irish General Medical Services scheme. Aspirin use in the year preceding prostate cancer diagnosis was identified from this linked prescription-claims data. Adjusted hazard ratios (HRs) and 95% confidence intervals (CIs) were estimated for associations between aspirin use and all-cause and prostate cancer-specific mortality. Associations between prescribed dose and dosing intensity were examined. The presence of effect modification by the type of treatment received and tumour characteristics was also assessed. Two thousand nine hundred and thirty-six men with a diagnosis of stage I-III prostate cancer (2001-2006) were identified (aspirin users, n = 1131). The median duration of patient follow-up was 5.5 years. In adjusted analyses, aspirin use was associated with a small, but non-significant, reduced risk of prostate cancer-specific mortality (HR = 0.88, 95% CI 0.67-1.15). In dose-response analyses, stronger associations with prostate cancer-specific mortality were observed in men with higher aspirin dosing intensity (HR = 0.73, 95% CI 0.51-1.05) and in men receiving >75 mg of aspirin (HR = 0.61, 95% CI 0.37-0.99). Analyses of effect modification by treatment type or tumour characteristics were non-significant. Consistent with prior studies, aspirin use was associated with a non-significant reduced risk of prostate cancer-specific mortality in men with localised prostate cancer. Men receiving higher doses of aspirin had a statistically significant reduced risk of prostate cancer-specific mortality. These findings regarding an aspirin dose require further investigation.

Effect Modification and Design Sensitivity in Observational Studies

In an observational study of treatment effects, subjects are not randomly assigned to treatment or control, so differing outcomes in treated and control groups may reflect a bias from nonrandom assignment rather than a treatment effect. After adjusting for measured pretreatment covariates, perhaps by matching, a sensitivity analysis determines the magnitude of bias from an unmeasured covariate that would need to be present to alter the conclusions of the naive analysis that presumes adjustments eliminated all bias. Other things being equal, larger effects tend to be less sensitive to bias than smaller effects. Effect modification is an interaction between a treatment and a pretreatment covariate controlled by matching, so that the treatment effect is larger at some values of the covariate than at others. In the presence of effect modification, it is possible that results are less sensitive to bias in subgroups experiencing larger effects. Two cases are considered: (i) an a priori grouping into a few categories based on covariates controlled by matching and (ii) a grouping discovered empirically in the data at hand. In case (i), subgroup specific bounds on p-values are combined using the truncated product of p-values. In case (ii), information that is fixed under the null hypothesis of no treatment effect is used to partition matched pairs in the hope of identifying pairs with larger effects. The methods are evaluated using an asymptotic device, the design sensitivity, and using simulation. Sensitivity analysis for a test of the global null hypothesis of no effect is converted to sensitivity analyses for subgroup analyses using closed testing. A study of an intervention to control malaria in Africa is used to illustrate.

Confounding and effect modification: distribution and measure.

The paper considers the properties of and relations between confounding and effect modification from the perspective of causal inference and with a distinction drawn as to how each of these two epidemiologic concepts can be defined both with respect to a distribution of potential outcomes or with respect to a specific effect measure. Both concepts are conditional on other covariates but the form this conditionality takes differs. Both concepts are also properties of the population, and both are relative to the specific exposure and to the specific outcome. For a particular population, the presence of confounding depends on how the exposure was assigned; the presence of effect modification does not. The possibility of confounding without effect modification and vice versa is discussed both with respect to distribution and measure. Discussion is given as to how confounding and effect modification relate to statistical models and to the relevance of the points made in the paper to data analysis and interpretation.

Similar Relation of Age and Height to Lung Function Among Whites, African Americans, and Hispanics

Current guidelines recommend separate spirometry reference equations for whites, African Americans, and Mexican Americans, but the justification for this recommendation is controversial. The authors examined the statistical justification for race/ethnic-specific reference equations in adults in the Third National Health and Nutrition Examination Survey (1988-1994) and the Multi-Ethnic Study of Atherosclerosis Lung Study (2000-2006). Spirometry was measured following American Thoracic Society guidelines. "Statistical justification" was defined as the presence of effect modification by race/ethnicity among never-smoking participants without respiratory disease or symptoms and was tested with interaction terms for race/ethnicity (× age and height) in regression models. There was no evidence of effect modification by race/ethnicity for forced expiratory volume in 1 second, forced vital capacity, or the forced expiratory volume in 1 second/forced vital capacity ratio among white, African-American, and Mexican-American men or women on an additive scale or a log scale. Interaction terms for race/ethnicity explained less than 1% of variability in lung function. The mean lung function for a given age, gender, and height was the same for whites and Mexican Americans but was lower for African Americans. Findings were similar in the Multi-Ethnic Study of Atherosclerosis Lung Study. The associations of age and height with lung function are similar across the 3 major US race/ethnic groups. Multiethnic rather than race/ethnic-specific spirometry reference equations are applicable for the US population.

Propensity scores in the presence of effect modification: A case study using the comparison of mortality on hemodialysis versus peritoneal dialysis

PurposeTo control for confounding bias from non-random treatment assignment in observational data, both traditional multivariable models and more recently propensity score approaches have been applied. Our aim was to compare a propensity score-stratified model with a traditional multivariable-adjusted model, specifically in estimating survival of hemodialysis (HD) versus peritoneal dialysis (PD) patients.MethodsUsing the Dutch End-Stage Renal Disease Registry, we constructed a propensity score, predicting PD assignment from age, gender, primary renal disease, center of dialysis, and year of first renal replacement therapy. We developed two Cox proportional hazards regression models to estimate survival on PD relative to HD, a propensity score-stratified model stratifying on the propensity score and a multivariable-adjusted model, and tested several interaction terms in both models.ResultsThe propensity score performed well: it showed a reasonable fit, had a good c-statistic, calibrated well and balanced the covariates. The main-effects multivariable-adjusted model and the propensity score-stratified univariable Cox model resulted in similar relative mortality risk estimates of PD compared with HD (0.99 and 0.97, respectively) with fewer significant covariates in the propensity model. After introducing the missing interaction variables for effect modification in both models, the mortality risk estimates for both main effects and interactions remained comparable, but the propensity score model had nearly as many covariates because of the additional interaction variables.ConclusionAlthough the propensity score performed well, it did not alter the treatment effect in the outcome model and lost its advantage of parsimony in the presence of effect modification.

Assumptions of IV Methods for Observational Epidemiology

Instrumental variable (IV) methods are becoming increasingly popular as they seem to offer the only viable way to overcome the problem of unobserved confounding in observational studies. However, some attention has to be paid to the details, as not all such methods target the same causal parameters and some rely on more restrictive parametric assumptions than others. We therefore discuss and contrast the most common IV approaches with relevance to typical applications in observational epidemiology. Further, we illustrate and compare the asymptotic bias of these IV estimators when underlying assumptions are violated in a numerical study. One of our conclusions is that all IV methods encounter problems in the presence of effect modification by unobserved confounders. Since this can never be ruled out for sure, we recommend that practical applications of IV estimators be accompanied routinely by a sensitivity analysis.

Different Methods of Balancing Covariates Leading to Different Effect Estimates in the Presence of Effect Modification

A number of covariate-balancing methods, based on the propensity score, are widely used to estimate treatment effects in observational studies. If the treatment effect varies with the propensity score, however, different methods can give very different answers. The authors illustrate this effect by using data from a United Kingdom–based registry of subjects treated with anti–tumor necrosis factor drugs for rheumatoid arthritis. Estimates of the effect of these drugs on mortality varied from a relative risk of 0.4 (95% confidence interval: 0.16, 0.91) to a relative risk of 1.3 (95% confidence interval: 0.8, 2.25), depending on the balancing method chosen. The authors show that these differences were due to a combination of an interaction between propensity score and treatment effect and to differences in weighting subjects with different propensity scores. Thus, the methods are being used to calculate average treatment effects in populations with very different distributions of effect-modifying variables, resulting in different overall estimates. This phenomenon highlights the importance of careful selection of the covariate-balancing method so that the overall estimate has a meaningful interpretation.

Inadequate prenatal care and its association with adverse pregnancy outcomes: A comparison of indices

BackgroundThe objectives of this study were to determine rates of prenatal care utilization in Winnipeg, Manitoba, Canada from 1991 to 2000; to compare two indices of prenatal care utilization in identifying the proportion of the population receiving inadequate prenatal care; to determine the association between inadequate prenatal care and adverse pregnancy outcomes (preterm birth, low birth weight [LBW], and small-for-gestational age [SGA]), using each of the indices; and, to assess whether or not, and to what extent, gestational age modifies this association.MethodsWe conducted a population-based study of women having a hospital-based singleton live birth from 1991 to 2000 (N = 80,989). Data sources consisted of a linked mother-baby database and a physician claims file maintained by Manitoba Health. Rates of inadequate prenatal care were calculated using two indices, the R-GINDEX and the APNCU. Logistic regression analysis was used to determine the association between inadequate prenatal care and adverse pregnancy outcomes. Stratified analysis was then used to determine whether the association between inadequate prenatal care and LBW or SGA differed by gestational age.ResultsRates of inadequate/no prenatal care ranged from 8.3% using APNCU to 8.9% using R-GINDEX. The association between inadequate prenatal care and preterm birth and LBW varied depending on the index used, with adjusted odds ratios (AOR) ranging from 1.0 to 1.3. In contrast, both indices revealed the same strength of association of inadequate prenatal care with SGA (AOR 1.4). Both indices demonstrated heterogeneity (non-uniformity) across gestational age strata, indicating the presence of effect modification by gestational age.ConclusionSelection of a prenatal care utilization index requires careful consideration of its methodological underpinnings and limitations. The two indices compared in this study revealed different patterns of utilization of prenatal care, and should not be used interchangeably. Use of these indices to study the association between utilization of prenatal care and pregnancy outcomes affected by the duration of pregnancy should be approached cautiously.