Association between coffee and incident heart failure: A prospective cohort study from the UK Biobank

BACKGROUND AND AIMS

The relationship between coffee consumption and heart failure (HF) incidence is inconclusive. This study aimed to explore the association between time-varying coffee consumption and incident HF using a longitudinal study design.

METHODS AND RESULTS

Data were obtained from the UK Biobank, comprising 497,503 adults (age, 56.5 ± 8.1 years; 54.6% women) who were free from HF at baseline in 2006-2010. The median follow-up time for the HF incidence was 11.9 years. Marginal structural models (MSM) were employed to adjust for potential time-varying confounders and account for bias caused by loss of follow-up. Furthermore, we used a restricted cubic spline to test and describe the nonlinear relationship between coffee consumption and HF risk. At baseline, 70.5% of participants reported drinking ≥1 cups/d coffee and 2.7% participants developed HF. After adjusting for potential confounders, we identified a nonlinear J-shaped association between coffee consumption and HF risk (P < 0.001). Compared with drinking coffee <1 cups/d, 1-2 cups/d (HR = 0.878; 95% CI: 0.838-0.920), 3-4 cups/d (HR = 0.920; 95% CI: 0.869-0.974) may be associated with a reduced risk of HF, while >6 cups/d (HR = 1.209; 95% CI: 1.056-1.385) may be associated with a higher risk of HF. However, sensitive analyses stratified by gender and smoking status indicated that >6 cups/d does not significantly increase the risk of HF. Additionally, the type of coffee was found to significant impact on the incidence of HF (P < 0.05).

CONCLUSION

In this large cohort of UK adults, moderate coffee consumption may reduce risk of HF incidence.

Semiparametric Bayesian inference for optimal dynamic treatment regimes via dynamic marginal structural models

Considerable statistical work done on dynamic treatment regimes (DTRs) is in the frequentist paradigm, but Bayesian methods may have much to offer in this setting as they allow for the appropriate representation and propagation of uncertainty, including at the individual level. In this work, we extend the use of recently developed Bayesian methods for Marginal Structural Models to arrive at inference of DTRs. We do this (i) by linking the observational world with a world in which all patients are randomized to a DTR, thereby allowing for causal inference and then (ii) by maximizing a posterior predictive utility, where the posterior distribution has been obtained from nonparametric prior assumptions on the observational world data-generating process. Our approach relies on Bayesian semiparametric inference, where inference about a finite-dimensional parameter is made all while working within an infinite-dimensional space of distributions. We further study Bayesian inference of DTRs in the double robust setting by using posterior predictive inference and the nonparametric Bayesian bootstrap. The proposed methods allow for uncertainty quantification at the individual level, thereby enabling personalized decision-making. We examine the performance of these methods via simulation and demonstrate their utility by exploring whether to adapt HIV therapy to a measure of patients' liver health, in order to minimize liver scarring.

Emulating a target trial using primary care electronic health records: SGLT-2i medications and Hemoglobin A1c

Substantial effort has been dedicated to conducting randomized controlled experiments to generate clinical evidence for diabetes treatment. Randomized controlled experiments are the gold standard to establish cause and effect. However, due to their high-cost and time-commitment, large observational databases such as those comprised of electronic health record (EHR) data collected in routine primary care may provide an alternative source to address such causal objectives. We used a Canadian primary care repository housed at University of Toronto to emulate a randomized experiment. We estimated the effectiveness of sodium-glucose co-transporter 2 inhibitors (SGLT-2i) medications for patients with diabetes using Hemoglobin A1c (HbA1c) as a primary outcome and marker for glycemic control from 2018 to 2021. We assumed an intention-to-treat analysis for prescribed treatment, with analyses based on the treatment assigned rather than the treatment eventually received. We defined the causal contrast of interest as the net change in HbA1c (%) between the group receiving standard of care versus the group receiving SGLT-2i medications. Using a counterfactual framework, marginal structural models demonstrated a reduction in mean HbA1c with theinitiation of SGLT-2i medications. These findings provided similar effect sizes to those from earlier clinical trials on assessing the effectiveness of SGLT-2i medications.

Estimation of marginal structural models under irregular visits and unmeasured confounder: calibrated inverse probability weights

Clinical information collected in electronic health records (EHRs) is becoming an essential source to emulate randomized experiments. Since patients do not interact with the healthcare system at random, the longitudinal information in large observational databases must account for irregular visits. Moreover, we need to also account for subject-specific unmeasured confounders which may act as a common cause for treatment assignment mechanism (e.g. glucose-lowering medications) while also influencing the outcome (e.g. Hemoglobin A1c). We used the calibration of longitudinal weights to improve the finite sample properties and to account for subject-specific unmeasured confounders. A Monte Carlo simulation study is conducted to evaluate the performance of calibrated inverse probability estimators using time-dependent treatment assignment and irregular visits with subject-specific unmeasured confounders. The simulation study showed that the longitudinal weights with calibrated restrictions improved the finite sample bias when compared to the stabilized weights. The application of the calibrated weights is demonstrated using the exposure of glucose lowering medications and the longitudinal outcome of Hemoglobin A1c. Our results support the effectiveness of glucose lowering medications in reducing Hemoglobin A1c among type II diabetes patients with elevated glycemic index ([Formula: see text]) using stabilized and calibrated weights.

Emulation of a target trial with sustained treatment strategies: an application to prostate cancer using both inverse probability weighting and the g-formula

As with many chronic illnesses, recurrent prostate cancer generally requires sustained treatment to prolong survival. However, initiating treatment immediately after recurrence may negatively impact quality of life without any survival gains. Therefore, we consider sustained strategies for initiating treatment based on specific characteristics of prostate-specific antigen (PSA), which can indicate disease progression. We define the protocol for a target trial comparing treatment strategies based on PSA doubling time, in which androgen deprivation therapy is initiated only after doubling time decreases below a certain threshold. Such a treatment strategy means the timing of treatment initiation (if ever) is not known at baseline, and the target trial protocol must explicitly specify the frequency of PSA monitoring until the threshold is met, as well as the duration of treatment. We describe these and other components of a target trial that need to be specified in order for such a trial to be emulated in observational data. We then use the parametric g-formula and inverse-probability weighted dynamic marginal structural models to emulate our target trial in a cohort of prostate cancer patients from clinics across the United States.

Longitudinal effects of lipid indices on incident cardiovascular diseases adjusting for time-varying confounding using marginal structural models: 25 years follow-up of two US cohort studies

Background

This study assesses the effect of blood lipid indices and lipid ratios on cardiovascular diseases (CVDs) using inverse probability-of-exposure weighted estimation of marginal structural models (MSMs).

Methods

A pooled dataset of two US representative cohort studies, including 16736 participants aged 42-84 years with complete information at baseline, was used. The effect of each lipid index, including low-density lipoprotein cholesterol (LDL-C), high-density lipoprotein cholesterol (HDL-C), triglyceride (TG), ratios of TC/HDL-C, LDL-C/HDL-C, and TG/HDL-C on coronary heart disease (CHD) and stroke were estimated using weighted Cox regression.

Results

There were 1638 cases of CHD and 1017 cases of stroke during a median follow-up of 17.1 years (interquartile range: 8.5 to 25.7). Compared to optimal levels, the risk of CVD outcomes increased substantially in high levels of TC, LDL-C, TC/HDL-C, and LDL-C/HDL-C. If everyone had always had high levels of TC (≥240 mg/dL), risk of CHD would have been 2.15 times higher, and risk of stroke 1.35 times higher than if they had always had optimal levels (<200 mg/dL). Moreover, if all participants had been kept at very high (≥190 mg/dL) levels of LDL-C, risk of CHD would have been 2.62 times higher and risk of stroke would have been 1.92 times higher than if all participants had been kept at optimal levels, respectively. Our results suggest that high levels of HDL-C may be protective for CHD, but not for stroke. There was also no evidence of an adverse effect of high triglyceride levels on stroke.

Conclusions

Using MSM, this study highlights the effect of TC and LDL-C on CVD, with a stronger effect on CHD than on stroke. There was no evidence for a protective effect of high levels of HDL-C on stroke. Besides, triglyceride was not found to affect stroke.

Marginal structural models for multilevel clustered data

Marginal structural models (MSMs), which adopt inverse probability treatment weighting in the estimating equations, are powerful tools to estimate the causal effects of time-varying exposures in the presence of time-dependent confounders. Motivated by the Conservation of Hearing Study (CHEARS) Audiology Assessment Arm (AAA) where repeated hearing measurements were clustered by study participants, time, and testing sites, we propose two methods to account for the multilevel correlation structure when fitting the MSMs. The first method directly models the covariance of the repeated outcomes when solving the weighted generalized estimating equations for MSMs, while the second two-stage analysis approach fits cluster-specific MSMs first and then combines the estimated parameters using mixed-effects meta-analysis. Finite sample simulation results suggest that our methods can obtain less biased and more efficient estimates of the parameters by accounting for the multilevel correlation. Moreover, we explore the effects of using fixed- or mixed-effects model to estimate the treatment probability on the parameter estimates of the MSMs in the presence of unmeasured cluster-level confounders. Lastly, we apply our methods to the CHEARS AAA data set, to estimate the causal effects of aspirin use on hearing loss.

Early-life exposure to Chinese famine and stroke risk in mid- to late life: the mediating roles of cognitive function and depression

BACKGROUND

Limited research has examined the role that famine exposure plays in adulthood stroke risk. We aim to explore the causal implications of early exposure to the Great Chinese Famine on stroke risk and determine whether these associations were mediated by cognitive function, and depression.

METHODS

We sampled 12,681 individuals aged 45 years and older from the China Health and Retirement Longitudinal Study (CHARLS) and divided them into fetally exposed, childhood-exposed, adolescence/adulthood-exposed and unexposed groups. Stroke was defined by self- or proxy-reported physician diagnosis. Based on a counterfactual framework, marginal structural models were used to estimate the natural direct effect and the natural indirect effects through cognitive function and depression for causal inference.

RESULTS

We found that early-life exposure to Chinese famine was directly related to increased stroke risk in mid- to late life. Cognitive function and depression accounted for a greater part of the effect for childhood famine exposure, mediating 36.35% (95%CI: 14.19, 96.19%) of the overall association between famine exposure and incident stroke, than for the fetal, adolescence/adulthood famine exposure groups. However, the natural indirect effect through depression was not significant in the fetally exposed group. The results were robust in the sensitivity analysis of model specification and unobserved confounding.

CONCLUSIONS

Our findings are consistent with the latency, pathway, and accumulation models, supporting the life-course theory. Early stages of life exposed to the Chinese Famine were associated with higher risk of stroke in mid- to late life. Enhanced cognitive and depression interventions may reduce stroke risk in middle-aged and older Chinese adults who exposure to famine in early life.

Ranking tailoring variables for constructing individualized treatment rules: an application to schizophrenia

As with many chronic conditions, matching patients with schizophrenia to the best treatment options is difficult. Selecting antipsychotic medication is especially challenging because many of the medications can have burdensome side effects. Adjusting or tailoring medications based on patients' characteristics could improve symptoms. However, it is often not known which patient characteristics are most helpful for informing treatment selection. In this paper, we address the challenge of identifying and ranking important variables for tailoring treatment decisions. We consider a value-search approach implemented through dynamic marginal structural models to estimate an optimal individualized treatment rule. We apply our methodology to the Clinical Antipsychotics Trial of Intervention and Effectiveness (CATIE) study for schizophrenia, to evaluate if some tailoring variables have greater potential than others for selecting treatments for patients with schizophrenia (Stroup et al., 2003).

Inequality in Place: Effects of Exposure to Neighborhood-Level Economic Inequality on Mortality

This study contributes to the debate on whether income inequality is harmful for health by addressing several analytical weaknesses of previous studies. Using the Panel Study of Income Dynamics in combination with tract-level measures of income inequality in the United States, we estimate the effects of differential exposure to income inequality during three decades of the life course on mortality. Our study is among the first to consider the implications of income inequality within U.S. tracts for mortality using longitudinal and individual-level data. In addition, we improve upon prior work by accounting for the dynamic relationship between local areas and individuals' health, using marginal structural models to account for changes in exposure to local income inequality. In contrast to other studies that found no significant relation between income inequality and mortality, we find that recent exposure to higher local inequality predicts higher relative risk of mortality among individuals at ages 45 or older.

Insomnia symptoms and incident heart failure: a population-based cohort study

AIMS

Heart failure (HF) is an ongoing epidemic and a serious clinical and public health issue. Currently, little is known about prospective associations between insomnia symptoms and HF incidence. We investigated the longitudinal associations between time-varying insomnia symptoms (difficulty initiating sleep, difficulty maintaining sleep, early-morning awakening, non-restorative sleep) and incident HF.

METHODS AND RESULTS

Data were obtained from the Health and Retirement Study in the US for a population-representative sample of 12,761 middle-aged and older adults (age ≥ 50 years; mean [SD] age, 66.7 [9.4] years; 57.7% females) who were free from HF at baseline in 2002. Respondents were followed for 16 years for incident HF. We employed marginal structural discrete-time survival analyses to adjust for potential time-varying biological, psycho-cognitive, and behavioral factors and to account for bias due to differential loss to follow-up. At baseline, 38.4% of the respondents reported experiencing at least one insomnia symptom. During the 16-year follow-up, 1,730 respondents developed incident HF. Respondents experiencing one (hazard ratio [HR]=1.22; 95% CI: 1.08-1.38), two (HR=1.45; 95% CI: 1.21-1.72), three (HR=1.66; 95% CI: 1.37-2.02), or four (HR=1.80; 95% CI: 1.25-2.59) insomnia symptoms had a higher hazard of incident HF than asymptomatic respondents. Respondents that had trouble initiating sleep (HR=1.17; 95%CI: 1.01-1.36), maintaining sleep (HR=1.14; 95% CI: 1.01-1.28), early-morning awakening (HR=1.20; 95% CI: 1.02-1.43), or non-restorative sleep (HR=1.25; 95% CI: 1.06-1.46) had a higher hazard of incident HF than asymptomatic respondents.

CONCLUSION

Insomnia symptoms, both cumulatively and individually, are associated with incident HF. Public health awareness and screening for insomnia symptoms in at-risk populations should be encouraged to reduce HF incidence.

Using marginal structural models to analyze randomized clinical trials with non-adherence and lost to follow up

BACKGROUND

In the presence of non-adherence and lost to follow up, results of an Intention to Treat (ITT) analysis may be biased as it is measuring the effect of assignment rather than the effect of treatment. Given that Marginal Structural Models (MSMs) adjust for such issues, this study examines the use of MSMs to assess the validity of ITT analyses in the presence of non-adherence and lost to follow up in an existing randomized clinical trial on asthma treatment.

METHODS

Inverse probability weights were obtained from a pooled logistic regression assessing the probability of staying on assigned treatment (adherence) and of remaining uncensored (censored) for subjects at each visit by treatment arm. Weights were then pooled into a MSM analysis using a Poisson generalized estimating equation with an independent correlation matrix.

RESULTS

Out of 488 participants, 174 (36%) did not adhere to the baseline assignment and 85 (17%) were lost to follow up by the end of the study. The adjusted relative risks (RR), and 95% confidence intervals (CI), obtained from the MSMs (theophylline vs. montelukast; RR=1.24; 95% CI: 0.83,1.84; theophylline vs. placebo: RR=1.01; 95% CI: 0.70,1.48; and montelukast vs. placebo: RR=0.83; 95% CI: 0.57,1.19) were nearly identical to that of the ITT analysis (theophylline vs. montelukast: RR=1.22; 95% CI: 0.82,1.86; theophylline vs. placebo: RR=0.99; 95% CI: 0.67,1.50; and montelukast vs. placebo: RR=0.82; 95% CI: 0.55,1.21).

CONCLUSION

Concordance between the results of ITT and MSMs indicate adherence and censoring may not invalidate ITT analysis. However, no adherence or censorship thresholds currently exist to assist researchers in determining when MSMs may be superior to ITT in the analysis of clinical trials with non-adherence or censorship issues, and therefore, MSMs should be conducted as a sensitivity analysis to the ITT approach in clinical trials.

Estimating the Effect of Depression on HIV Transmission Risk Behaviors Among People Who Inject Drugs in Vietnam: A Causal Approach

The burden of depression and HIV is high among people who inject drugs (PWID), yet the effect of depression on transmission risk behaviors is not well understood in this population. Using causal inference methods, we analyzed data from 455 PWID living with HIV in Vietnam 2009-2013. Study visits every 6 months over 2 years measured depressive symptoms in the past week and injecting and sexual behaviors in the prior 3 months. Severe depressive symptoms (vs. mild/no symptoms) increased injection equipment sharing (risk difference [RD] = 3.9 percentage points, 95% CI -1.7, 9.6) but not condomless sex (RD = -1.8, 95% CI -6.4, 2.8) as reported 6 months later. The cross-sectional association with injection equipment sharing at the same visit (RD = 6.2, 95% CI 1.4, 11.0) was stronger than the longitudinal effect. Interventions on depression among PWID may decrease sharing of injection equipment and the corresponding risk of HIV transmission.Clinical trial registration ClinicalTrials.gov NCT01689545.

Long-term effects of asthma medication on asthma symptoms: an application of the targeted maximum likelihood estimation

Background

Long-term effectiveness of asthma control medication has been shown in clinical trials but results from observational studies with children and adolescents are lacking. Marginal structural models estimated using targeted maximum likelihood methods are a novel statistiscal approach for such studies as it allows to account for time-varying confounders and time-varying treatment. Therefore, we aimed to calculate the long-term risk of reporting asthma symptoms in relation to control medication use in a real-life setting from childhood to adulthood applying targeted maximum likelihood estimation.

Methods

In the prospective cohort study SOLAR (Study on Occupational Allergy Risks) we followed a German subsample of 121 asthmatic children (9–11 years old) of the ISAAC II cohort (International Study of Asthma and Allergies in Childhood) until the age of 19 to 24. We obtained self-reported questionnaire data on asthma control medication use at baseline (1995–1996) and first follow-up (2002–2003) as well as self-reported asthma symptoms at baseline, first and second follow-up (2007–2009). Three hypothetical treatment scenarios were defined: early sustained intervention, early unsustained intervention and no treatment at all. We performed longitudinal targeted maximum likelihood estimation combined with Super Learner algorithm to estimate the relative risk (RR) to report asthma symptoms at SOLAR I and SOLAR II in relation to the different hypothetical scenarios.

Results

A hypothetical intervention of early sustained treatment was associated with a statistically significant risk increment of asthma symptoms at second follow-up when compared to no treatment at all (RR: 1.51, 95% CI: 1.19–1.83) or early unsustained intervention (RR:1.38, 95% CI: 1.11–1.65).

Conclusions

While we could confirm the tagerted maximum likelihood estimation to be a usable and robust statistical tool, we did not observe a beneficial effect of asthma control medication on asthma symptoms. Because of potential due to the small sample size, lack of data on disease severity and reverse causation our results should, however, be interpreted with caution.

Direct and indirect pathways between low income status and becoming a high-cost health care user in Ontario, Canada: a mediation analysis of health risk behaviors

PURPOSE

We investigated natural direct and indirect pathways between low-income status and high-cost user (HCU) transitions considering health risk behaviors as potential mediators.

METHODS

We analyzed data from respondents (aged 18+) from four pooled cycles of the Canadian Community Health Survey (2005-2010/2011) linked to administrative data in Ontario, Canada. HCUs were defined as the top 5% of the population, ranked by cost consumption in any of the five years after survey interview. Low-income status was defined from the provincial distribution of self-reported household income, with missing values imputed from neighborhood-level data. In mediation analyses based on marginal structural models, we quantified the contributions of smoking, physical inactivity, alcohol consumption, and body mass index to income-HCU associations.

RESULTS

115,091 respondents (representative of 9,661,764 Ontarians) were included in the study, of which 7.2% became HCUs. The odds of becoming HCUs were 1.36 times (95% CI: 1.25-1.48) greater for low (vs. high) income status respondents. Smoking, physical activity, alcohol consumption, and body mass index contributed 9.4%, 6.5%, 10.6%, and 4.4% to this association, respectively. Tests for exposure-mediator interactions were not statistically significant.

CONCLUSIONS

Health risk behaviors only partially explain income inequalities in future HCU transitions.

Associations between cognitive and affective job insecurity and incident purchase of psychotropic drugs: A prospective cohort study of Swedish employees

BACKGROUND

Previous research suggests that job insecurity is associated with poor mental health, but research examining how different aspects of job insecurity relate to clinical measures of poor mental health are lacking. We aimed to investigate the association between cognitive and affective job insecurity and incident purchases of psychotropic drugs.

METHODS

We included 14,586 employees participating in the Swedish Longitudinal Occupational Survey of Health (SLOSH), who answered questions on cognitive and/or affective job insecurity in 2010, 2012 or 2014. Respondents were followed in the Swedish Prescribed Drug Register (2.5 years on average). We investigated the association between job insecurity and incident psychotropic drugs with marginal structural Cox models.

RESULTS

Affective job insecurity was associated with an increased risk of purchasing any psychotropic drugs (Hazard Ratio (HR) 1.40 (95% Confidence Interval (CI) 1.04-1.89)) while cognitive job insecurity was not (HR 1.15 (95% CI 0.92-1.43)). Cognitive and affective job insecurity were both associated with antidepressants, affective job insecurity with anxiolytics, but no association was found with sedatives. Women and younger workers seemed to have higher risk compared to men and older workers, but differences were not statistically significant.

LIMITATIONS

Although job insecurity and psychotropic drugs were assessed through independent sources and several covariates were considered, unmeasured confounding cannot be ruled out.

CONCLUSIONS

The findings support that affective job insecurity is a risk factor for psychotropic drug treatment, that it may be relevant to distinguish between different types of job insecurity, and to consider sex and age as moderating factors.

Metformin use and risk of cancer in patients with type 2 diabetes: a cohort study of primary care records using inverse probability weighting of marginal structural models

BACKGROUND

Previous studies provide conflicting evidence on whether metformin is protective against cancer. When studying time-varying exposure to metformin, covariates such as body mass index (BMI) and glycated haemoglobin (HbA1c) may act as both confounders and causal pathway variables, and so cannot be handled adequately by standard regression methods. Marginal structural models (MSMs) with inverse probability of treatment weights (IPTW) can correctly adjust for such confounders. Using this approach, the main objective of this study was to estimate the effect of metformin on cancer risk compared with risk in patients with T2DM taking no medication.

METHODS

Patients with incident type 2 diabetes (T2DM) were identified in the Clinical Practice Research Datalink (CPRD), a database of electronic health records derived from primary care in the UK. Patients entered the study at diabetes diagnosis or the first point after this when they had valid HbA1c and BMI measurements, and follow-up was split into 1-month intervals. Logistic regression was used to calculate IPTW; then the effect of metformin on all cancers (including and excluding non-melanoma skin cancer) and breast, prostate, lung, colorectal and pancreatic cancers was estimated in the weighted population.

RESULTS

A total of 55 629 T2DM patients were alive and cancer-free at their study entry; 2530 people had incident cancer during a median follow-up time of 2.9 years [interquartile range (IQR) 1.3-5.4 years]. Using the MSM approach, the hazard ratio (HR) for all cancers, comparing treatment with metformin with no glucose-lowering treatment, was 1.02 (0.88-1.18). Results were robust to a range of sensitivity analyses and remained consistent when estimating the treatment effect by length of exposure. We also found no evidence of a protective effect of metformin on individual cancer outcomes.

CONCLUSIONS

We find no evidence that metformin has a causal association with cancer risk.

The roll-out of a health insurance program and its impact on the supply of healthcare services: a new method to evaluate time-varying continuous interventions

BACKGROUND

We analyze the effects of the Mexican universal health insurance program, Seguro Popular, on key variables associated with the provision of healthcare services. Given that the program was introduced gradually over a period that lasted more than a decade, the dynamics of the roll-out of the program and its reaction to the expansion of healthcare services it caused should be accounted for when evaluating the program.

METHODS

We present a new semiparametric procedure to analyze time-varying continuous interventions. This is accomplished by bringing together the literatures on continuous and on dynamic treatments. Our approach allows the researcher to estimate mean and quantile dose-response functions by applying local regression methods to appropriately weighted samples that control for time-dependent confounding.

RESULTS

Using administrative data, we show compelling evidence that Seguro Popular has incremented the human and physical resources available for healthcare services over the period 2001-2013. Moreover, we show that these effects have been heterogeneously distributed.

CONCLUSIONS

The program has proven most helpful in less vulnerable territories, leaving behind those in greater need.

Long-Term Supportive Housing is Associated with Decreased Risk for New HIV Diagnoses Among a Large Cohort of Homeless Persons in New York City

It is unknown whether providing housing to persons experiencing homelessness decreases HIV risk. Housing, including access to preventive services and counseling, might provide a period of transition for persons with HIV risk factors. We assessed whether the new HIV diagnosis rate was associated with duration of supportive housing. We linked data from a cohort of 21,689 persons without a previous HIV diagnosis who applied to a supportive housing program in New York City (NYC) during 2007-2013 to the NYC HIV surveillance registry. We used time-dependent Cox modeling to compare new HIV diagnoses among recipients of supportive housing (defined a priori, for program evaluation purposes, as persons who spent > 7 days in supportive housing; n = 6447) and unplaced applicants (remainder of cohort), after balancing the groups on baseline characteristics with propensity score weights. Compared with unplaced applicants, persons who received ≥ 3 continuous years of supportive housing had decreased risk for new HIV diagnosis (HR 0.10; CI 0.01-0.99). Risk of new HIV diagnosis decreased with longer duration placement in supportive housing. Supportive housing might aid in primary HIV prevention.

The role of venues in structuring HIV, sexually transmitted infections, and risk networks among men who have sex with men

Background

Venues form part of the sampling frame for time-location sampling, an approach often used for HIV surveillance. While sampling location is often regarded as a nuisance factor, venues may play a central role in structuring risk networks. We investigated individual reports of risk behaviors and infections among men who have sex with men (MSM) attending different venues to examine structuring of HIV risk behaviors. However, teasing apart ‘risky people’ from ‘risky places’ is difficult, as individuals cannot be randomized to attend different venues. However, we can emulate this statistically using marginal structural models, which inversely weight individuals according to their estimated probability of attending the venue.

Methods

We conducted a cross-sectional survey of 609 MSM patrons of 14 bars in San Diego, California, recruited using the Priorities for Local AIDS Control Efforts (PLACE) methodology, which consists of a multi-level identification and assessment of venues for HIV risk through population surveys.

Results and discussion

Venues differed by many factors, including participants’ reported age, ethnicity, number of lifetime male partners, past sexually transmitted infection (STI), and HIV status. In multivariable marginal structural models, venues demonstrated structuring of HIV+ status, past STI, and methamphetamine use, independently of individual-level characteristics.

Conclusions

Studies using time-location sampling should consider venue as an important covariate, and the use of marginal structural models may help to identify risky venues. This may assist in widespread, economically feasible and sustainable targeted surveillance and prevention. A more mechanistic understanding of how ‘risky venues’ emerge and structure risk is needed.

Electronic supplementary material

The online version of this article (10.1186/s12889-018-5140-3) contains supplementary material, which is available to authorized users.

Application of causal inference methods in the analyses of randomised controlled trials: a systematic review

BACKGROUND

Applications of causal inference methods to randomised controlled trial (RCT) data have usually focused on adjusting for compliance with the randomised intervention rather than on using RCT data to address other, non-randomised questions. In this paper we review use of causal inference methods to assess the impact of aspects of patient management other than the randomised intervention in RCTs.

METHODS

We identified papers that used causal inference methodology in RCT data from Medline, Premedline, Embase, Cochrane Library, and Web of Science from 1986 to September 2014, using a forward citation search of five seminal papers, and a keyword search. We did not include studies where inverse probability weighting was used solely to balance baseline characteristics, adjust for loss to follow-up or adjust for non-compliance to randomised treatment. Studies where the exposure could not be assigned were also excluded.

RESULTS

There were 25 papers identified. Nearly half the papers (11/25) estimated the causal effect of concomitant medication on outcome. The remainder were concerned with post-randomisation treatment regimens (sequential treatments, n =5 ), effects of treatment timing (n = 2) and treatment dosing or duration (n = 7). Examples were found in cardiovascular disease (n = 5), HIV (n = 7), cancer (n = 6), mental health (n = 4), paediatrics (n = 2) and transfusion medicine (n = 1). The most common method implemented was a marginal structural model with inverse probability of treatment weighting.

CONCLUSIONS

Examples of studies which exploit RCT data to address non-randomised questions using causal inference methodology remain relatively limited, despite the growth in methodological development and increasing utilisation in observational studies. Further efforts may be needed to promote use of causal methods to address additional clinical questions within RCTs to maximise their value.

Does being physically active prevent future disability in older people? Attenuated effects when taking time-dependent confounders into account

Background

Causal experimental evidence that physical activity prevents disability in older people is sparse. Being physically active has nonetheless been shown to be associated with disability-free survival in observational studies. Observational studies are, however, prone to bias introduced by time-dependent confounding. Time-dependent confounding occurs when an exposure (e.g. being physically active at some time-point) potentially affects the future status of a confounder (such as depression sometime later), and both variables have an effect on latter outcome (i.e. disability). “Conventional” analysis with e.g. Cox-regression is the mainstay when analyzing longitudinal observational studies. Unfortunately, it does not provide unbiased estimates in the presence of time-dependent confounding. Marginal structural models (MSM) – a relatively new class of causal models – have the potential to adequately account for time-dependent confounding.

Here we analyze the effect of older people being physically active on disability, in a large long-term observational study. We address time-dependent confounding by using marginal structural models and provide a non-technical practical demonstration of how to implement this type of modeling.

Methods

Data is from 639 elderly individuals ascertained in the European multi-center Leukoaraiosis and Disability study (LADIS), followed-up yearly over a period of three years.

We estimated the effect of self-reported physical activity on the probability to transit to instrumental disability in the presence of a large set of potential confounders.

We compare the results of “conventional” modeling approaches to those estimated using marginal structural models, highlighting discrepancies.

Results

A “conventional” Cox-regression-like adjustment for salient baseline confounders signals a significant risk reduction under physical activity for later instrumental disability (OR 0.62, 95% CI 0.44–0.90). However, given MSM estimation, the effect is attenuated towards null (OR 1.00, 95% CI 0.57–1.76).

Conclusions

Contrary to most reports, we did not find that physical activity in older people prevents future instrumental disability, when taking time-dependent confounding into account. This result may be due to the characteristics our particular study population. It is, however, also conceivable that previous evidence neglected the effect of this type of bias.

We suggest that analysts of longitudinal observational studies consider marginal structural models as a further modeling approach.

Electronic supplementary material

The online version of this article (10.1186/s12877-017-0657-3) contains supplementary material, which is available to authorized users.

Performance of the marginal structural cox model for estimating individual and joined effects of treatments given in combination

BACKGROUND

The Marginal Structural Cox Model (Cox-MSM), an alternative approach to handle time-dependent confounder, was introduced for survival analysis and applied to estimate the joint causal effect of two time-dependent nonrandomized treatments on survival among HIV-positive subjects. Nevertheless, Cox-MSM performance in the case of multiple treatments has not been fully explored under different degree of time-dependent confounding for treatments or in case of interaction between treatments. We aimed to evaluate and compare the performance of the marginal structural Cox model (Cox-MSM) to the standard Cox model in estimating the treatment effect in the case of multiple treatments under different scenarios of time-dependent confounding and when an interaction between treatment effects is present.

METHODS

We specified a Cox-MSM with two treatments including an interaction term for situations where an adverse event might be caused by two treatments taken simultaneously but not by each treatment taken alone. We simulated longitudinal data with two treatments and a time-dependent confounder affected by one or the two treatments. To fit the Cox-MSM, we used the inverse probability weighting method. We illustrated the method to evaluate the specific effect of protease inhibitors combined (or not) to other antiretroviral medications on the anal cancer risk in HIV-infected individuals, with CD4 cell count as time-dependent confounder.

RESULTS

Overall, Cox-MSM performed better than the standard Cox model. Furthermore, we showed that estimates were unbiased when an interaction term was included in the model.

CONCLUSION

Cox-MSM may be used for accurately estimating causal individual and joined treatment effects from a combination therapy in presence of time-dependent confounding provided that an interaction term is estimated.

Social interaction and cognitive decline: Results of a 7-year community intervention

INTRODUCTION

There are few intervention studies that demonstrated linking social participation to lower risk of cognitive decline. We examined prospectively the protective effect of a community intervention program promoting social participation on the incidence of cognitive disability.

METHODS

The baseline was established in a survey of community-dwelling older people aged 65 years old or more in July 2006 (2793 respondents, response rate 48.5%). The setting was Taketoyo town in Japan, where municipal authorities launched an intervention that was based on the establishment of community-based centers called "salons," where the town's senior residents could congregate and participate in social activities, ranging from arts and crafts, games, and interactive activities with preschool children. Three salons were established in May 2010, and a total of 10 salons were in operation by 2013. We recorded the frequency of salon participation among survey respondents till 2013 and conducted two follow-up surveys (in 2010 and 2013) to collect information about health status and behaviors. The onset of cognitive disability was followed from May 2007 to January 2014. We used the marginal structural models to evaluate the effect of program.

RESULTS

The range of prevalence of cognitive disability was from 0.2% to 2.5% during the observation period. The proportion of respondents who participates to salons increased over time to about 11.7%. The frequency of salon participation was protectively associated with cognitive decline, even after adjusting for time-dependent covariates and attrition (odds ratio = 0.73, 95% confidence interval: 0.54-0.99).

DISCUSSION

Our study suggests that operating community salons that encourage social interactions, light physical activity, and cognitive activities among older participants may be effective for preventing cognitive decline. In future studies, we need to understand what sorts of activities (e.g., those involving light physical activity vs. purely intellectual activities) are most effective in maintaining cognitive function.

A graphical perspective of marginal structural models: An application for the estimation of the effect of physical activity on blood pressure

Estimating causal effects requires important prior subject-matter knowledge and, sometimes, sophisticated statistical tools. The latter is especially true when targeting the causal effect of a time-varying exposure in a longitudinal study. Marginal structural models are a relatively new class of causal models that effectively deal with the estimation of the effects of time-varying exposures. Marginal structural models have traditionally been embedded in the counterfactual framework to causal inference. In this paper, we use the causal graph framework to enhance the implementation of marginal structural models. We illustrate our approach using data from a prospective cohort study, the Honolulu Heart Program. These data consist of 8006 men at baseline. To illustrate our approach, we focused on the estimation of the causal effect of physical activity on blood pressure, which were measured at three time points. First, a causal graph is built to encompass prior knowledge. This graph is then validated and improved utilizing structural equation models. We estimated the aforementioned causal effect using marginal structural models for repeated measures and guided the implementation of the models with the causal graph. By employing the causal graph framework, we also show the validity of fitting conditional marginal structural models for repeated measures in the context implied by our data.

Causal inference and longitudinal data: a case study of religion and mental health

PURPOSE

We provide an introduction to causal inference with longitudinal data and discuss the complexities of analysis and interpretation when exposures can vary over time.

METHODS

We consider what types of causal questions can be addressed with the standard regression-based analyses and what types of covariate control and control for the prior values of outcome and exposure must be made to reason about causal effects. We also consider newer classes of causal models, including marginal structural models, that can assess questions of the joint effects of time-varying exposures and can take into account feedback between the exposure and outcome over time. Such feedback renders cross-sectional data ineffective for drawing inferences about causation.

RESULTS

The challenges are illustrated by analyses concerning potential effects of religious service attendance on depression, in which there may in fact be effects in both directions with service attendance preventing the subsequent depression, but depression itself leading to lower levels of the subsequent religious service attendance.

CONCLUSIONS

Longitudinal designs, with careful control for prior exposures, outcomes, and confounders, and suitable methodology, will strengthen research on mental health, religion and health, and in the biomedical and social sciences generally.

On Bayesian estimation of marginal structural models

The purpose of inverse probability of treatment (IPT) weighting in estimation of marginal treatment effects is to construct a pseudo-population without imbalances in measured covariates, thus removing the effects of confounding and informative censoring when performing inference. In this article, we formalize the notion of such a pseudo-population as a data generating mechanism with particular characteristics, and show that this leads to a natural Bayesian interpretation of IPT weighted estimation. Using this interpretation, we are able to propose the first fully Bayesian procedure for estimating parameters of marginal structural models using an IPT weighting. Our approach suggests that the weights should be derived from the posterior predictive treatment assignment and censoring probabilities, answering the question of whether and how the uncertainty in the estimation of the weights should be incorporated in Bayesian inference of marginal treatment effects. The proposed approach is compared to existing methods in simulated data, and applied to an analysis of the Canadian Co-infection Cohort.

Performance of the marginal structural models under various scenarios of incomplete marker's values: a simulation study

Marginal structural models (MSMs) have been proposed for estimating a treatment's effect, in the presence of time-dependent confounding. We aimed to evaluate the performance of the Cox MSM in the presence of missing data and to explore methods to adjust for missingness. We simulated data with a continuous time-dependent confounder and a binary treatment. We explored two classes of missing data: (i) missed visits, which resemble clinical cohort studies; (ii) missing confounder's values, which correspond to interval cohort studies. Missing data were generated under various mechanisms. In the first class, the source of the bias was the extreme treatment weights. Truncation or normalization improved estimation. Therefore, particular attention must be paid to the distribution of weights, and truncation or normalization should be applied if extreme weights are noticed. In the second case, bias was due to the misspecification of the treatment model. Last observation carried forward (LOCF), multiple imputation (MI), and inverse probability of missingness weighting (IPMW) were used to correct for the missingness. We found that alternatives, especially the IPMW method, perform better than the classic LOCF method. Nevertheless, in situations with high marker's variance and rarely recorded measurements none of the examined method adequately corrected the bias.

Association of long-duration breastfeeding and dental caries estimated with marginal structural models

PURPOSE

To estimate the association between breastfeeding 24 months or beyond and severe early childhood caries (S-ECC).

METHODS

Within a birth cohort (n = 715) from low-income families in Porto Alegre, Brazil, the age 38-month prevalence of S-ECC (≥4 affected tooth surfaces or ≥1 affected maxillary anterior teeth) was compared over breastfeeding duration categories using marginal structural models to account for time-dependent confounding by other feeding habits and child growth. Additional analyses assessed whether daily breastfeeding frequency modified the association of breastfeeding duration and S-ECC. Multiple imputation and censoring weights were used to address incomplete covariate information and missing outcomes, respectively. Confidence intervals (CIs) were estimated using bootstrap resampling.

RESULTS

Breastfeeding 24 months or beyond was associated with the highest adjusted population-average S-ECC prevalence (0.45; 95% CI, 0.36 to 0.54) compared with breastfeeding less than 6 months (0.22; 95% CI, 0.15 to 0.28), 6-11 months (0.38; 95% CI, 0.25 to 0.53), or 12-23 months (0.39; 95% CI, 0.20 to 0.56). High-frequency breastfeeding enhanced the association between long-duration breastfeeding and caries (excess prevalence due to interaction: 0.13; 80% CI, -0.03 to 0.30).

CONCLUSIONS

In this population, breastfeeding 24 months or beyond, particularly if frequent, was associated with S-ECC. Dental health should be one consideration, among many, in evaluating health outcomes associated with breastfeeding 24 months or beyond.

Does academic achievement during childhood and adolescence benefit later health?

PURPOSE

Educational disparities in health persist after adjustment for income and occupation, suggesting that other purely cognitive and psychosocial mechanisms may be involved. Unlike occupation- or income-mediated effects, effects of cognitive and psychosocial gains-as reflected in academic achievement-may be apparent even before schooling is completed.

METHODS

We used data spanning 10 years on a national U.S. cohort of 2546 children aged 3-14 years at baseline to estimate the effects of academic achievement, measured by standardized tests of cognitive achievement, on future health. We used marginal structural models to address potential mutual influence of achievement and health on each other over time.

RESULTS

One SD higher academic achievement 1997-2002 was associated with a lower prevalence of poorer health status in 2007 in girls (prevalence ratio = 0.87 [(95% confidence interval) 0.78-0.97]) but not in boys (prevalence ratio = 0.96 [0.86-1.08]). Higher achievement was also weakly associated with lower body mass index and less psychological distress among girls only.

CONCLUSIONS

Academic achievement may benefit future health but a number of questions remain unanswered, including reasons for the gender differences and how academic achievement-related health disparities may progress over the life course and interact with other social determinants of health.

A cross-validation deletion-substitution-addition model selection algorithm: Application to marginal structural models

The cross-validation deletion-substitution-addition (cvDSA) algorithm is based on data-adaptive estimation methodology to select and estimate marginal structural models (MSMs) for point treatment studies as well as models for conditional means where the outcome is continuous or binary. The algorithm builds and selects models based on user-defined criteria for model selection, and utilizes a loss function-based estimation procedure to distinguish between different model fits. In addition, the algorithm selects models based on cross-validation methodology to avoid "over-fitting" data. The cvDSA routine is an R software package available for download. An alternative R-package (DSA) based on the same principles as the cvDSA routine (i.e., cross-validation, loss function), but one that is faster and with additional refinements for selection and estimation of conditional means, is also available for download. Analyses of real and simulated data were conducted to demonstrate the use of these algorithms, and to compare MSMs where the causal effects were assumed (i.e., investigator-defined), with MSMs selected by the cvDSA. The package was used also to select models for the nuisance parameter (treatment) model to estimate the MSM parameters with inverse-probability of treatment weight (IPTW) estimation. Other estimation procedures (i.e., G-computation and double robust IPTW) are available also with the package.