A double robust approach to causal effects in case-control studies

The Targeted Maximum Likelihood estimation to estimate the causal effects of the previous tuberculosis treatment in Multidrug-resistant tuberculosis in Sudan

INTRODUCTION

This study used Targeted Maximum Likelihood Estimation (TMLE) as a double robust method to estimate the causal effect of previous tuberculosis treatment history on the occurrence of multidrug-resistant tuberculosis (MDR-TB). TMLE is a method to estimate the marginal statistical parameters in case-control study design. The aim of this study was to estimate the causal effect of the previous tuberculosis treatment on the occurrence of MDR-TB using TMLE in Sudan.

METHOD

A case-control study design combined with TMLE was used to estimate parameters. Cases were MDR-TB patients and controls were and patients who cured from tuberculosis. The history of previous TB treatment was considered the main exposure, and MDR-TB as an outcome. A designed questionnaire was used to collect a set of covariates including age, time to reach a health facility, number of times stopping treatment, gender, education level, and contact with MDR-TB cases. TMLE method was used to estimate the causal association of parameters. Statistical analysis was carried out with ltmle package in R-software. Result presented in graph and tables.

RESULTS

A total number of 430 cases and 860 controls were included in this study. The estimated risk difference of the previous tuberculosis treatment was (0.189, 95% CI; 0.161, 0.218) with SE 0.014, and p-value (<0.001). In addition, the estimated risk ratio was (16.1, 95% CI; 12.932, 20.001) with SE = 0.014 and p-value (<0.001).

CONCLUSION

Our findings indicated that previous tuberculosis treatment history was determine as a risk factor for MDR-TB in Sudan. Also, TMLE method can be used to estimate the risk difference and the risk ratio in a case-control study design.

Evaluating the heterogeneous effect of a modifiable risk factor on suicide: The case of vitamin D deficiency

OBJECTIVES

To illustrate the use of machine learning methods to search for heterogeneous effects of a target modifiable risk factor on suicide in observational studies. The illustration focuses on secondary analysis of a matched case-control study of vitamin D deficiency predicting subsequent suicide.

METHODS

We describe a variety of machine learning methods to search for prescriptive predictors; that is, predictors of significant variation in the association between a target risk factor and subsequent suicide. In each case, the purpose is to evaluate the potential value of selective intervention on the target risk factor to prevent the outcome based on the provisional assumption that the target risk factor is causal. The approaches illustrated include risk modeling based on the super learner ensemble machine learning method, Least Absolute Shrinkage and Selection Operator (Lasso) penalized regression, and the causal forest algorithm.

RESULTS

The logic of estimating heterogeneous intervention effects is exposited along with the illustration of some widely used methods for implementing this logic.

CONCLUSIONS

In addition to describing best practices in using the machine learning methods considered here, we close with a discussion of broader design and analysis issues in planning an observational study to investigate heterogeneous effects of a modifiable risk factor.

Relation of Pre-Stroke Aspirin Use With Cerebral Infarct Volume and Functional Outcomes

Objective

We investigated (1) the associations of pre‐stroke aspirin use with thrombus burden, infarct volume, hemorrhagic transformation, early neurological deterioration (END), and functional outcome, and (2) whether stroke subtypes modify these associations in first‐ever ischemic stroke.

Methods

This multicenter magnetic resonance imaging (MRI)‐based study included 5,700 consecutive patients with acute first‐ever ischemic stroke, who did not undergo intravenous thrombolysis or endovascular thrombectomy, from May 2011 through February 2014. Propensity score‐based augmented inverse probability weighting was performed to estimate adjusted effects of pre‐stroke aspirin use.

Results

The mean age was 67 years (41% women), and 15.9% (n = 907) were taking aspirin before stroke. Pre‐stroke aspirin use (vs nonuse) was significantly related to a reduced infarct volume (by 30%), particularly in large artery atherosclerosis stroke (by 45%). In cardioembolic stroke, pre‐stroke aspirin use was associated with a ~50% lower incidence of END (adjusted difference = −5.4%, 95% confidence interval [CI] = −8.9 to −1.9). Thus, pre‐stroke aspirin use was associated with ~30% higher likelihood of favorable outcome (3‐month modified Rankin Scale score < 3), particularly in large artery atherosclerosis stroke and cardioembolic stroke (adjusted difference = 7.2%, 95% CI = 1.8 to 12.5 and adjusted difference = 6.4%, 95% CI = 1.7 to 11.1, respectively). Pre‐stroke aspirin use (vs nonuse) was associated with 85% less frequent cerebral thrombus‐related susceptibility vessel sign (SVS) in large artery atherosclerosis stroke (adjusted difference = −1.4%, 95% CI = −2.1 to −0.8, p < 0.001) and was associated with ~40% lower SVS volumes, particularly in cardioembolic stroke (adjusted difference = −0.16 cm³, 95% CI = −0.29 to −0.02, p = 0.03). Moreover, pre‐stroke aspirin use was not significantly associated with hemorrhagic transformation (adjusted difference = −1.1%, p = 0.09).

Interpretation

Pre‐stroke aspirin use associates with improved functional independence in patients with first‐ever ischemic large arterial stroke by reducing infarct volume and/or END, likely by decreasing thrombus burden, without increased risk of hemorrhagic transformation. ANN NEUROL 2021;90:763–776

Estimating the Marginal Causal Effect and Potential Impact of Waterpipe Smoking on Risk of Multiple Sclerosis Using the Targeted Maximum Likelihood Estimation Method: A Large, Population-Based Incident Case-Control Study

There are few if any reports regarding the role of lifetime waterpipe smoking in the etiology of multiple sclerosis (MS). In a population-based incident case-control study conducted in Tehran, Iran, we investigated the association between waterpipe smoking and MS, adjusted for confounders. Cases (n = 547) were patients aged 15-50 years identified from the Iranian Multiple Sclerosis Society between 2013 and 2015. Population-based controls (n = 1,057) were persons aged 15-50 years recruited through random digit telephone dialing. A doubly robust estimation method, the targeted maximum likelihood estimator (TMLE), was used to estimate the marginal risk ratio and odds ratio for the association between waterpipe smoking and MS. The estimated risk ratio and odds ratio were both 1.70 (95% confidence interval: 1.34, 2.17). The population attributable fraction was 21.4% (95% confidence interval: 4.0, 38.8). Subject to the limitations of case-control studies in interpreting associations causally, these results suggest that waterpipe use, or strongly related but undetermined factors, increases the risk of MS. Further epidemiologic studies, including nested case-control studies, are needed to confirm these findings.

The causal effect and impact of reproductive factors on breast cancer using super learner and targeted maximum likelihood estimation: a case-control study in Fars Province, Iran

OBJECTIVES

The relationship between reproductive factors and breast cancer (BC) risk has been investigated in previous studies. Considering the discrepancies in the results, the aim of this study was to estimate the causal effect of reproductive factors on BC risk in a case-control study using the double robust approach of targeted maximum likelihood estimation.

METHODS

This is a causal reanalysis of a case-control study done between 2005 and 2008 in Shiraz, Iran, in which 787 confirmed BC cases and 928 controls were enrolled. Targeted maximum likelihood estimation along with super Learner were used to analyze the data, and risk ratio (RR), risk difference (RD), andpopulation attributable fraction (PAF) were reported.

RESULTS

Our findings did not support parity and age at the first pregnancy as risk factors for BC. The risk of BC was higher among postmenopausal women (RR = 3.3, 95% confidence interval (CI) = (2.3, 4.6)), women with the age at first marriage ≥20 years (RR = 1.6, 95% CI = (1.3, 2.1)), and the history of oral contraceptive (OC) use (RR = 1.6, 95% CI = (1.3, 2.1)) or breastfeeding duration ≤60 months (RR = 1.8, 95% CI = (1.3, 2.5)). The PAF for menopause status, breastfeeding duration, and OC use were 40.3% (95% CI = 39.5, 40.6), 27.3% (95% CI = 23.1, 30.8) and 24.4% (95% CI = 10.5, 35.5), respectively.

CONCLUSIONS

Postmenopausal women, and women with a higher age at first marriage, shorter duration of breastfeeding, and history of OC use are at the higher risk of BC.

Understanding Marginal Structural Models for Time-Varying Exposures: Pitfalls and Tips

Epidemiologists are increasingly encountering complex longitudinal data, in which exposures and their confounders vary during follow-up. When a prior exposure affects the confounders of the subsequent exposures, estimating the effects of the time-varying exposures requires special statistical techniques, possibly with structural (ie, counterfactual) models for targeted effects, even if all confounders are accurately measured. Among the methods used to estimate such effects, which can be cast as a marginal structural model in a straightforward way, one popular approach is inverse probability weighting. Despite the seemingly intuitive theory and easy-to-implement software, misunderstandings (or "pitfalls") remain. For example, one may mistakenly equate marginal structural models with inverse probability weighting, failing to distinguish a marginal structural model encoding the causal parameters of interest from a nuisance model for exposure probability, and thereby failing to separate the problems of variable selection and model specification for these distinct models. Assuming the causal parameters of interest are identified given the study design and measurements, we provide a step-by-step illustration of generalized computation of standardization (called the g-formula) and inverse probability weighting, as well as the specification of marginal structural models, particularly for time-varying exposures. We use a novel hypothetical example, which allows us access to typically hidden potential outcomes. This illustration provides steppingstones (or "tips") to understand more concretely the estimation of the effects of complex time-varying exposures.

Dental alterations on panoramic radiographs of patients with osteogenesis imperfecta in relation to clinical diagnosis, severity, and bisphosphonate regimen aspects: a STROBE-compliant case-control study

OBJECTIVE

This study aimed to assess the prevalence of dental findings on panoramic radiographs (PRs) of patients with osteogenesis imperfecta (OI) and correlate these results with epidemiologic and medical data.

STUDY DESIGN

A case-control study was conducted with 24 patients with OI and 48 sex- and age-matched controls. Demographic, clinical, and bisphosphonate regimen-related data were recorded. The outcome variables were the presence or absence of dental alterations in PRs. Mann-Whitney U test, Pearson's χ² test, and multinomial logistic regression analysis (95% confidence interval) were used (significance level of 5%).

RESULTS

OI type 4 demonstrated a high prevalence (62.5%), followed by type 1 (37.5%). With regard to prevalence associated with severity, the moderate form was the most prevalent (P = .028). The mean time of intravenous pamidronate regimen was 6.6 ± 4.4 years. Dentinogenesis imperfecta was observed in 75% of patients with OI, and this group showed a high prevalence of dental abnormalities in comparison with controls (P < .05). Bisphosphonate therapy was associated with ectopic teeth (P = .007) and tooth impaction (P = .033). Pulp obliteration was significant with bisphosphonate treatment over a period of 7 years (P = .026).

CONCLUSIONS

This study found a significant prevalence of dental alterations in patients with OI, and certain alterations were associated with bisphosphonate therapy, indicating its influence on the dentin-related physiopathology.

Estimating the Marginal Causal Effect of Fish Consumption during Adolescence on Multiple Sclerosis: A Population-Based Incident Case-Control Study

BACKGROUND

Adolescence is considered as a critical time period in multiple sclerosis (MS) etiology. Nonetheless, there are insufficient reports regarding the potential role of fresh and canned fish consumptions during adolescence in MS etiology. The authors investigated the association between fresh and canned fish consumptions and MS.

METHODS

This was a population-based incident case-control study conducted in Tehran. Cases (n = 547) identified from Iranian Multiple Sclerosis Society between August 7, 2013, and November 17, 2015 were included in the study. Population-based controls (n = 1,057) were recruited by random digit telephone dialing without any matching. Inverse-probability-of-treatment weighing (IPTW) using 2 sets of propensity scores and model-based standardization were used to separately estimate the marginal odds ratio between fresh and canned fish consumptions in adolescence and MS.

RESULTS

The marginal OR for fresh fish was 0.72 (95% CI 0.58-0.90; p = 0.005) in both IPTW analyses. Similarly, the marginal OR for canned fish consumption was 0.75 (95% CI 0.60-0.95; p = 0.014).The model-based standardized OR was 0.72 (95% CI 0.58-0.91; p = 0.008) for fresh and 0.73 (95% CI 0.59-0.94; p = 0.006) for canned fish consumption in adolescence.

DISCUSSION

Subject to limitation of case-control studies in interpreting associations causally, this study suggests that both fresh and canned fish consumptions in adolescence can decrease the risk of MS.

Anticholinergic drugs and risk of dementia: case-control study

OBJECTIVES

To estimate the association between the duration and level of exposure to different classes of anticholinergic drugs and subsequent incident dementia.

DESIGN

Case-control study.

SETTING

General practices in the UK contributing to the Clinical Practice Research Datalink.

PARTICIPANTS

40 770 patients aged 65-99 with a diagnosis of dementia between April 2006 and July 2015, and 283 933 controls without dementia.

INTERVENTIONS

Daily defined doses of anticholinergic drugs coded using the Anticholinergic Cognitive Burden (ACB) scale, in total and grouped by subclass, prescribed 4-20 years before a diagnosis of dementia.

MAIN OUTCOME MEASURES

Odds ratios for incident dementia, adjusted for a range of demographic and health related covariates.

RESULTS

14 453 (35%) cases and 86 403 (30%) controls were prescribed at least one anticholinergic drug with an ACB score of 3 (definite anticholinergic activity) during the exposure period. The adjusted odds ratio for any anticholinergic drug with an ACB score of 3 was 1.11 (95% confidence interval 1.08 to 1.14). Dementia was associated with an increasing average ACB score. When considered by drug class, gastrointestinal drugs with an ACB score of 3 were not distinctively linked to dementia. The risk of dementia increased with greater exposure for antidepressant, urological, and antiparkinson drugs with an ACB score of 3. This result was also observed for exposure 15-20 years before a diagnosis.

CONCLUSIONS

A robust association between some classes of anticholinergic drugs and future dementia incidence was observed. This could be caused by a class specific effect, or by drugs being used for very early symptoms of dementia. Future research should examine anticholinergic drug classes as opposed to anticholinergic effects intrinsically or summing scales for anticholinergic exposure.

TRIAL REGISTRATION

Registered to the European Union electronic Register of Post-Authorisation Studies EUPAS8705.

Estimation of the marginal effect of regular drug use on multiple sclerosis in the Iranian population

There are only few reports regarding the role of lifetime drug or substance use in multiple sclerosis (MS) etiology. In this study, we investigated the potential effect of drug or substance exposure on the onset of MS diagnosis. We conducted a population-based incident case control study in Tehran. Cases (n = 547) were 15-50 years old persons with MS identified from the Iranian Multiple Sclerosis Society (IMSS) register during August 7, 2013, and November 17, 2015. Population-based controls (n = 1057) were 15-50 years old and were recruited by random digit telephone dialing. Inverse-probability-of-treatment weighing (IPTW) using two sets of propensity scores (PSs) was used to estimate marginal incidence odds ratios (ORs) for MS contrasting pre-specified substance use. The estimated marginal OR was 6.03 (95% confidence interval: 3.54;10.3, using trimmed weights at the 95th percentile of the stabilized weight distribution) in both IPTW analyses comparing lifetime substance use (opioids, cannabis, inhalants, hallucinogens and stimulants) for at least one time monthly during a six-months or longer period vs. no such history of drug use. Subject to limitation of causal claims based on case-control studies, this study suggests that monthly drug or substance use for a period of at least six consecutive months, may increase the risk of MS by factor 3.5 or higher.

Targeted Estimation of Marginal Absolute and Relative Associations in Case-Control Data: An Application in Social Epidemiology

BACKGROUND

Case-control studies are useful for rare outcomes, but typical analyses limit investigators to parametric estimation of conditional odds ratios. Several methods exist for obtaining marginal risk differences and risk ratios in a case-control setting, including a recently described semiparametric targeted approach optimized for rare outcomes.

METHODS

Using case-control data from a study of neighborhood poverty and very preterm birth, we demonstrate estimation of marginal risk differences and risk ratios and compare a parametric substitution estimator based on maximum likelihood estimation with targeted maximum likelihood estimation (TMLE), and a refinement of TMLE for rare outcomes that incorporates bounds on the conditional risk.

RESULTS

In this illustration, living in a neighborhood with high poverty was associated with a higher risk of very preterm birth for white women. The estimated risk differences (cases/100) were 0.6 (95% confidence interval [CI]: 0.1, 1.1) from maximum likelihood estimation, 0.5 (95% CI: -1.1, 2.1) from TMLE, and 0.5 (95% CI: 0.0, 1.0) from the rare outcomes refinement. The rare outcomes refinement, which incorporates knowledge that the conditional risk is small, produced more precise estimates than TMLE. A similar pattern was observed for the relative risk.

CONCLUSION

Absolute and relative associations estimated from case-control data using a semiparametric targeted approach allow the scientific question to determine the analysis and avoid unwarranted parametric assumptions. A rare outcomes refinement provided more precise estimates than TMLE, and thus is well suited for the study of rare outcomes.

Targeted Maximum Likelihood Estimation for Pharmacoepidemiologic Research

Supplemental Digital Content is available in the text.

Background:

Targeted maximum likelihood estimation has been proposed for estimating marginal causal effects, and is robust to misspecification of either the treatment or outcome model. However, due perhaps to its novelty, targeted maximum likelihood estimation has not been widely used in pharmacoepidemiology. The objective of this study was to demonstrate targeted maximum likelihood estimation in a pharmacoepidemiological study with a high-dimensional covariate space, to incorporate the use of high-dimensional propensity scores into this method, and to compare the results to those of inverse probability weighting.

Methods:

We implemented the targeted maximum likelihood estimation procedure in a single-point exposure study of the use of statins and the 1-year risk of all-cause mortality postmyocardial infarction using data from the UK Clinical Practice Research Datalink. A range of known potential confounders were considered, and empirical covariates were selected using the high-dimensional propensity scores algorithm. We estimated odds ratios using targeted maximum likelihood estimation and inverse probability weighting with a variety of covariate selection strategies.

Results:

Through a real example, we demonstrated the double robustness of targeted maximum likelihood estimation. We showed that results with this method and inverse probability weighting differed when a large number of covariates were included in the treatment model.

Conclusions:

Targeted maximum likelihood can be used in high-dimensional covariate settings. In high-dimensional covariate settings, differences in results between targeted maximum likelihood and inverse probability weighted estimation are likely due to sensitivity to (near) positivity violations. Further investigations are needed to gain better understanding of the advantages and limitations of this method in pharmacoepidemiological studies.

Matching and Imputation Methods for Risk Adjustment in the Health Insurance Marketplaces

New state-level health insurance markets, denoted Marketplaces, created under the Affordable Care Act, use risk-adjusted plan payment formulas derived from a population ineligible to participate in the Marketplaces. We develop methodology to derive a sample from the target population and to assemble information to generate improved risk-adjusted payment formulas using data from the Medical Expenditure Panel Survey and Truven MarketScan databases. Our approach requires multi-stage data selection and imputation procedures because both data sources have systemic missing data on crucial variables and arise from different populations. We present matching and imputation methods adapted to this setting. The long-term goal is to improve risk-adjustment estimation utilizing information found in Truven MarketScan data supplemented with imputed Medical Expenditure Panel Survey values.

Invited commentary: Some advantages of the relative excess risk due to interaction (RERI)--towards better estimators of additive interaction

In the accompanying commentary, Rose and van der Laan (Am J Epidemiol. 2014;179(6):663-669) criticize the relative excess risk due to interaction (RERI) measure, the use of additive interaction, and the weighting approach we developed to assess RERI with case-control data. In this commentary, we note some of the advantages of using additive measures of interaction, such as RERI, in making decisions about targeting interventions toward certain subgroups and in assessing mechanistic interaction. We discuss the relationship between Rose and van der Laan's estimator for case-control data and the one we had previously proposed. We also develop a new doubly robust estimator for determining the RERI with case-control data when the prevalence or incidence of the outcome is known.