A Cautionary Note on Using Propensity Score Calibration to Control for Unmeasured Confounding Bias When the Surrogacy Assumption Is Absent.

Conventional propensity score methods encounter challenges when unmeasured confounding is present, as it becomes impossible to accurately estimate the gold-standard propensity score when data on certain confounders are unavailable. Propensity score calibration (PSC) addresses this issue by constructing a surrogate for the gold-standard propensity score under the surrogacy assumption. This assumption posits that the error-prone propensity score, based on observed confounders, is independent of the outcome when conditioned on the gold-standard propensity score and the exposure. However, this assumption implies that confounders cannot directly impact the outcome and that their effects on the outcome are solely mediated through the propensity score. This raises concerns regarding the applicability of PSC in practical settings where confounders can directly affect the outcome. While PSC aims to target a conditional treatment effect by conditioning on a subject's unobservable propensity score, the causal interest in the latter case lies in a conditional treatment effect conditioned on a subject's baseline characteristics. Our analysis reveals that PSC is generally biased unless the effects of confounders on the outcome and treatment are proportional to each other. Furthermore, we identify 2 sources of bias: 1) the noncollapsibility of effect measures, such as the odds ratio or hazard ratio and 2) residual confounding, as the calibrated propensity score may not possess the properties of a valid propensity score.

External adjustment of unmeasured confounders in a case-control study of benzodiazepine use and cancer risk.

AIMS: Previous studies have reported diverging results on the association between benzodiazepine use and cancer risk. METHODS: We investigated this association in a matched case-control study including incident cancer cases during 2002-2009 in the Danish Cancer Registry (n = 94 923) and age- and sex-matched (1:8) population controls (n = 759 334). Long-term benzodiazepine use was defined as ≥500 defined daily doses 1-5 years prior to the index date. We implemented propensity score (PS) calibration using external information on confounders available from a survey of the Danish population. Two PSs were used: The error-prone PS using register-based confounders and the calibrated PS based on both register- and survey-based confounders, retrieved from the Health Interview Survey. RESULTS: Register-based data showed that cancer cases had more diagnoses, higher comorbidity score and more co-medication then population controls. Survey-based data showed lower self-rated health, more self-reported diseases, and more smokers as well as subjects with sedentary lifestyle among benzodiazepine users. By PS calibration, the odds ratio for cancer overall associated with benzodiazepine use decreased from 1.16 to 1.09 (95% confidence interval 1.00-1.19) and for smoking-related cancers from 1.20 to 1.10 (95% confidence interval 1.00-1.21). CONCLUSION: We conclude that the increased risk observed in the solely register-based study could partly be attributed to unmeasured confounding.

Prenatal triptan exposure and parent-reported early childhood neurodevelopmental outcomes: an application of propensity score calibration to adjust for unmeasured confounding by migraine severity.

PURPOSE: Triptan medications are serotonin agonists used to treat migraine, a chronic pain condition highly prevalent in women of reproductive age. Data on the safety of triptans during pregnancy are scant. We sought to quantify the association of prenatal triptan exposure on neurodevelopment in 3-year-old children. METHODS: Using data from the Norwegian Mother and Child Cohort Study, we used propensity score matching to examine associations between prenatal triptan exposure and psychomotor function, communication, and temperament. We used an external validation study to perform propensity calibration to adjust effect estimates for confounders unmeasured in the main study (migraine severity, type, and maternal attitudes towards medication use). RESULTS: We identified 4204 women who reported migraine headache at baseline, of which 375 (8.9%) reported using a triptan greater than or equal to once during pregnancy. Children with prenatal triptan exposure had 1.37-fold greater unadjusted odds of fine motor problems (95% confidence interval (CI): 1.06-1.77), which decreased after propensity score matching (odds ratio (OR): 1.29, 95%CI 0.97-1.73) and was further attenuated after calibration (OR: 1.25, 95%CI 0.89-1.74). We observed no increased risk for gross motor or communication problems, and no differences in temperament. Adjustment for migraine severity using propensity score calibration had a moderate impact on effect estimates, with percent changes ranging from 2.4% to 50%. CONCLUSIONS: Prenatal triptan exposure was not associated with psychomotor function, communication problems, or temperament in 3-year-old children. Adjustment for migraine severity reduced effect estimates and should be considered in future studies of the safety of triptans during pregnancy. Copyright © 2015 John Wiley & Sons, Ltd.

Comparative safety of antipsychotic medications in elderly stroke survivors: A nationwide claim data and stroke registry linkage cohort study.

BACKGROUND: Antipsychotics remain the first choice of treatment for post-stroke psychosis, despite an increased risk of mortality reported in elderly patients. We aimed to compare the mortality risk among antipsychotics in elderly patients with stroke using the stroke registry for external adjustment. METHODS: We conducted a retrospective cohort study to identify patients aged above 65 years who were admitted for stroke in the National Health Insurance Database (NHID) from 2002 to 2014. The first date of antipsychotic use after the stroke hospitalization was defined as the index date. Covariates including diseases, medications and external information on smoking, BMI, stroke severity and disability, that were unavailable in the NHID were obtained from the linked Multicenter Stroke Registry (MSR) and used for propensity score calibration (PSC). The main outcome was one-year all-cause mortality. RESULTS: Stroke patients in the NHID prescribed with haloperidol, quetiapine and risperidone numbered 22,235, 28,702 and 8 663, respectively. In the PSC-adjusted analyses, haloperidol [adjusted hazard ratio (aHR) = 1.22; 95% CI 1.18-1.27] and risperidone (aHR = 1.31; 95% CI 1.24-1.38) users had a higher mortality risk than quetiapine users. When the dosage was higher than 0.5 defined daily dose (DDD), haloperidol and risperidone users had a significant mortality risk as compared with those taking a lower dose. CONCLUSIONS: In post-stroke elderly patients, quetiapine would pose less mortality risk than risperidone and haloperidol at doses higher than 0.5 DDD. When haloperidol or risperidone is indicated, starting with a lower dose is suggested to avoid excess risk.

Mortality after Transplantation for Hepatocellular Carcinoma: A Study from the European Liver Transplant Registry.

Background and Aims: Prognosis after liver transplantation differs between hepatocellular carcinoma (HCC) arising in cirrhotic and non-cirrhotic livers and aetiology is poorly understood. The aim was to investigate differences in mortality after liver transplantation between these patients. Methods: We included patients from the European Liver Transplant Registry transplanted due to HCC from 1990 to November 2016 and compared cirrhotic and non-cirrhotic patients using propensity score (PS) calibration of Cox regression estimates to adjust for unmeasured confounding. Results: We included 22,787 patients, of whom 96.5% had cirrhosis. In the unadjusted analysis, non-cirrhotic patients had an increased risk of overall mortality with a hazard ratio (HR) of 1.37 (95% confidence interval [CI] 1.23-1.52). However, the HR approached unity with increasing adjustment and was 1.11 (95% CI 0.99-1.25) when adjusted for unmeasured confounding. Unadjusted, non-cirrhotic patients had an increased risk of HCC-specific mortality (HR 2.62, 95% CI 2.21-3.12). After adjustment for unmeasured confounding, the risk remained significantly increased (HR 1.62, 95% CI 1.31-2.00). Conclusions: Using PS calibration, we showed that HCC in non-cirrhotic liver has similar overall mortality, but higher HCC-specific mortality. This may be a result of a more aggressive cancer form in the non-cirrhotic liver as higher mortality could not be explained by tumour characteristics or other prognostic variables.

Misclassification in administrative claims data: quantifying the impact on treatment effect estimates.

Misclassification is present in nearly every epidemiologic study, yet is rarely quantified in analysis in favor of a focus on random error. In this review, we discuss past and present wisdom on misclassification and what measures should be taken to quantify this influential bias, with a focus on bias in pharmacoepidemiologic studies. To date, pharmacoepidemiology primarily utilizes data obtained from administrative claims, a rich source of prescription data but susceptible to bias from unobservable factors including medication sample use, medications filled but not taken, health conditions that are not reported in the administrative billing data, and inadequate capture of confounders. Due to the increasing focus on comparative effectiveness research, we provide a discussion of misclassification in the context of an active comparator, including a demonstration of treatment effects biased away from the null in the presence of nondifferential misclassification. Finally, we highlight recently developed methods to quantify bias and offer these methods as potential options for strengthening the validity and quantifying uncertainty of results obtained from pharmacoepidemiologic research.