Data-Driven Simulations to Assess the Impact of Study Imperfections in Time-to-Event Analyses.

Quantitative bias analysis (QBA) permits assessment of the expected impact of various imperfections of the available data on the results and conclusions of a particular real-world study. This article extends QBA methodology to multivariable time-to-event analyses with right-censored endpoints, possibly including time-varying exposures or covariates. The proposed approach employs data-driven simulations, which preserve important features of the data at hand while offering flexibility in controlling the parameters and assumptions that may affect the results. First, the steps required to perform data-driven simulations are described, and then two examples of real-world time-to-event analyses illustrate their implementation and the insights they may offer. The first example focuses on the omission of an important time-invariant predictor of the outcome in a prognostic study of cancer mortality, and permits separating the expected impact of confounding bias from non-collapsibility. The second example assesses how imprecise timing of an interval-censored event - ascertained only at sparse times of clinic visits - affects its estimated association with a time-varying drug exposure. The simulation results also provide a basis for comparing the performance of two alternative strategies for imputing the unknown event times in this setting. The R scripts that permit the reproduction of our examples are provided.

Immunosuppression and cancer risk in kidney transplant recipients: A retrospective cohort study.

We assessed whether contemporary immunosuppression agents were associated with cancer among kidney transplant recipients (KTR), and if this association varied by age and sex. We studied a retrospective province-wide cohort of primary KTR (1997-2016). Employing multivariable Cox models, we estimated associations of cumulative doses of prednisone, mycophenolate and tacrolimus administered over the past 10 years, lagged by 2 years, with the incidence of primary malignant neoplasms (PMN). We assessed interactions with age and sex. To assess the impact of exposure recency, we used weighted cumulative exposure (WCE) modeling. Among 1064 KTR, 108 (10.2%) developed PMN over median follow-up of 73 months (interquartile range: 32-120). Adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) of 0.96 (0.64-1.43), 1.34 (0.96-1.86), and 1.06 (0.88-1.29) were estimated for cumulative daily doses of prednisone (5 mg), mycophenolate (1000 mg), and tacrolimus (2 mg) administered continuously over the past 10 years, respectively. PMN risk associated with cumulative tacrolimus exposure was modified by age (interaction p = .035) and was more pronounced in 15-year and 30-year-old KTR (aHRs of 1.57 [1.08-2.28] and 1.31 [1.03-1.66], respectively) in comparison to older KTR. PMN risk increase associated with higher cumulative mycophenolate dose was more pronounced in females (aHR = 1.86 [1.15-3.00]) than in males (aHR = 1.16 [0.74-1.81]; interaction p = .131). WCE analyses suggested increased PMN risk the higher the mycophenolate doses taken 5-10 years ago. A trend toward increased PMN risk with long-term mycophenolate exposure, particularly in females, and more pronounced risk with long-term tacrolimus exposure in younger KTR, identify opportunities for tailored immunosuppression to mitigate cancer risk.

Investigating the Shape and Strength of the Relationship Between Maternal Weight Gain and Gestational Age at Delivery in Twin and Singleton Pregnancies.

Both inadequate and excessive maternal weight gain are correlated with preterm delivery in singleton pregnancies, yet this relationship has not been adequately studied in twins. We investigated the relationship between time-varying maternal weight gain and gestational age at delivery in twin pregnancies and compared it with that in singletons delivered in the same study population. We used serial weight measurements abstracted from charts for twin and singleton pregnancies delivered during 1998-2013 in Pittsburgh, Pennsylvania. Our exposure was time-varying weight gain z score, calculated using gestational age-standardized and prepregnancy body mass index-stratified twin- and singleton-specific charts, and our outcome was gestational age at delivery. Our analyses used a flexible extension of the Cox proportional hazards model that allowed for nonlinear and time-dependent effects. We found a U-shaped relationship between weight gain z score and gestational age at delivery among twin pregnancies (lowest hazard of delivery observed at z score = 1.2), which we attributed to increased hazard of early preterm spontaneous delivery among pregnancies with low weight gain and increased hazard of late preterm delivery without labor among pregnancies with high weight gain. Our findings may be useful for updating provisional guidelines for maternal weight gain in twin pregnancies.

Comparison of Different Modeling Approaches for Prescription Opioid Use and Its Association With Adverse Events.

Previous research linking opioid prescribing to adverse drug events has failed to properly account for the time-varying nature of opioid exposure. This study aimed to explore how the risk of opioid-related emergency department visits, readmissions, or deaths (composite outcome) varies with opioid dose and duration, comparing different novel modeling techniques. A prospective cohort of 1,511 hospitalized patients discharged from 2 McGill-affiliated hospitals in Montreal, 2014-2016, was followed from the first postdischarge opioid dispensation until 1 year after discharge. Marginal structural Cox proportional hazards models and their flexible extensions were used to explore the association between time-varying opioid use and the composite outcome. Weighted cumulative exposure models assessed cumulative effects of past use and explored how its impact depends on the recency of exposure. The patient mean age was 69.6 (standard deviation = 14.9) years; 57.7% were male. In marginal structural model analyses, current opioid use was associated with a 71% increase in the hazard of opioid-related adverse events (adjusted hazard ratio = 1.71, 95% confidence interval: 1.21, 2.43). The weighted cumulative exposure results suggested that the risk cumulates over the previous 50 days of opioid consumption. Flexible modeling techniques helped assess how the risk of opioid-related adverse events may be associated with time-varying opioid exposures while accounting for nonlinear relationships and the recency of past use.

Adapting SIMEX to correct for bias due to interval-censored outcomes in survival analysis with time-varying exposure.

Many clinical and epidemiological applications of survival analysis focus on interval-censored events that can be ascertained only at discrete times of clinic visits. This implies that the values of time-varying covariates are not correctly aligned with the true, unknown event times, inducing a bias in the estimated associations. To address this issue, we adapted the simulation-extrapolation (SIMEX) methodology, based on assessing how the estimates change with the artificially increased time between clinic visits. We propose diagnostics to choose the extrapolating function. In simulations, the SIMEX-corrected estimates reduced considerably the bias to the null and generally yielded a better bias/variance trade-off than conventional estimates. In a real-life pharmacoepidemiological application, the proposed method increased by 27% the excess hazard of the estimated association between a time-varying exposure, representing the 2-year cumulative duration of past use of a hypertensive medication, and the hazard of nonmelanoma skin cancer (interval-censored events). These simulation-based and real-life results suggest that the proposed SIMEX-based correction may help improve the accuracy of estimated associations between time-varying exposures and the hazard of interval-censored events in large cohort studies where the events are recorded only at relatively sparse times of clinic visits/assessments. However, these advantages may be less certain for smaller studies and/or weak associations.

Associations of 4 Nurse Staffing Practices With Hospital Mortality.

BACKGROUND: Cross-sectional studies of hospital-level administrative data have suggested that 4 nurse staffing practices-using adequate staffing levels, higher proportions of registered nurses (RNs) (skill mix), and more educated and experienced RNs-are each associated with reduced hospital mortality. To increase the validity of this evidence, patient-level longitudinal studies assessing the simultaneous associations of these staffing practices with mortality are required. METHODS: A dynamic cohort of 146,349 adult medical, surgical, and intensive care patients admitted to a Canadian University Health Center was followed for 7 years (2010-2017). We used a multivariable Cox proportional hazards model to estimate the associations between patients' time-varying cumulative exposure to measures of RN understaffing, skill mix, education, and experience, each relative to nursing unit and shift means, and the hazard of in-hospital mortality, while adjusting for patient and nursing unit characteristics, and modeling the current nursing unit of hospitalization as a random effect. RESULTS: Overall, 4854 in-hospital deaths occurred during 3,478,603 patient-shifts of follow-up (13.95 deaths/10,000 patient-shifts). In multivariable analyses, every 5% increase in the cumulative proportion of understaffed shifts was associated with a 1.0% increase in mortality (hazard ratio: 1.010; 95% confidence interval: 1.002-1.017; P=0.009). Moreover, every 5% increase in the cumulative proportion of worked hours by baccalaureate-prepared RNs was associated with a 2.0% reduction of mortality (hazard ratio: 0.980; 95% confidence interval: 0.965-0.995, P=0.008). RN experience and skill mix were not significantly associated with mortality. CONCLUSION: Reducing the frequency of understaffed shifts and increasing the proportion of baccalaureate-prepared RNs are associated with reduced hospital mortality.

Determination of the time-dependent association between ciprofloxacin consumption and ciprofloxacin resistance using a weighted cumulative exposure model compared with standard models.

OBJECTIVES: To obtain comprehensive insight into the association of ciprofloxacin use at different times in the past with the current risk of detecting resistance. METHODS: This retrospective nested case-control study of ciprofloxacin users used Dutch data from the PHARMO Database Network and one laboratory for the period 2003-14. Cases and controls were selected as patients with an antibiotic susceptibility test (AST) indicating ciprofloxacin resistance or susceptibility, respectively. We performed univariable and multivariable conditional logistic regression analyses, defining time-dependent exposure using standard definitions (current ciprofloxacin use, used 0-30, 31-90, 91-180 and 181-360 days ago) and a flexible weighted cumulative effect (WCE) model with four alternative time windows of past doses (0-30, 0-90, 0-180 and 0-360 days). RESULTS: The study population consisted of 230 cases and 909 controls. Under the standard exposure definitions, the association of ciprofloxacin use with resistance decreased with time [current use: adjusted OR 6.8 (95% CI 3.6-12.4); used 181-360 days ago: 1.3 (0.8-1.9)]. Under the 90 day WCE model (best-fitting model), more recent doses were more strongly associated with resistance than past doses, as was longer or repeated treatment. The 180 day WCE model, which fitted the data equally well, suggested that doses taken 91-180 days ago were also significantly associated with resistance. CONCLUSIONS: The estimates for the association between ciprofloxacin use at different times and resistance show that ciprofloxacin prescribers should consider ciprofloxacin use 0-180 days ago to ensure that patients receive suitable treatment. The OR of ciprofloxacin resistance could be reduced by eliminating repeated ciprofloxacin prescription within 180 days and by treating for no longer than necessary.

Long-term effectiveness of catheter ablation in patients with atrial fibrillation and heart failure.

AIMS: Randomized trials suggest reductions in all-cause mortality and heart failure (HF) rehospitalizations with catheter ablation (CA) in patients with atrial fibrillation (AF) and HF. Whether these results can be replicated in a real-world population with long-term follow-up or varies over time is unknown. We sought to evaluate the long-term effectiveness of CA in reducing the incidence of all-cause mortality, HF hospitalizations, stroke, and major bleeding in AF-HF patients. METHODS AND RESULTS: In a cohort of patients newly diagnosed with AF-HF in Quebec, Canada (2000-2017), CA patients were matched 1:2 to controls on time and frequency of hospitalizations. Confounders were controlled for using inverse probability of treatment weighting. Multivariable Cox models adjusted for the presence of cardiac electronic implantable devices and medication use during follow-up, and the effect of time since CA was modelled with B-splines. For non-fatal outcomes, the Lunn-McNeil approach was used to account for the competing risk of death. Among 101 933 AF-HF patients, 451 underwent CA and were matched to 899 controls. Over a median follow-up of 3.8 years, CA was associated with a statistically significant reduction in all-cause mortality [hazard ratio 0.4 (95% confidence interval 0.2-0.7)], but no difference in stroke or major bleeding. The hazard of HF rehospitalization for CA patients, relative to non-CA patients, varied with time since CA (P = 0.01), with a reduction in HF rehospitalizations until approximately 3 years post-CA. CONCLUSION: Compared with matched non-CA patients, CA was associated with a long-term reduction in all-cause mortality and a reduction in HF rehospitalizations until 3 years post-CA.

Nonlinear and time-dependent effects of sparsely measured continuous time-varying covariates in time-to-event analysis.

Many flexible extensions of the Cox proportional hazards model incorporate time-dependent (TD) and/or nonlinear (NL) effects of time-invariant covariates. In contrast, little attention has been given to the assessment of such effects for continuous time-varying covariates (TVCs). We propose a flexible regression B-spline-based model for TD and NL effects of a TVC. To account for sparse TVC measurements, we added to this model the effect of time elapsed since last observation (TEL), which acts as an effect modifier. TD, NL, and TEL effects are estimated with the iterative alternative conditional estimation algorithm. Furthermore, a simulation extrapolation (SIMEX)-like procedure was adapted to correct the estimated effects for random measurement errors in the observed TVC values. In simulations, TD and NL estimates were unbiased if the TVC was measured with a high frequency. With sparse measurements, the strength of the effects was underestimated but the TEL estimate helped reduce the bias, whereas SIMEX helped further to correct for bias toward the null due to "white noise" measurement errors. We reassessed the effects of systolic blood pressure (SBP) and total cholesterol, measured at two-year intervals, on cardiovascular risks in women participating in the Framingham Heart Study. Accounting for TD effects of SBP, cholesterol and age, the NL effect of cholesterol, and the TEL effect of SBP improved substantially the model's fit to data. Flexible estimates yielded clinically important insights regarding the role of these risk factors. These results illustrate the advantages of flexible modeling of TVC effects.

Flexible Modeling of the Association Between Cumulative Exposure to Low-Dose Ionizing Radiation From Cardiac Procedures and Risk of Cancer in Adults With Congenital Heart Disease.

Adults with congenital heart disease are increasingly being exposed to low-dose ionizing radiation (LDIR) from cardiac procedures. In a recent study, Cohen et al. (Circulation. 2018;137(13):1334-1345) reported an association between increased LDIR exposure and cancer incidence but did not explore temporal relationships. Yet, the impact of past exposures probably accumulates over years, and its strength may depend on the amount of time elapsed since exposure. Furthermore, LDIR procedures performed shortly before a cancer diagnosis may have been ordered because of early symptoms of cancer, raising concerns about reversal causality bias. To address these challenges, we combined flexible modeling of cumulative exposures with competing-risks methodology to estimate separate associations of time-varying LDIR exposure with cancer incidence and all-cause mortality. Among 24,833 patients from the Quebec Congenital Heart Disease Database, 602 had incident cancer and 500 died during a follow-up period of up to 15 years (1995-2010). Initial results suggested a strong association of cancer incidence with very recent LDIR exposures, likely reflecting reverse causality bias. When exposure was lagged by 2 years, an increased cumulative LDIR dose from the previous 2-6 years was associated with increased cancer incidence, with a stronger association for women. These results illustrate the importance of accurate modeling of temporal relationships between time-varying exposures and health outcomes.

Direct-Acting Antivirals for Hepatitis C: Predictors of Early Discontinuation in the Real World.

BACKGROUND: The treatment for hepatitis C virus (HCV) infection has evolved over time, and direct-acting antivirals (DAA) have revolutionized HCV therapy. OBJECTIVES: To (a) assess early treatment discontinuation and (b) identify predictors of early discontinuation in a cohort of patients receiving second-generation DAAs. METHODS: We identified HCV patients newly prescribed simeprevir/sofosbuvir (SIM/SOF), ledipasvir/sofosbuvir (LDV/SOF), ombitasvir/paritaprevir/ritonavir + dasabuvir (OPrD), sofosbuvir/velpatasvir (SOF/VEL), elbasvir/grazoprevir (EBR/GZR), and glecaprevir/pibrentasvir (GLE/PIB) between 2014 and 2017. Early discontinuation was defined as duration of therapy less than 8 weeks. Multivariable logistic regression was performed to evaluate the association of drug regimens and potential predictive factors to early discontinuation. RESULTS: We identified 26,098 DAA-treated patients: 67.8% with LDV/SOF, 9.9% with OPrD, 8.5% with SIM/SOF, 7.8% with SOF/VEL, 5.2% with EBR/GZR, and 0.8% with GLE/PIB. With approval of new therapies in 2016 and 2017, use of OPrD, LDV/SOF, and SIM/SOF declined substantially. At baseline, there was some heterogeneity of past HCV drug use and comorbidity across groups; patients on SIM/SOF had the highest frequency of previous interferon, cirrhosis, and decompensated cirrhosis. Most HCV patients received therapy for 8-12 weeks; fewer patients went through 16-week and 24-week therapy courses. Early discontinuation rates (95% CI) were 7.1% (6.0-8.2) for SIM/SOF, 3.2% (2.9-3.5) for LDV/SOF, 9.6% (8.5-10.7) for OPrD, 3.1% (2.3-3.8) for SOF/VEL, 4.2% (3.1-5.3) for EBR/GZR, and 2.5% (0.3-4.7) for GLE/PIB. In multivariable analyses, versus OPrD, patients starting other drug regimens were less likely to discontinue therapy early. Early discontinuation was more common in women, patients with baseline anemia, and Medicare and Medicaid patients. CONCLUSIONS: These real-world data confirm low rates of early discontinuation in users of second-generation DAAs. Future research focusing on socio-economic and sex/gender issues may help further optimize care for patients with HCV. DISCLOSURES: This study was funded by the Canadian Institutes of Health Research. Klein has received grants for investigator-initiated trials from ViiV Healthcare, Janssen, Gilead, and Merck, as well as consulting fees from ViiV Healthcare, Merck, and AbbVie. Feld has received research support and/or scientific consulting fees from AbbVie, Contravir, Enanta, Gilead, Janssen, Merck, and Wako. All other authors have no conflicts of interest to declare. Results from this study were presented as a poster at the 34th International Conference of Phamacoepidemiology and Therapeutic Risk Management; August 22-26, 2018; Prague, Czech Republic.

Associations Between Methotrexate Use and the Risk of Cardiovascular Events in Patients with Elderly-onset Rheumatoid Arthritis.

OBJECTIVE: We evaluated the associations between time-varying methotrexate (MTX) use and risk of cardiovascular events (CVE) in patients with rheumatoid arthritis (RA). METHODS: We studied an inception cohort of 23,994 patients with RA diagnosed after their 65th birthday. Multivariable Cox regression models were fit to evaluate the associations between time-varying MTX use, controlling for other risk factors, and time to CVE. Alternative models assessed the cumulative duration of MTX use over the (1) first year, (2) previous year (recent use), and (3) entire duration of followup. We also assessed whether the strength of the association varied over time. RESULTS: Over 115,453 patient-years (PY), 3294 (13.7%) patients experienced a CVE (28.5 events per 1000 PY; 95% CI 27.6-29.5). In the multivariable analyses, the model assessing time-varying continuous use in the most recent year yielded the best fit. Increasing recent MTX use was associated with lower CVE risks (HR 0.79 for continuous use vs no use in past 12 months, 95% CI 0.70-0.88; p < 0.0001). Greater MTX use in the first year after cohort entry was also protective (HR 0.84, 95% CI 0.72-0.96; p = 0.0048), but this effect decreased with increasing followup. In contrast, longer MTX use during the entire followup was not clearly associated with CVE risk (HR 0.98, 95% CI 0.95-1.01; p = 0.1441). CONCLUSION: We observed about a 20% decrease in CVE associated with recent continuous MTX use. Greater MTX use in the first year of cohort entry also appeared to be important in the association between MTX and CVE risk.

Asthma is not associated with the need for surgery in Crohn's disease when controlling for smoking status: a population-based cohort study.

PURPOSE: Growing evidence suggests asthma and Crohn's disease commonly cooccur. However, the impact of asthma on the prognosis of Crohn's disease is unknown. The aim of our study was to assess the effect of asthma on the need for intestinal resection in patients with Crohn's disease while adjusting for smoking status, imputed from a smaller, secondary data set. PATIENTS AND METHODS: Using health administrative data from a universally funded healthcare plan in Alberta, Canada, we conducted a cohort study to assess the effect of asthma on the need for surgery in patients with Crohn's disease diagnosed between 2002 and 2008 (N=2,113). Validated algorithms were used to identify incident cases of Crohn's disease, cooccurring asthma, and intestinal resection. The association between asthma and intestinal resection was estimated using multivariable Cox proportional hazards regression. Smoking status was imputed using a novel method using martingale residuals, derived from a data set of 485 patients enrolled in the Alberta Inflammatory Bowel Disease Consortium (2007 to 2014) who completed environmental questionnaires. All analyses were adjusted for age, sex, rural/urban status, and mean neighborhood income quintile. RESULTS: Asthma did not increase the risk of surgery in the health administrative data when not adjusting for smoking status (HR 1.03, 95% CI 0.81 to 1.29). The association remained nonsignificant after imputing smoking status in the health administrative data (HR 1.03, 95% CI 0.81 to 1.29). CONCLUSION: Although asthma is associated with an increased risk of Crohn's disease, co-occurring asthma is not associated with the risk of surgery in these patients. This null association persisted after adjusting for smoking status. This study described a novel method to adjust for confounding (smoking status) in time-to-event analyses, even when the confounding variable is unmeasured in health administrative data.

Acute vs cumulative benefits of metformin use in patients with type 2 diabetes and heart failure.

AIMS: To evaluate the association between metformin use and heart failure (HF) exacerbation in people with type 2 diabetes (T2D) and pre-existing HF using alternative exposure models. MATERIALS AND METHODS: We analysed data for patients with T2D and incident HF from a national US insurance claims database. We compared the results of several multivariable Cox models where time-varying use of metformin was modelled as: (1) current use; (2) total duration of past use; and (3) use within the past 30 days or 10 days. The outcome was defined as time to HF-related hospitalization. We then re-analysed the data using flexible weighted cumulative exposure (WCE) models. RESULTS: A total of 7620 patients with diabetes and incident HF were analysed. The mean (SD) patient age was 54 (8) years, and 58% (n = 4440) were men. In all, 3799 individuals (50%) were exposed to metformin, and 837 HF hospitalizations (11%) occurred (mean follow-up 1.7 years). Results of conventional models suggested potential acute benefits in reducing HF exacerbation with metformin use in the past 10 days (adjusted hazard ratio [aHR] 0.76, 95% confidence interval [CI] 0.60-0.97), while WCE models, which provided a better fit for the data, suggested lack of a systematic effect (aHR 0.91, 95% CI 0.69-1.20). CONCLUSIONS: Our results suggest that cumulative metformin exposure does not decrease the risk of HF-related exacerbation. Use of other anti-hyperglycaemic agents with proven efficacy in patients with HF should also be considered as treatment options in this population.

Risk of acute myocardial infarction with real-world NSAIDs depends on dose and timing of exposure.

PURPOSE: Real-life use of nonsteroidal anti-inflammatory drugs (NSAIDs) is dynamic. This study aimed to characterize the temporal association between time-varying NSAID exposure and acute myocardial infarction (MI). METHODS: Nested case-control analyses were conducted on a Quebec administrative health cohort. NSAID dose, confounders, and outcome status were determined for each day of follow-up. To better account for dose and timing of past exposures, flexible weighted cumulative exposure models were also fitted. RESULTS: The cohort consisted of 233 816 older adults including 21 256 acute MI cases. Dose-related increased risks of MI were found with current use of all NSAIDs. In models not accounting for duration of use, ORs (95%CI) for the most common current daily dose vs. no current exposure were: celecoxib 200 mg: 1.16 (1.10, 1.22), diclofenac 150 mg: 1.59 (1.38, 1.84), ibuprofen 1200 mg: 1.42 (1.17, 1.74), naproxen 750 mg: 1.38 (1.21, 1.58), and rofecoxib 25 mg: 1.54 (1.43, 1.66). Weighted cumulative exposure models confirmed that all NSAIDs-including naproxen-are associated with an increased risk of MI and suggested that doses taken up to 3 weeks ago for rofecoxib, ibuprofen, and naproxen and up to 75 days ago for diclofenac and celecoxib may contribute to the current MI risk. However, the celecoxib risk seems to require continuous use for more than 30 days, whereas for other NSAIDs, a heightened MI risk occurs within 7 days. CONCLUSIONS: Weighted cumulative exposure analysis uncovered NSAID-specific differences in the immediate MI risk and how this risk seems to accumulate. KEY POINTS Accurate assessment of drug safety requires an etiologically correct model encompassing all relevant aspects of exposure. Weighted cumulative exposure models suggest that the relative importance of past doses on the risk of MI differs among NSAIDs. All common NSAIDs are associated with an increased MI risk. Celecoxib MI risk seems to depend on continuously using the drug for more than 30 days, whereas for ibuprofen, rofecoxib, diclofenac, and naproxen, a heightened MI risk occurs within 7 days of use.

The missing cause approach to unmeasured confounding in pharmacoepidemiology.

Unmeasured confounding is a major threat to the validity of pharmacoepidemiological studies of medication safety and effectiveness. We propose a new method for detecting and reducing the impact of unobserved confounding in large observational database studies. The method uses assumptions similar to the prescribing preference-based instrumental variable (IV) approach. Our method relies on the new 'missing cause' principle, according to which the impact of unmeasured confounding by (contra-)indication may be detected by assessing discrepancies between the following: (i) treatment actually received by individual patients and (ii) treatment that they would be expected to receive based on the observed data. Specifically, we use the treatment-by-discrepancy interaction to test for the presence of unmeasured confounding and correct the treatment effect estimate for the resulting bias. Under standard IV assumptions, we first proved that unmeasured confounding induces a spurious treatment-by-discrepancy interaction in risk difference models for binary outcomes and then simulated large pharmacoepidemiological studies with unmeasured confounding. In simulations, our estimates had four to six times smaller bias than conventional treatment effect estimates, adjusted only for measured confounders, and much smaller variance inflation than unbiased but very unstable IV estimates, resulting in uniformly lowest root mean square errors. The much lower variance of our estimates, relative to IV estimates, was also observed in an application comparing gastrointestinal safety of two classes of anti-inflammatory drugs. In conclusion, our missing cause-based method may complement other methods and enhance accuracy of analyses of large pharmacoepidemiological studies.

The Longterm Effect of Early Intensive Treatment of Seniors with Rheumatoid Arthritis: A Comparison of 2 Population-based Cohort Studies on Time to Joint Replacement Surgery.

OBJECTIVE: Disease-modifying antirheumatic drugs (DMARD) have the greatest effect when initiated early. We evaluated the influence of early exposure to DMARD on time to joint replacement surgery among patients with incident rheumatoid arthritis (RA). METHOD: Using a common protocol, we undertook 2 independent population-based cohort studies of patients with incident RA aged 66 years or older in Ontario (ON) and Quebec (QC) covering the period 2000-2013. We used Cox proportional hazards regression with time-dependent variables measuring duration of drug use in the first year, separately for methotrexate (MTX) and other DMARD, adjusting for baseline demographics, clinical factors, and other potentially confounding drug exposures. Our outcome measure was any joint replacement derived from standardized procedure codes. Adjusted HR and 95% CI were estimated. RESULTS: Among 20,918 ON and 6754 QC patients with RA followed for a median of 4.5 years, 2201 and 494 patients underwent joint replacement surgery for crude event rates of 2.0 and 1.4 per 100 person-years, respectively. Greater cumulative exposure to MTX (HR 0.97, 95% CI 0.95-0.98) and other DMARD (HR 0.98, 95% CI 0.97-0.99) in the first year after diagnosis was associated with longer times to joint replacement in ON, corresponding to a 2-3% decrease in the hazard of surgery with each additional month of early use. Similar results were observed in QC. CONCLUSION: Greater duration of exposure to DMARD soon after RA diagnosis was associated with delays to joint replacement surgery in both provinces. Early intensive treatment of RA may ultimately reduce demand for joint replacement surgery.

Risk of Incident Diabetes Mellitus Associated With the Dosage and Duration of Oral Glucocorticoid Therapy in Patients With Rheumatoid Arthritis.

OBJECTIVE: To quantify the risk of incident diabetes mellitus (DM) associated with the dosage, duration, and timing of glucocorticoid (GC) use in patients with rheumatoid arthritis (RA). METHODS: We undertook a cohort study using 2 databases: a UK primary care database (the Clinical Practice Research Datalink [CPRD]) including 21,962 RA patients (1992-2009) and the US National Data Bank for Rheumatic Diseases (NDB) including 12,657 RA patients (1998-2013). Information on the dosage and timing of GC use was extracted. DM in the CPRD was defined using Read codes, at least 2 prescriptions for oral antidiabetic medication, or abnormal blood test results. DM in the NDB was defined through patient self-reports. Data were analyzed using time-dependent Cox models and a novel weighted cumulative dose (WCD) model that accounts for dosage, duration, and timing of treatment. RESULTS: The hazard ratio (HR) was 1.30 (95% confidence interval [95% CI] 1.17-1.45) and 1.61 (95% CI 1.37-1.89) in current GC users compared to nonusers in the CPRD and the NDB, respectively. A range of conventional statistical models consistently confirmed increases in risk with the GC dosage and duration. The WCD model showed that recent GC use contributed the most to the current risk of DM, while doses taken >6 months previously did not influence current risk. In the CPRD, 5 mg of prednisolone equivalent dose for the last 1, 3, and 6 months was significantly associated with HRs of 1.20, 1.43, and 1.48, respectively, compared to nonusers. CONCLUSION: GC use is a clinically important and quantifiable risk factor for DM. Risk is influenced by the dosage and treatment duration, although only for GC use within the last 6 months.

Early medication use in new-onset rheumatoid arthritis may delay joint replacement: results of a large population-based study.

INTRODUCTION: Use of disease-modifying anti-rheumatic drugs (DMARDs) in rheumatoid arthritis (RA) may prevent joint damage and potentially reduce joint replacement surgeries. We assessed the association between RA drug use and joint replacement in Quebec, Canada. METHODS: A cohort of new-onset RA patients was identified from Quebec's physician billing and hospitalization databases from 2002-2011. The outcome was defined using procedure codes submitted by orthopedic surgeons. Medication use was obtained from pharmacy databases. We used alternative Cox regression models with time-dependent variables measuring the cumulative effects of past use during different time windows (one model focussing on the first year after cohort entry) for methotrexate (MTX), and other DMARDs. Models were adjusted for baseline sociodemographics, co-morbidity and prior health service use, time-dependent cumulative use of other drugs (anti-tumor necrosis factor [anti-TNF] agents, other biologics, cyclooxygenase-2 inhibitors [COXIBs], nonselective nonsteroidal antiinflammatory drugs [NSAIDs], and systemic steroids), and markers of disease severity. RESULTS: During follow-up, 608 joint replacements occurred among 11,333 patients (median follow-up: 4.6 years). The best-fitting model relied on the cumulative early use (within the first year after cohort entry) of MTX and of other DMARDs, with an interaction between MTX and other DMARDs. In this model, greater exposure within the first year, to either MTX (adjusted hazard ratio, HR = 0.95 per 1 month, 95% confidence interval, 95% CI 0.93-0.97) or other DMARDs (HR = 0.97, 95% CI 0.95-0.99) was associated with longer time to joint replacement. CONCLUSIONS: Our results suggest that longer exposure to either methotrexate (MTX) or other DMARDs within the first year after RA diagnosis is associated with longer time to joint replacement surgery.

Evidence of subgroup-specific treatment effect in the absence of an overall effect: is there really a contradiction?

PURPOSE: Interaction and subgroup analyses remain controversial topics in epidemiology. A recent theoretical paper suggested that a combination of no overall treatment-outcome association and treatment effect limited to a single subgroup would imply a clinically implausible interaction, with opposite treatment effects in the two subgroups. However, this argument was based entirely on point estimates and ignored sampling error and statistical inference. METHODS: We simulated hypothetical studies in which treatment truly affected the outcome in only one subgroup, with no effect in the other subgroup. We generated 1000 random samples for three study designs (small clinical study, case-control, and large cohort), and different values of total sample size (N), relative size of the affected subgroup, and treatment effect. We estimated the frequency of significant results for tests of overall and subgroup-specific treatment effects, and treatment-by-subgroup interaction. RESULTS: Combination of statistically non-significant overall treatment effect and significant treatment-by-subgroup interaction occurred frequently, especially if the affected subgroup was proportionally smaller, even in studies with high power to detect the overall effect (e.g. in 37.1% of samples with N = 20 000, with 600 outcomes, and an effect (odds ratio of 1.5) limited to 30% of subjects). Furthermore, in most samples with a significant interaction, subgroup analyses correctly indicated that the significant effect was limited to one subgroup. CONCLUSION: In studies where the treatment truly affects the risks in only one subgroup, a non-significant overall effect will often coincide with a statistically significant treatment-by-subgroup interaction. Thus, a non-significant overall effect should not prevent testing plausible interactions.