Bias In Treatment Effect Research Articles

Consequences of handling missing data for treatment response in osteoarthritis: a simulation study

To understand how handling of missing data influences the statistical power and bias of treatment effects in randomised controlled trials of painful knee osteoarthritis (OA). We simulated trials with missing data (withdrawals) due to lack-of-efficacy. Outcome measures were response/non-response according to the Outcome Measures in Rheumatology-Osteoarthritis Research Society International (OMERACT-OARSI) set of responder criteria, Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain and physical function from the WOMAC questionnaire, and patient global assessment. We used five methods for managing missing data: ignoring the missing data, last and baseline observation carried forward (LOCF and BOCF), and multiple imputation with two different strategies. The treatment effect was then analysed by appropriate univariate and longitudinal statistical methods, and power, bias and mean squared error (MSE) was assessed by comparing the estimated treatment effect in the trials with missing data with the estimated treatment effect on the trials without missing data. The best imputation method in terms of high power and low bias/MSE was our implementation of regression multiple imputation. The most conservative method was the data augmentation Markov chain Monte Carlo (MCMC) multiple imputation. The LOCF, BOCF and the complete-case methods were not particularly conservative and gave relatively low power and high bias. The analysis on the WOMAC pain scale gave less bias and higher power than the OMERACT-OARSI responder outcome measure. Multiple imputation of missing data may be used to decrease bias/MSE and increase power in OA trials. These results can guide investigators in the choice of outcome measures and especially how missing data can be handled.

Evaluating treatment effectiveness in patient subgroups: a comparison of propensity score methods with an automated matching approach

Propensity score (Pscore) matching and inverse probability of treatment weighting (IPTW) can remove bias due to observed confounders, if the Pscore is correctly specified. Genetic Matching (GenMatch) matches on the Pscore and individual covariates using an automated search algorithm to balance covariates. This paper compares common ways of implementing Pscore matching and IPTW, with Genmatch for balancing time-constant baseline covariates}. The methods are considered when estimates of treatment effectiveness are required for patient subgroups, and the treatment allocation process differs by subgroup. We apply these methods in a prospective cohort study that estimates the effectiveness of Drotrecogin alfa activated, for subgroups of patients with severe sepsis. In a simulation study we compare the methods when the Pscore is correctly specified, and then misspecified by ignoring the subgroup-specific treatment allocation. The simulations also consider poor overlap in baseline covariates, and different sample sizes. In the case study, GenMatch reports better covariate balance than IPTW or Pscore matching. In the simulations with correctly specified Pscores, good overlap and reasonable sample sizes, all methods report minimal bias. When the Pscore is misspecified, GenMatch reports the least imbalance and bias. With small sample sizes, IPTW is the most efficient approach, but all methods report relatively high bias of treatment effects. This study shows that overall GenMatch achieves the best covariate balance for each subgroup, and is more robust to Pscore misspecification than common alternative Pscore approaches.

Treatment effect bias in randomised controlled trials using surrogate outcomes: a preliminary cohort study analysis

Background Ideally, decisions on the value of health technologies should be based on evidence from well-conducted clinical trials that assess clinically important final patientrelevant outcomes, such as mortality or impaired quality of life. Pressure to reduce the delay in the availability of technologies to patients has led to an increased reliance on the use of surrogate outcomes [1]. A key tenant of surrogate outcomes is unbiased quantification of the predictive treatment effect on the final patient-relevant clinical outcomes. This study compares the treatment outcome of trials that use a surrogate versus a final patient-relevant primary outcome.

Estimation Bias in Complete-Case Analysis in Crossover Studies with Missing Data

Crossover designs are used often in clinical trials. It is not uncommon that subjects discontinue before completing all treatment periods in a crossover study. Despite availability of statistical methodologies utilizing all available data and software for obtaining valid inferences under the assumption of missing at random (MAR), naïve approaches, such as the complete case (CC) analysis, which is only valid with a strong assumption of missing completely at random are still widely used in practice. In this article, we obtain the analytical form of the estimation bias of treatment effects with CC for linear mixed models. We use simulation studies to examine the inflation of Type I error and efficiency loss in the inferences with CC under MAR. Invalidity and inefficiency of two other commonly used approaches for defining analyzed data in the presence of missing data, including data from at least two periods in three period crossover and available cases for a specific comparison of interest, are also demonstrated through simulation studies.

Mortality Probability Model III and Simplified Acute Physiology Score II: Assessing Their Value in Predicting Length of Stay and Comparison to APACHE IV

<h3>Background</h3> To develop and compare ICU length-of-stay (LOS) risk-adjustment models using three commonly used mortality or LOS prediction models. <h3>Methods</h3> Between 2001 and 2004, we performed a retrospective, observational study of 11,295 ICU patients from 35 hospitals in the California Intensive Care Outcomes Project. We compared the accuracy of the following three LOS models: a recalibrated acute physiology and chronic health evaluation (APACHE) IV-LOS model; and models developed using risk factors in the mortality probability model III at zero hours (MPM₀) and the simplified acute physiology score (SAPS) II mortality prediction model. We evaluated models by calculating the following: (1) grouped coefficients of determination; (2) differences between observed and predicted LOS across subgroups; and (3) intraclass correlations of observed/expected LOS ratios between models. <h3>Results</h3> The grouped coefficients of determination were APACHE IV with coefficients recalibrated to the LOS values of the study cohort (APACHE IVrecal) [<i>R</i>² = 0.422], mortality probability model III at zero hours (MPM₀ III) [<i>R</i>² = 0.279], and simplified acute physiology score (SAPS II) [<i>R</i>² = 0.008]. For each decile of predicted ICU LOS, the mean predicted LOS vs the observed LOS was significantly different (p ≤ 0.05) for three, two, and six deciles using APACHE IVrecal, MPM₀ III, and SAPS II, respectively. Plots of the predicted vs the observed LOS ratios of the hospitals revealed a threefold variation in LOS among hospitals with high model correlations. <h3>Conclusions</h3> APACHE IV and MPM₀ III were more accurate than SAPS II for the prediction of ICU LOS. APACHE IV is the most accurate and best calibrated model. Although it is less accurate, MPM₀ III may be a reasonable option if the data collection burden or the treatment effect bias is a consideration.

Missing data in randomized controlled trials of rheumatoid arthritis with radiographic outcomes: A simulation study

To assess the impact, in terms of statistical power and bias of treatment effect, of approaches to dealing with missing data in randomized controlled trials of rheumatoid arthritis with radiographic outcomes. We performed a simulation study. The missingness mechanisms we investigated copied the process of withdrawal from trials due to lack of efficacy. We compared 3 methods of managing missing data: all available data (case-complete), last observation carried forward (LOCF), and multiple imputation. Data were then analyzed by classic t-test (comparing the mean absolute change between baseline and final visit) or F test (estimation of treatment effect with repeated measurements by a linear mixed-effects model). With a missing data rate close to 15%, the treatment effect was underestimated by 18% as estimated by a linear mixed-effects model with a multiple imputation approach to missing data. This bias was lower than that obtained with the case-complete approach (-25%) or LOCF approach (-35%). This statistical approach (combination of multiple imputation and mixed-effects analysis) was moreover associated with a power of 70% (for a 90% nominal level), whereas LOCF was associated with a power of 55% and a case-complete power of 58%. Analysis with the t-test gave qualitatively equivalent but poorer quality results, except when multiple imputation was applied. Our simulation study demonstrated multiple imputation, offering the smallest bias in treatment effect and the highest power. These results can help in planning trials, especially in choosing methods of imputation and data analysis.

Dealing with the common econometric problems of count data with excess zeros, endogenous treatment effects, and attrition bias

We use experimental panel data to estimate impacts of managed care on medical care use counts free of endogenous treatment effect bias. Our count regressions also allow for excess zeros and correct for possible attrition bias.