Effectiveness of initial methotrexate-based treatment approaches in early rheumatoid arthritis: an elicitation of rheumatologists' beliefs.

OBJECTIVES: To quantify rheumatologists' beliefs about the effectiveness of triple therapy (MTX + HCQ + SSZ) and other commonly used initial treatments for RA. METHODS: In a Bayesian belief elicitation exercise, 40 rheumatologists distributed 20 chips, each representing 5% of their total weight of belief on the probability that a typical patient with moderate-severe early RA would have an ACR50 response within 6 months with MTX (oral and s.c.), MTX + HCQ (dual therapy) and triple therapy. Parametric distributions were fit, and used to calculate pairwise median relative risks (RR), with 95% credible intervals, and estimate sample sizes for new trials to shift these beliefs. RESULTS: In the pooled analysis, triple therapy was perceived to be superior to MTX (RR 1.97; 1.35, 2.89) and dual therapy (RR 1.32; 1.03, 1.73). A pessimistic subgroup (n = 10) perceived all treatments to be similar, whereas an optimistic subgroup (n = 10) believed triple therapy to be most effective of all (RR 4.03; 2.22, 10.12). Similar variability was seen for the comparison between oral and s.c. MTX. Assuming triple therapy is truly more effective than MTX, a trial of 100 patients would be required to convince the pessimists; if triple therapy truly has no-modest effect (RR <1.5), a non-inferiority trial of 475 patients would be required to convince the optimists. CONCLUSION: Rheumatologists' beliefs regarding the effectiveness of triple therapy vary, which may partially explain the variability in its use. Owing to the strength of beliefs, some may be reluctant to shift, even with new evidence.

Ensemble learning methods of inference for spatially stratified infectious disease systems.

Individual level models are a class of mechanistic models that are widely used to infer infectious disease transmission dynamics. These models incorporate individual level covariate information accounting for population heterogeneity and are generally fitted in a Bayesian Markov chain Monte Carlo (MCMC) framework. However, Bayesian MCMC methods of inference are computationally expensive for large data sets. This issue becomes more severe when applied to infectious disease data collected from spatially heterogeneous populations, as the number of covariates increases. In addition, summary statistics over the global population may not capture the true spatio-temporal dynamics of disease transmission. In this study we propose to use ensemble learning methods to predict epidemic generating models instead of time consuming Bayesian MCMC method. We apply these methods to infer disease transmission dynamics over spatially clustered populations, considering the clusters as natural strata instead of a global population. We compare the performance of two tree-based ensemble learning techniques: random forest and gradient boosting. These methods are applied to the 2001 foot-and-mouth disease epidemic in the U.K. and evaluated using simulated data from a clustered population. It is shown that the spatially clustered data can help to predict epidemic generating models more accurately than the global data.

Stage Shifting by Modifying the Determinants of Breast Cancer Stage at Diagnosis: A Simulation Study.

BACKGROUND: Breast cancer is the most common cancer in Canadian women; nearly 25% of women diagnosed with cancer have breast cancer. The early detection of breast cancer is a major challenge because tumours often grow without causing symptom. The diagnosis of breast cancer at an early stage (stages I and II) improves survival outcomes because treatments are more effective and better tolerated. To better inform the prevention of and screening for breast cancer, simulations using modifiable rather than non-modifiable risk factors may be helpful in shifting the stage at diagnosis downward. METHODS: Breast cancer stages were simulated using the data distributions from Alberta's Tomorrow Project participants who developed breast cancer. Using multivariable partial proportional odds regression models, modifiable lifestyle factors associated with the stage of cancer at diagnosis were evaluated. The proportions or mean levels of these lifestyle factors in the simulated population were systematically changed, then multiplied by their corresponding estimated odds ratios from the real data example. The effects of these changes were evaluated singly as well as cumulatively. RESULTS: Increasing total dietary protein (g/day) intake was the single most important lifestyle factor in shifting the breast cancer stage downwards followed by decreasing total dietary energy intake (kcal/day). Increasing the proportion of women who spend time in the sun between 11 am and 4 pm in the summer months, who have had a mammogram, who have been pregnant or reducing the proportion who are in stressful situations had much smaller effects. The percentage of Stage I diagnoses could be increased by approximately 12% with small modifications of these lifestyle factors. CONCLUSION: Shifting the breast cancer stage at diagnosis of a population may be achieved through changes to lifestyle factors. This proof of principle study that evaluated multiple factors associated with the stage at diagnosis in a population can be expanded to other cancers as well, providing opportunities for cancer prevention programs to target specific factors and identify populations at higher risk.

Patient preferences for maintenance therapy in Crohn's disease: A discrete-choice experiment.

OBJECTIVE: To quantify patient preferences for maintenance therapy of Crohn's disease and understand the impact on treatment selection. METHODS: We conducted a discrete-choice experiment in patients with Crohn's disease (n = 155) to measure the importance of attributes relevant to choosing between different medical therapies for maintenance of Crohn's disease. The attributes included efficacy and withdrawals due to adverse events, as well as dosing and other rare risks of treatment. From the discrete-choice experiment we estimated the part-worth (importance) of each attribute level, and explored preference heterogeneity through latent class analysis. We then used the part-worths to apply weights across each outcome from a prior network meta-analysis to estimate patients' preferred treatment in pairwise comparisons and for the overall group of treatments. RESULTS: The discrete-choice experiment revealed that maintaining remission was the most important attribute. Patients would accept a rare risk of infection or cancer for a 14% absolute increased chance of remission. Latent class analysis demonstrated that 45% of the cohort was risk averse, either to adverse events or requiring a course of prednisone. When these preferences were used in modelling studies to compare pairs of treatments, there was a ≥ 78% probability that all biologic treatments were preferred to azathioprine and methotrexate, based on the balance of benefits and harms. When comparing all treatments, adalimumab was preferred by 53% of patients, who were motivated by efficacy, and vedolizumab was preferred by 30% who were driven by the preference to avoid risks. However, amongst biologic treatment options, there was considerable uncertainty regarding the preferred treatment at the individual patient level. CONCLUSION: Patients with Crohn's disease from our population were, on average, focused on the benefits of treatment, supporting intensive treatment approaches aimed at maintaining remission. Important preference heterogeneity was identified, however, highlighting the importance of shared decision making when selecting treatments.

Joint Estimation of Remission and Response for Methotrexate-Based DMARD Options in Rheumatoid Arthritis: A Bivariate Network Meta-Analysis.

OBJECTIVE: To jointly estimate American College of Rheumatology (ACR50) response (a more commonly reported outcome) and remission (a more clinically relevant outcome) for methotrexate (MTX)-based treatment options in rheumatoid arthritis (RA). METHODS: We conducted a bivariate network meta-analysis (NMA) to compare MTX monotherapy and MTX-based conventional and biologic disease-modifying antirheumatic drug (DMARD) combinations for RA. The correlation between the outcomes was derived from an incident RA cohort study, whereas the treatment effects were derived from randomized trials in the network of evidence. The analyses were conducted separately for MTX-naïve and MTX-inadequate response (IR) populations in a Bayesian framework with uninformative priors. RESULTS: From the cohort study, the correlation between ACR50 response and Disease Activity Score 28 remission at 6 months was moderate (Pearson correlation coefficient = 0.58). In the bivariate NMA for MTX-naïve populations, most combinations of MTX with either biologic or tofacitinib were statistically superior to MTX alone for both ACR50 response and remission. Triple therapy (MTX + sulfasalazine + hydroxychloroquine) was the only nonbiologic DMARD statistically superior to MTX for either ACR50 response (odds ratio [OR] 95% credible interval: 2.1 [1.0, 4.3]) or remission (OR: 2.5 [1.0, 5.8]). In the MTX-IR analysis, all treatments except MTX + sulfasalazine were statistically superior to MTX alone. Compared to analyzing the outcomes separately, the bivariate model often resulted in more precise estimates and allowed remission to be estimated for all treatments. CONCLUSION: Borrowing the strength of correlation between outcomes allowed us to demonstrate a statistically significant benefit for remission across most MTX-based DMARD combinations, including triple therapy.

Supervised learning and prediction of spatial epidemics.

Parameter estimation for mechanistic models of infectious disease can be computationally intensive. Nsoesie et al. (2011) introduced an approach for inference on infectious disease data based on the idea of supervised learning. Their method involves simulating epidemics from various infectious disease models, and using classifiers built from the epidemic curve data to predict which model were most likely to have generated observed epidemic curves. They showed that the classification approach could fairly identify underlying characteristics of the disease system, without fitting various transmission models via, say, Bayesian Markov chain Monte Carlo. We extend this work to the case where the underlying infectious disease model is inherently spatial. Our goal is to compare the use of global epidemic curves for building the classifier, with the use of spatially stratified epidemic curves. We demonstrate these methods on simulated data and apply the method to analyze a tomato spotted wilt virus epidemic dataset.