Donna M Dryden

Risk of bias versus quality assessment of randomised controlled trials: cross sectional study.

Objectives To evaluate the risk of bias tool, introduced by the Cochrane Collaboration for assessing the internal validity of randomised trials, for inter-rater agreement, concurrent validity compared with the Jadad scale and Schulz approach to allocation concealment, and the relation between risk of bias and effect estimates. Design Cross sectional study. Study sample 163 trials in children. Main outcome measures Inter-rater agreement between reviewers assessing trials using the risk of bias tool (weighted κ), time to apply the risk of bias tool compared with other approaches to quality assessment (paired t test), degree of correlation for overall risk compared with overall quality scores (Kendall’s τ statistic), and magnitude of effect estimates for studies classified as being at high, unclear, or low risk of bias (metaregression). Results Inter-rater agreement on individual domains of the risk of bias tool ranged from slight (κ=0.13) to substantial (κ=0.74). The mean time to complete the risk of bias tool was significantly longer than for the Jadad scale and Schulz approach, individually or combined (8.8 minutes (SD 2.2) per study v 2.0 (SD 0.8), P<0.001). There was low correlation between risk of bias overall compared with the Jadad scores (P=0.395) and Schulz approach (P=0.064). Effect sizes differed between studies assessed as being at high or unclear risk of bias (0.52) compared with those at low risk (0.23). Conclusions Inter-rater agreement varied across domains of the risk of bias tool. Generally, agreement was poorer for those items that required more judgment. There was low correlation between assessments of overall risk of bias and two common approaches to quality assessment: the Jadad scale and Schulz approach to allocation concealment. Overall risk of bias as assessed by the risk of bias tool differentiated effect estimates, with more conservative estimates for studies at low risk.

The epidemiology of traumatic spinal cord injury in Alberta, Canada

OBJECTIVES To describe the incidence and pattern of traumatic spinal cord injury and cauda equina injury (SCI) in a geographically defined region of Canada. METHODS The study period was April 1, 1997 to March 31, 2000. Data were gathered from three provincial sources: administrative data from the Alberta Ministry of Health and Wellness, records from the Alberta Trauma Registry, and death certificates from the Office of the Medical Examiner. RESULTS From all three data sources, 450 cases of SCI were identified. Of these, 71 (15.8%) died prior to hospitalization. The annual incidence rate was 52.5/million population (95% CI: 47.7, 57.4). For those who survived to hospital admission, the incidence rate was 44.3/million/year (95% CI: 39.8, 48.7). The incidence rates for males were consistently higher than for females for all age groups. Motor vehicle collisions accounted for 56.4% of injuries, followed by falls (19.1%). The highest incidence of motor vehicle-related SCI occurred to those between 15 and 29 years (60/million/year). Fall-related injuries primarily occurred to those older than 60 years (45/million/year). Rural residents were 2.5 times as likely to be injured as urban residents. CONCLUSION Prevention strategies for SCI should target males of all ages, adolescents and young adults of both sexes, rural residents, motor vehicle collisions, and fall prevention for those older than 60 years.

Systematic Review: Implantable Cardioverter Defibrillators for Adults with Left Ventricular Systolic Dysfunction

Left ventricular (LV) systolic dysfunction carries a high risk for sudden cardiac death (1). Implantable cardioverter defibrillators (ICDs) can potentially mitigate this risk by delivering rapid life-saving therapy and have been substantially refined since their initial development in the late 1970s (2). Randomized, controlled trials (RCTs) have tested the efficacy of ICDs in high-risk individuals. We previously reported a systematic review of 8 RCTs (3 RCTs of secondary prevention in survivors of sudden cardiac death; 5 RCTs of primary prevention in patients without a history of ventricular arrhythmias) demonstrating a 26% reduction in all-cause mortality and a 57% reduction in sudden cardiac death with ICDs (3). Since then, additional RCTs of primary prevention have been published, and questions have arisen about the generalizability of the RCT results for ICDs to clinical practice. In particular, it is uncertain whether the benefits of ICDs seen in the trials extend to nontrial populations and whether the risks associated with ICDs may be higher in clinical practice than reported in trials. Given the public policy implications, we extended our previous systematic review of the efficacy (that is, the risks and benefits of a therapy when tested under ideal circumstances) (4) of ICDs in patients with LV systolic dysfunction by updating it with recently published RCTs that examined efficacy. In addition, we expanded the review to include data from observational studies to determine the effectiveness (that is, the risks and benefits of a therapy when tested under usual clinical practice conditions) and safety of ICDs when used in clinical practice. Methods A study protocol meeting Cochrane criteria, including all of the elements described briefly in the following sections, was developed and followed by the study authors in conjunction with the Agency for Healthcare Research and Quality (AHRQ). Search Strategy We sought studies published between 1980 and 27 April 2007 by searching MEDLINE, Ovid MEDLINE In-Process & Other Non-Indexed Citations, Cochrane Central Register of Controlled Trials, Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, Health Technology Assessment Database, EMBASE, Science Citation Index Expanded (via Web of Science), International Pharmaceutical Abstracts, PubMed, National Library of Medicine Gateway, OCLC ProceedingsFirst and PapersFirst, Computer Retrieval of Information on Scientific Projects, various trial registries (including the National Research Register [United Kingdom], Australian Clinical Trials Registry, ClinicalTrials.gov, and Current Controlled Trials), and U.S. Food and Drug Administration reports. In addition, we hand-searched abstracts from the annual Heart Rhythm Society meetings and the reference lists of review articles and included studies; we also contacted authors of included studies for additional citations and information. Unpublished studies and individual-patient data were sought from device manufacturers, including Medtronic (Minneapolis, Minnesota), Guidant Corp. (Indianapolis, Indiana), and St. Jude Medical (St. Paul, Minnesota). The search was not limited by language or publication status. The search terms included Medtronic InSync, ELA medical, Guidant, St. Jude, implantable defibrillators, implantable cardioverter defibrillators, AICD, ICD, single chamber ICD, dual chamber ICD, congestive heart failure, CHF, chronic heart failure, and heart diseases. A full list of search strategies (adapted for each database) and search results are available at www.ahrq.gov/clinic/tp/defibtp.htm (5). Study Selection We selected original research studies that had at least 25 participants and reported mortality or peri- or postimplantation complications with ICDs in adult patients with LV systolic dysfunction (left ventricular ejection fraction [LVEF]0.35, regardless of whether the patients had heart failure symptoms). To address efficacy questions, we restricted the analyses to RCTs. To address effectiveness questions, we expanded our inclusion criteria to include observational studies with contemporaneous comparison groups (such as cohort studies) and RCTs that did not report efficacy outcomes. To address safety questions, we included evidence from both RCTs and observational studies (including those without contemporaneous control groups, such as case series and registry data). Data Extraction and Analysis Study selection, quality assessment, and data extraction were completed by several investigators in duplicate and independently, using the methods recommended by the Quality of Reporting of Meta-analyses (QUOROM) group (6). We assessed quality by using the methods of Schulz and colleagues (7), the 5-item Jadad scale (8), and the 27-point Downs and Black scale (9). Publication bias was assessed visually by using funnel plots and quantitatively by using the rank correlation test (10), the graphical test (11), and the trim-and-fill method (12). Random-effects models were used to calculate pooled relative risks (RRs) in Review Manager 4.2.5 (Cochrane Collaboration, Copenhagen, Denmark). The length of study follow-up versus all-cause mortality was plotted for each study, and inverse varianceweighted least-squares regression was used to create a best-fit line. Postimplantation complications were expressed per 100 patient-years (calculated by multiplying the frequency of events in each study by the duration of follow-up, and standardizing to a denominator of 100) and are unadjusted rates. All results were reported with 95% CIs and, where appropriate, SDs or SEs. Statistical heterogeneity was quantified by using the I 2 statistic (13). In addition to examining for differences in point estimates across study designs and study quality, we explored device efficacy in different patient subgroups by using meta-regression. Covariates tested included presence of cardiac resynchronization therapy, length of follow-up, ischemic etiology, New York Heart Association (NYHA) class, age, QRS interval, LVEF, and primary versus secondary prevention. Role of the Funding Source The funding source (AHRQ, U.S. Department of Health and Human Services) had no role in the collection, analysis, or interpretation of the data or in the decision to submit the manuscript for publication. Results Literature Search From 4439 citations (Figure 1), we identified 12 RCTs (8516 patients) for the ICD efficacy review (1426), 53 studies (26840 patients from 5 nonefficacy RCTs and 48 observational studies [25 retrospective and 23 prospective]) for the ICD effectiveness review (2778), and 64 studies (86809 patients from 11 efficacy RCTs, 10 RCTs without efficacy outcomes, and 43 observational studies [24 retrospective and 19 prospective]) for the ICD safety review (1417, 1927, 29, 30, 34, 3740, 4245, 47, 48, 52, 54, 60, 61, 6366, 69, 70, 7275, 78101). A full list of search strategies, search results, detailed quality assessments for each included study, and tests for publication bias are available at www.ahrq.gov/clinic/tp/defibtp.htm (5). No publication bias was seen on the funnel plots. Figure 1. Flow diagram of study identification and selection. RCTs with Efficacy Data The 12 efficacy RCTs varied in quality (ranging from 1 to 3 on the Jadad scale) and duration (ranging from 15 to 66 months). All but 2 trials (16, 19) evaluated single-chamber ICDs (although no trials reported protocol adherence to single-chamber vs. dual-chamber ICDs). All patients in the RCTs had LV systolic dysfunction: Mean LVEF ranged from 0.21 to 0.28 in the primary prevention trials and from 0.32 to 0.46 in the secondary prevention trials. Most patients also had symptoms of heart failure: 50% had NYHA class II symptoms at baseline; 36%, class III symptoms; and 3%, class IV symptoms. Eleven percent of trial participants were in NYHA class I at baseline (Appendix Table 1). The mean age of RCT participants was 61 years (SD, 4), 74% were male, and 59% had ischemic heart disease. Appendix Table 1. Description of Randomized Trials Included in the Review Use of ICDs reduced all-cause mortality in patients with LV systolic dysfunction by 20% (95% CI, 10% to 29%; I 2= 44.4%) (Figure 2), largely because of a 54% relative reduction (CI, 37% to 63%; I 2= 0%) in sudden cardiac deaths. In patients with LV systolic dysfunction, ICDs were equally beneficial in reducing all-cause mortality in both primary prevention trials (RR, 0.81 [CI, 0.69 to 0.95]; I 2= 53.1% across 9 RCTs) and secondary prevention trials (RR, 0.77 [CI, 0.65 to 0.91]; I 2= 13.2% across 3 RCTs) (P for this indirect comparison= 0.56). Figure 2. Effect of implantable cardioverter defibrillator ( ICDs ) on all-cause mortality in randomized trials. A single trial included cardiac resynchronization therapy in both study groups for its comparison of ICDs versus control (19). All-cause mortality (RR, 0.83 [CI, 0.66 to 1.05]) was similar to that reported from the remainder of the studies, which did not contain cardiac resynchronization therapy in either study group (RR, 0.79 [CI, 0.69 to 0.91]) (P for indirect comparison= 0.92). Only 1 trial reported a statistically significant difference in the effect of ICDs across NYHA classes: The mortality benefits were greater in patients with NYHA class II symptoms than in those with NYHA class III symptoms in the Sudden Cardiac Death in Heart Failure Trial (P< 0.001 for interaction term of NYHA class and mortality) (22). In a series of univariate meta-regression sensitivity analyses, none of the covariates we examined (duration of follow-up, primary vs. secondary prevention, ischemic cause, presence of cardiac resynchronization therapy, NYHA class, mean age, mean LVEF, or mean QRS duration) contributed to the moderate statistical heterogeneity observed in our meta-analysis of all-cause mortality. In addition, our estimate of treatment effect was not associated with study quality. Implantable cardioverter defibrillators

Systematic review of knowledge translation strategies in the allied health professions

BackgroundKnowledge translation (KT) aims to close the research-practice gap in order to realize and maximize the benefits of research within the practice setting. Previous studies have investigated KT strategies in nursing and medicine; however, the present study is the first systematic review of the effectiveness of a variety of KT interventions in five allied health disciplines: dietetics, occupational therapy, pharmacy, physiotherapy, and speech-language pathology.MethodsA health research librarian developed and implemented search strategies in eight electronic databases (MEDLINE, CINAHL, ERIC, PASCAL, EMBASE, IPA, Scopus, CENTRAL) using language (English) and date restrictions (1985 to March 2010). Other relevant sources were manually searched. Two reviewers independently screened the titles and abstracts, reviewed full-text articles, performed data extraction, and performed quality assessment. Within each profession, evidence tables were created, grouping and analyzing data by research design, KT strategy, targeted behaviour, and primary outcome. The published descriptions of the KT interventions were compared to the Workgroup for Intervention Development and Evaluation Research (WIDER) Recommendations to Improve the Reporting of the Content of Behaviour Change Interventions.ResultsA total of 2,638 articles were located and the titles and abstracts were screened. Of those, 1,172 full-text articles were reviewed and subsequently 32 studies were included in the systematic review. A variety of single (n = 15) and multiple (n = 17) KT interventions were identified, with educational meetings being the predominant KT strategy (n = 11). The majority of primary outcomes were identified as professional/process outcomes (n = 25); however, patient outcomes (n = 4), economic outcomes (n = 2), and multiple primary outcomes (n = 1) were also represented. Generally, the studies were of low methodological quality. Outcome reporting bias was common and precluded clear determination of intervention effectiveness. In the majority of studies, the interventions demonstrated mixed effects on primary outcomes, and only four studies demonstrated statistically significant, positive effects on primary outcomes. None of the studies satisfied the four WIDER Recommendations.ConclusionsAcross five allied health professions, equivocal results, low methodological quality, and outcome reporting bias limited our ability to recommend one KT strategy over another. Further research employing the WIDER Recommendations is needed to inform the development and implementation of effective KT interventions in allied health.

Benefits and Harms of Treating Gestational Diabetes Mellitus: A Systematic Review and Meta-analysis for the U.S. Preventive Services Task Force and the National Institutes of Health Office of Medical Applications of Research

BACKGROUND Outcomes of treating gestational diabetes mellitus (GDM) are not well-established. PURPOSE To summarize evidence about the maternal and neonatal benefits and harms of treating GDM. DATA SOURCES 15 electronic databases from 1995 to May 2012, gray literature, Web sites of relevant organizations, trial registries, and reference lists. STUDY SELECTION English-language randomized, controlled trials (n = 5) and cohort studies (n = 6) of women without known preexisting diabetes. DATA EXTRACTION One reviewer extracted data, and a second reviewer verified them. Two reviewers independently assessed methodological quality and evaluated strength of evidence for primary outcomes by using a Grading of Recommendations Assessment, Development and Evaluation approach. DATA SYNTHESIS All studies compared diet modification, glucose monitoring, and insulin as needed with no treatment. Women who were treated had more prenatal visits than those in control groups. Moderate evidence showed fewer cases of preeclampsia, shoulder dystocia, and macrosomia in the treated group. Evidence was insufficient for maternal weight gain and birth injury. Low evidence showed no difference between groups for neonatal hypoglycemia. Evidence was insufficient for long-term metabolic outcomes among offspring. No difference was found for cesarean delivery (low evidence), induction of labor (insufficient evidence), small-for-gestational-age neonates (moderate evidence), or admission to a neonatal intensive care unit (low evidence). LIMITATIONS Evidence is low or insufficient for many outcomes of greatest clinical importance. The strongest evidence supports reductions in intermediate outcomes; however, other factors (for example, maternal weight and gestational weight gain) may impart greater risk than GDM, particularly when glucose levels are modestly elevated. CONCLUSION Treating GDM results in less preeclampsia, shoulder dystocia, and macrosomia; however, current evidence does not show an effect on neonatal hypoglycemia or future poor metabolic outcomes. There is little evidence of short-term harm of treating GDM other than an increased demand for services.

Testing the Newcastle Ottawa Scale showed low reliability between individual reviewers

OBJECTIVES To assess inter-rater reliability and validity of the Newcastle Ottawa Scale (NOS) used for methodological quality assessment of cohort studies included in systematic reviews. STUDY DESIGN AND SETTING Two reviewers independently applied the NOS to 131 cohort studies included in eight meta-analyses. Inter-rater reliability was calculated using kappa (κ) statistics. To assess validity, within each meta-analysis, we generated a ratio of pooled estimates for each quality domain. Using a random-effects model, the ratios of odds ratios for each meta-analysis were combined to give an overall estimate of differences in effect estimates. RESULTS Inter-rater reliability varied from substantial for length of follow-up (κ = 0.68, 95% confidence interval [CI] = 0.47, 0.89) to poor for selection of the nonexposed cohort and demonstration that the outcome was not present at the outset of the study (κ = -0.03, 95% CI = -0.06, 0.00; κ = -0.06, 95% CI = -0.20, 0.07). Reliability for overall score was fair (κ = 0.29, 95% CI = 0.10, 0.47). In general, reviewers found the tool difficult to use and the decision rules vague even with additional information provided as part of this study. We found no association between individual items or overall score and effect estimates. CONCLUSION Variable agreement and lack of evidence that the NOS can identify studies with biased results underscore the need for revisions and more detailed guidance for systematic reviewers using the NOS.

Comparative effectiveness of pain management interventions for hip fracture: a systematic review.

BACKGROUND Pain management is integral to the management of hip fracture. PURPOSE To review the benefits and harms of pharmacologic and nonpharmacologic interventions for managing pain after hip fracture. DATA SOURCES 25 electronic databases (January 1990 to December 2010), gray literature, trial registries, and reference lists, with no language restrictions. STUDY SELECTION Multiple reviewers independently and in duplicate screened 9357 citations to identify randomized, controlled trials (RCTs); nonrandomized, controlled trials (non-RCTs); and cohort studies of pain management techniques in older adults after acute hip fracture. DATA EXTRACTION Independent, duplicate data extraction and quality assessment were conducted, with discrepancies resolved by consensus or a third reviewer. Data extracted included study characteristics, inclusion and exclusion criteria, participant characteristics, interventions, and outcomes. DATA SYNTHESIS 83 unique studies (64 RCTs, 5 non-RCTs, and 14 cohort studies) were included that addressed nerve blockade (n = 32), spinal anesthesia (n = 30), systemic analgesia (n = 3), traction (n = 11), multimodal pain management (n = 2), neurostimulation (n = 2), rehabilitation (n = 1), and complementary and alternative medicine (n = 2). Overall, moderate evidence suggests that nerve blockades are effective for relieving acute pain and reducing delirium. Low-level evidence suggests that preoperative traction does not reduce acute pain. Evidence was insufficient on the benefits and harms of most interventions, including spinal anesthesia, systemic analgesia, multimodal pain management, acupressure, relaxation therapy, transcutaneous electrical neurostimulation, and physical therapy regimens, in managing acute pain. LIMITATIONS No studies evaluated outcomes of chronic pain or exclusively examined participants from nursing homes or with cognitive impairment. Systemic analgesics (narcotics, nonsteroidal anti-inflammatory drugs) were understudied during the search period. CONCLUSION Nerve blockade seems to be effective in reducing acute pain after hip fracture. Sparse data preclude firm conclusions about the relative benefits or harms of many other pain management interventions for patients with hip fracture. PRIMARY FUNDING SOURCE Agency for Healthcare Research and Quality.

A Descriptive Analysis of Overviews of Reviews Published between 2000 and 2011

Background Overviews of systematic reviews compile data from multiple systematic reviews (SRs) and are a new method of evidence synthesis. Objectives To describe the methodological approaches in overviews of interventions. Design Descriptive study. Methods We searched 4 databases from 2000 to July 2011; we handsearched Evidence-based Child Health: A Cochrane Review Journal. We defined an overview as a study that: stated a clear objective; examined an intervention; used explicit methods to identify SRs; collected and synthesized outcome data from the SRs; and intended to include only SRs. We did not restrict inclusion by population characteristics (e.g., adult or children only). Two researchers independently screened studies and applied eligibility criteria. One researcher extracted data with verification by a second. We conducted a descriptive analysis. Results From 2,245 citations, 75 overviews were included. The number of overviews increased from 1 in 2000 to 14 in 2010. The interventions were pharmacological (n = 20, 26.7%), non-pharmacological (n = 26, 34.7%), or both (n = 29, 38.7%). Inclusion criteria were clearly stated in 65 overviews. Thirty-three (44%) overviews searched at least 2 databases. The majority reported the years and databases searched (n = 46, 61%), and provided key words (n = 58, 77%). Thirty-nine (52%) overviews included Cochrane SRs only. Two reviewers independently screened and completed full text review in 29 overviews (39%). Methods of data extraction were reported in 45 (60%). Information on quality of individual studies was extracted from the original SRs in 27 (36%) overviews. Quality assessment of the SRs was performed in 28 (37%) overviews; at least 9 different tools were used. Quality of the body of evidence was assessed in 13 (17%) overviews. Most overviews provided a narrative or descriptive analysis of the included SRs. One overview conducted indirect analyses and the other conducted mixed treatment comparisons. Publication bias was discussed in 18 (24%) overviews. Conclusions This study shows considerable variation in the methods used for overviews. There is a need for methodological rigor and consistency in overviews, as well as empirical evidence to support the methods employed.

Behavioral Programs for Type 2 Diabetes Mellitus: A Systematic Review and Network Meta-analysis

In 2012, 29.1 million Americans had diabetes with costs of

Risk factors for methamphetamine use in youth: a systematic review

245 billion (1), representing 11% of the total U.S. health care expenditure (2). Although tight glycemic control may reduce the risk for microvascular complications in type 2 diabetes mellitus (T2DM) (3), behavioral and pharmacologic management of body weight, blood pressure, and cholesterol levels are often needed to reduce the risk for mortality and macrovascular complications. Moreover, other patient-centered outcomes, such as diabetes-related distress and depression, are important to address (4). Health care experts recommend that anyone with diabetes adopt and adhere to multiple self-care behaviors, including healthy eating, being active, monitoring, taking medication, problem-solving, healthy coping, and reducing risks (5). Approaches to support behavior change include diabetes self-management education (DSME) with or without an added support (clinical, behavioral, psychosocial, or educational) phase, and lifestyle programs. Because knowledge acquisition insufficiently promotes behavioral changes (6), recommendations for DSME have shifted from traditional didactic educational services to more patient-centered methodologies that incorporate interaction, problem-solving, and other behavioral approaches. Although evidence shows that diabetes-specific behavioral interventions can be effective, which combination of program components and delivery mechanisms is most effective is unclear (711). We conducted a network meta-analysis to identify factors related to program components and delivery mechanisms that moderate the effectiveness of multicomponent behavioral programs for T2DM. Methods Key informants, a technical expert panel, and public commentary informed our methods. A protocol and a peer- and public-reviewed technical report were produced for the Agency for Healthcare Research and Quality (AHRQ) and are available online (www.ahrq.gov/research/findings/evidence-based-reports/). Data Sources and Searches A research librarian searched the following bibliographic databases from 1993 to January 2015: Ovid MEDLINE (Appendix Table 1) and Ovid MEDLINE In-Process & Other Non-Indexed Citations, Cochrane Central Register of Controlled Trials via the Cochrane Library, EMBASE via Ovid, CINAHL Plus with Full Text via EBSCOhost, PsycINFO via Ovid, Scopus, and PubMed via the National Center for Biotechnology Information Databases. We reviewed the reference lists of relevant systematic reviews and of all included studies. We also searched ClinicalTrials.gov and the World Health Organization International Clinical Trials Registry Platform, relevant conference proceedings (2011 through 2014), and the U.S. Federal Register. Appendix Table 1. Search Strategy for MEDLINE* Study Selection We included studies conducted in highly developed countries published in English after 1993 (because medical management for diabetes intensified after this time). We included randomized, controlled trials done in community or outpatient health settings and involving adults that compared a behavioral program with usual care (medical management provided to all participants), an active control (intervention not meeting our definition of behavioral program), or another behavioral program (comparative effectiveness study). A behavioral program was a multicomponent, diabetes-specific program that included repeated interactions with trained individuals over at least 4 weeks, and that consisted of DSME using a behavioral approach or another program format including at least a structured dietary or physical activity intervention with another component (Appendix). We excluded abstracts and studies in which the intervention was a disease or care management program (for example, with active adjustment of diabetes-related medications) (12) or a quality improvement program incorporating strategies targeting health systems or providers (13). Other exclusion criteria included studies 1) focusing on patients with newly diagnosed (1 year) disease; 2) with no outcome of interest to this review (for example, only C-reactive protein), or in which the only difference between the study groups was a factor outside of the reviews scope (for example, low- vs. high-fat diet); and 3) in which 25% or more of the participants had type 1 diabetes mellitus (unless results were reported for those with T2DM). Two reviewers independently screened all titles and abstracts, and the full text of any publication marked for inclusion was retrieved. Two reviewers independently assessed the full texts by using a priori inclusion criteria and a standard form. We resolved disagreements by consensus or consultation with a third reviewer. Data Extraction and Quality Assessment One reviewer extracted data by using a structured form created in the Systematic Review Data Repository (available at http://srdr.ahrq.gov/) (14); a second reviewer verified data. Two reviewers independently applied the Cochrane risk of bias tool (15). Discrepancies were resolved through discussion. Data Synthesis and Analysis With input from technical experts, we categorized behavioral programs by various component and delivery factors (Table). We separated DSME and DSME plus support, in recognition that the support phase of the latter was often of lower intensity (less frequent contacts) and focused on different content, such as psychosocial support. Table. Categorization of Program Components and Delivery Factors To serve as an overview of program effectiveness and help interpret our primary analysis of program moderation, we performed pairwise meta-analyses by using the HartungKnappSidikJonkman random-effects model (16, 17) for multiple behavioral, clinical, and health outcomes, as well as health care utilization and program acceptability (the full report is available at www.ahrq.gov/research/findings/evidence-based-reports/). We defined thresholds for clinical importance where there was guidance: For hemoglobin A1c (HbA1c), we used a reduction of at least 0.4% (for example, 7.6% vs. 8.0%) (18); for quality-of-life measures and other patient-reported outcomes, we used a conservative value of one-half SD (19, 20). We then conducted a network meta-analysis that allowed simultaneous evaluation of a suite of comparisons and considered both direct and indirect evidence while preserving the within-study randomization. To assure the transitivity within the network, we categorized all behavioral programs and comparators into nodes. The nodes for behavioral programs were formed on the basis of different combinations of variables in our program categorization (Appendix Table 2); we identified all plausible nodes differing by only one variable (for example, a level within the intensity category) and then filled the nodes with the applicable interventions on the basis of our coding. The nodes for the comparator groups were categorized as usual care, active non-DSME control (education interventions not meeting our criteria), and active other control (for example, stand-alone dietary or physical activity interventions). Appendix Table 2. Characteristics of Studies of Behavioral Programs for T2DM Appendix Table 2Continued Appendix Table 2Continued Appendix Table 2Continued Appendix Table 2Continued Appendix Table 2Continued Appendix Table 2Continued Appendix Table 2Continued Appendix Table 2Continued The analysis was conducted by using a Bayesian network model to compare all interventions simultaneously and to use all available information on treatment effects in a single analysis (21, 22). These methods ensure that correlation in multigroup trials is preserved. Mean differences (MDs) were modeled using noninformative prior distributions. A normal prior distribution with mean 0 and large variance (10000) was used for each of the trial means, whereas their between study variance had a uniform prior with range 0 to 2. These priors were checked for influence with sensitivity analyses. Markov chain Monte Carlo simulations using WinBugs software were performed to obtain simultaneous estimates of all interventions compared with placebo, as well as estimates of which interventions were the best. A burn-in sample of 20000 iterations was followed by 300000 iterations used to compute estimates. A sensitivity analysis that thinned the amount of used data to every 10th iteration was also conducted to check for proper chain convergence. The analysis was checked for consistency by contrasting direct and indirect estimates in each triangular and quadratic loop by using the methods described by Veroniki and colleagues (23). Results are presented as estimates of the treatment effects (MD) relative to usual care, with 95% credible intervals. To examine different population subgroups, we conducted subgroup analyses of the pairwise meta-analysis results for HbA1c at longest follow-up in comparison with usual care and active controls; subgroups were defined on the basis of study-level baseline HbA1c (<7% vs. 7%), age (<65 vs. 65 years), and ethnicity (75 vs. <75% nonwhite), according to categories that were defined a priori. For baseline HbA1c level and age, we performed subgroup analyses of the network meta-analysis; the analysis was rerun for studies having a mean baseline HbA1c level of 7% or greater and for those with a mean participant age younger than 65 years. For subgroups based on race/ethnicity, the number of trials in either subgroup was insufficient to perform a network meta-analysis. Role of the Funding Source This project was funded under contract 290-2012-000131 from the AHRQ, U.S. Department of Health and Human Services. Staff at AHRQ participated in development of the scope of the work and reviewed drafts of the manuscript. Approval by AHRQ was required before the manuscript could be submitted for publication, but the authors are solely responsible for its content and the decision to submit for publication. AHRQ staff did not participate in the conduct of the review,