Trustworthy clinical practice guidelines assist health care professionals in selecting the management options that optimize patient health outcomes. The development of trustworthy guidelines requires the consideration of many aspects and the involvement of multiple contributors, often working in groups. The guideline panel plays the key role in the development process as it is responsible for prioritizing topics that should be covered as part of the guideline effort, formulating questions, reviewing the evidence, developing and agreeing on the recommendations, and endorsing the final guideline document. Ensuring transparency throughout the process by appropriately organizing and documenting panel activities is an essential standard that is used to assess the credibility of a developed guideline and its resulting recommendations. The adoption of conceptual frameworks that systematically guides panel members in their decision-making process (e.g. the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Evidence to Decision (EtD) frameworks) can aid in the formulation of methodologically sound recommendations. In this dissertation, I used the example of a guideline on diagnosis and treatment of autism spectrum disorders to describe how rigorous research methods can support guideline panels in the development process from early stages to the formulation of recommendations. In another prominent guideline development effort with the American Society of Hematology, I have identified two steps in the process where panel members may benefit from further support and addressed these gaps by conceptualizing and developing novel approaches. The first approach comprises modelling baseline risk estimates for patient-important outcomes when only surrogate data is available. The second approach proposes a method to estimate decision thresholds for judgments on health benefits and harms using the GRADE EtD framework. While these approaches are tailored to address specific guideline panel needs, guideline methodologists could use the underlying concepts to find solutions to aid guideline panels in other steps of the development process. / Thesis / Doctor of Philosophy (PhD) / Clinical practice guidelines assist health care professionals in selecting management options that can best improve the health outcomes of their patients. The development of trustworthy guidelines is a complex process that requires the contribution of several entities. The guideline panel, which typically comprises different experts (clinicians, patient representatives, experts in research methodologies) plays the key role in this process as it is responsible for selecting the most important questions to address in the guideline, reviewing the evidence supporting an option, agreeing on the recommendations, and endorsing the final guideline document. To ensure that the process of developing guidelines is transparent and that the recommendations are credible, it is important that panel activities are well documented and follow rigorous methods. Structured frameworks, such as the Grading of Recommendations Assessment, Development and Evaluation (GRADE) Evidence to Decision (EtD) approach, have been developed to systematically guide the panel members and to minimize the error that could be introduced while making decisions. In this thesis, I describe the development of an approach and its application for comprehensive guideline development by the Italian National Health Institute, to describe rigorous guideline development and propose two novel approaches to further assist panel members in enhancing their guideline development. The first of these two enhancements to guideline development describes how to derive a modelled estimate of the risk of having certain health conditions when this data is not directly available in the medical literature. The second of the two enhancements is a method to support guideline panels in judging how substantial the desirable and undesirable effects of health interventions are. Both approaches were tailored to fit specific needs but can be adapted to inform the improvement of other steps in the guideline development process.

Background: Despite the many harms and limited efficacy of opioids in managing chronic noncancer pain (CNCP), they are commonly prescribed for these patients in North America. One of the harms associated with prolonged opioid use is opioid use disorder (OUD); however, the risk of addiction is uncertain. We systematically reviewed observational studies to establish the prevalence of (OUD), and to explore factors associated with OUD in patients with CNCP. Methods: We searched MEDLINE, EMBASE, CINAHL, Cochrane Library, and PsycINFO from inception to December 2018 to identify studies that explored the prevalence of OUD or risk factors for OUD in patients with CNCP. Two specialists in addiction medicine reviewed each potentially eligible study, blinded to results, to ensure their outcome met DSM-5 criteria for OUD. We pooled estimates of OUD across eligible studies using random-effects models. When possible, we pooled estimates of association with OUD for all independent variables reported by more than one study. Results: Twenty-two studies reported the prevalence of OUD, and six studies reported the association of 36 factors with OUD in patients with CNCP. The pooled prevalence of OUD was 20% (95% CI: 15% to 25%); however, we found evidence for small study effects (interaction p<0.001). When restricted to larger studies (≥900 patients), the pooled prevalence of OUD was 5.8% (95% CI: 2.8% to 9.5%; moderate certainty evidence). The prevalence of OUD was not associated with level of certainty of OUD criteria, under- or overestimation of instruments compared to DSM-5 criteria, severity of OUD, or risk of bias (interaction p values ranged from 0.34 to 0.92). Moderate certainty evidence demonstrated an association between OUD and male sex (OR 1.50 [95% CI: 1.05 to 2.14]; absolute risk increase (ARI) 2.7% [95% CI: 0.3% more to 5.8% more]), current smokers (OR 1.63; [95% CI: 1.25 to 2.12]; ARI 3.3% [1.3% more to 5.7% more]), and a history of mental health disorders (OR 1.49 [95% CI: 1.17 to 1.89]; ARI 2.6% [95% CI: 0.9% more to 4.6% more]). Low certainty evidence demonstrated an association between OUD and younger age (OR for every 10-year decrement, 1.60 [95% CI: 1.11 to 2.30]; ARI, 3.2% for every 10-year decrement [95% CI: 0.6% more to 6.6% more]). Moderate certainty evidence suggested no association between OUD and a history of alcohol abuse/dependence (OR 1.32 [95% CI: 0.84 to 2.07]; ARI 1.7% [95% CI: 0.9% less to 5.5% more]), and low certainty evidence suggested no association between OUD and a history of drug abuse (OR 1.51 [95% CI: 0.75 to 3.02]; ARI 2.7% [95% CI: 1.4% less to 9.9% more]). Conclusion: Moderate certainty evidence suggests that 6% of CNCP patients prescribed opioids will develop OUD. Younger men who smoke, with a history of mental health disorders, are at higher risk. Additional research is needed to establish the association between OUD and a history of drug or alcohol abuse. / Thesis / Master of Science (MSc) / Opioids are commonly prescribed for patients with chronic pain that is not due to cancer; however, long-term opioid use inevitably leads to physical dependence and may result in addiction. Prior studies have reported extremely variable rates of opioid use disorder (OUD) following prescription for chronic noncancer pain, ranging from less than 1% to more than 50%, which has led to considerable confusion. My systematic review found moderate certainty evidence that the prevalence of OUD following prescription for chronic pain is 5.8% (95% CI: 2.8% to 9.5%). Patients who were younger, current smokers, males, and had a history of mental health disorders, had a higher risk of developing OUD. These findings will help support shared care decision-making between patients with chronic pain considering opioid therapy and their healthcare providers.