Stanford Pragmatic Effectiveness Comparison

Overview

Chronic pain is a major healthcare problem with an annual cost of above $600 billion. The quality of data available for treatments of chronic pain is not optimal. Generalizability of explanatory randomized controlled trial data is problematic as these trials exclude up to 90% of patients: leaving out real-world patients with serious medical and psychological comorbidities. Pragmatic trials embedded in patient care compare effectiveness of currently used treatments in real-world application leading to findings that generalize to broader range of patients. The changes in clinical practice and workflow necessary to integrate this type of research within patient care present pragmatic challenges. In this research, the overall objective is to overcome these challenges using an open-source learning health care system - CHOIR. CHOIR is currently used to track patients' clinical trajectory and treatment response across multiple academic sites resulting in over 25 publications characterizing chronic pain. Through the pilot studies, the investigators have already developed a point-of-care randomization for CHOIR that facilitates integration of research and patient care by allowing the physicians to randomize patients during clinic visits. The investigators have already demonstrated feasibility of the randomization and data collection platform in two ongoing pilot pragmatic clinical trials. The investigators are proposing to better integrate pragmatic research within clinical practice through conducting a randomized comparative effectiveness trial in 450 patients with chronic pain comparing effectiveness of anti-convulsants and anti-depressants (two most commonly used classes of medications for treatment of chronic pain). The investigators will also perform a qualitative interview with all physicians in our clinic to study the impact of integrating pragmatic research into clinical care. The investigators will use the data available in CHOIR as well as the real-world data generated from this clinical trial to build, validate and test a model to predict what clinical characteristics can predict response to either of these classes of medications. The proposed study is the first step to use flexible point-of-care randomization to compare effectiveness of different treatments in different subgroups of patients whenever equipoise exists. The prediction model will guide decision making process of clinicians choosing between these medications based on clinical characteristics of individual patients.

This study aims to increase generalizability of clinical data for current treatments of chronic pain by integrating pragmatic comparative effectiveness research within patient care. Chronic pain is a major healthcare problem with annual cost of above $600 billion. The quality of data available for treatments of chronic pain is not optimal. Generalizability of explanatory randomized controlled trial data is problematic as these trials exclude up to 90% of patients: leaving out real-world patients with serious medical and psychological comorbidities. Thus, national institutions and societies (including National Institute of Health) have called for pragmatic trials embedded in patient care to compare effectiveness of currently used treatments leading to findings that generalize to broader range of patients. The long-term goal is to integrate pragmatic research within clinical pain practice; the investigators will then collect high-quality longitudinal data in the form of randomized comparative effectiveness research. This research will be a platform for development of precision health in the field of pain medicine. Personalized treatment options for each patient will not be possible without understanding how effective these treatments are in real-world application. The changes in clinical practice and workflow necessary to integrate this type of research within patient care present pragmatic challenges. In this research, the overall objective is to overcome these challenges using an open-source learning health care system - CHOIR. CHOIR is currently used to track patients' clinical trajectory and treatment response across multiple academic sites resulting in over 25 publications characterizing chronic pain. Through pilot studies, the investigators have already developed a point-of-care randomization for CHOIR that facilitates integration of research and patient care by allowing the physicians to randomize patients during clinic visits. The investigators have already demonstrated feasibility of the randomization and data collection platform in two ongoing pilot pragmatic clinical trials. To move this forward, the investigators are proposing to integrate pragmatic research within clinical practice through conducting a randomized comparative effectiveness trial in 450 patients with chronic pain pursuing the following aims: Aim 1: To compare effectiveness of anti-convulsants and anti-depressants in patients with chronic pain using point-of-care randomization. Anti-convulsants and anti-depressants are two most-commonly used classes of medications for treating chronic pain. Clinicians usually choose between them based on clinical judgement; these decisions often lack clinical evidence. The investigators need to understand how effective these medications are in real-world applications, and which patients respond better to each class of medication. The investigators will thus randomize the patients in our clinic to either anti-convulsants or anti-depressants. Randomization: The participants are going to be randomized by the treating pain provider using point-of-care randomization during the clinic visit. Point-of-care randomization will be done through CHOIR using computer generated random numbers. The randomization algorithm will disregard the patients who are randomized but decline to participate in the study to maintain a balanced number of patients in each arm. Intervention: The participants will be randomized to receiving either classes of medications. Then using a patient centered approach, the participant and their treating pain provider will choose one medication within the class the patient are randomized to. The decision will be based on participants' comorbidities, goals of care, preferences and best interest. For example, if a participant is randomized to receive an anti-convulsant and wants to avoid any weight gain, topiramate would be an appropriate choice; or if a participant is randomized to receive an anti-depressant and has poor sleep, a tricyclic anti-depressant would be an appropriate choice. Participants will continue the medications for at least 6 months provided that they can tolerate the side effects, consider the treatment effective, and are willing to continue the treatment. Titration of doses based on participants' tolerance and response is allowed based on clinical decision of the treating pain provider. Concomitant use of other analgesics or pain medications as well as pain interventions is also allowed. The investigators allow the participants to stop medications in consultation with their primary pain provider if they cannot tolerate side effects or experience no benefit. In these circumstances, we encourage the patients and their physicians to choose another medication within the class of randomization; however, the investigators will not limit the choices of participants. Data collection: The participants will receive electronic surveys similar to their routine clinical care. The investigators will send out the surveys at one, three, and six months after randomization. These surveys are going to be abbreviated to include average and worst Numerical Rating Scale (11-point NRS) of pain; pain catastrophizing scale; National Institute of Health (NIH) Patient Reported Outcomes Measurement Information System (PROMIS) measures for pain interference, depression, anxiety, and sleep disturbance; and questions about medication compliance, dose and adverse events.24-30 Data about use of other analgesics and medications, and pain interventions will also be recorded at these intervals from participants' electronic health records. If the participants do not complete the surveys within five days of receipt of the survey link, they will be contacted to be reminded of the survey. The investigators will also record participants' baseline characteristics including age, sex, main pain diagnosis, other pain diagnoses, psychological comorbidities, medical comorbidities, medication use, and education. Observation Arm: The patients who do not enter the randomization arms (patient refusal or contraindication to receive one class of medications) will be included in an observation arm after obtaining informed consent. The investigators will collect data from these patients identical to what explained above for randomized patients. Aim 2: To characterize the level of satisfaction of physicians with pragmatic comparative effectiveness trials. The investigators will perform a qualitative interview with all physicians in our clinic before recruitment phase and then every six months until the end of recruitment phase. Hypothesis. Integrating pragmatic randomized comparative effectiveness research will not affect the satisfaction of physicians in our clinic with the flow of clinical care. Aim 3: To characterize factors impacting responsiveness to anti-convulsants and anti-depressants in treating chronic pain. The investigators will use reduction in patients' reported pain as response to a treatment. The investigators will use multivariate logistic regression and machine learning algorithms to identify predictive factors for response to these two treatment arms. The investigators predict that lower socioeconomic status; longer duration of pain; widespread pain; and worse scores in pain catastrophizing, physical function, pain interference, depression, anxiety, and social function will predict lower responses to anti-convulsants and anti-depressants. Summary: Successful completion of these aims will advance the ability to: (1) use flexible point-of-care randomization to compare the effectiveness of different treatments in different subgroups of patients whenever equipoise exists; (2) apply the identified predictive factors to develop decision support systems to determine best choice of medication for individual patients; inching closer to precision health; and (3) use the feedback from our physicians to better integrate research and clinical care.