Treatment guidelines implemented into healthcare protocols are assigning patients into a respective treatment pathway based on predefined thresholds. Evidence on the benefit of these treatments comes from randomized clinical trials (RCTs) that study these effects in research settings only and do not have the statistical power to examine how effects vary by patients‘ sociodemographic characteristics and comorbidities. We will use large-scale real-world data from CPRD to study the causal effects of type 2 diabetes mellitus (T2DM) related treatment intensification protocols (intensification from monotherapy to dual therapy and dual therapy to triple therapy) in routine care and will examine in detail how these effects vary by highly granular patient subgroups. To do so, we will employ regression discontinuity (RD), a causal inference technique that is widely used in econometrics and psychology but rarely in clinical research. RD allows us to determine causal effects in observational data under minimal assumptions.
Previous studies investigating the health effects of T2DM therapy administered to T2DM patients and corresponding treatment pathways including times and conditions leading to treatment intensification are largely limited to controlled clinical trials. While these studies have shown the efficacy of T2DM therapy intensification in improving key health outcomes, they might not be able to fully capture treatment effectiveness during routine care and frequently lack the necessary time horizon to study long-run benefits and risks. Non-experimental studies, in contrast, may fail to establish causality due to insufficient control of confounding factors. This study seeks to measure the effect of treatment intensification in T2DM patients on short-, mid-, and long-term clinical outcomes (all-cause mortality, HbA1c levels and other T2DM related outcomes) in a routine care set-up. To establish causality, we make use of a regression discontinuity (RD) design taking advantage of major UK clinical guidelines (NICE) recommending different T2DM treatment pathways based on threshold rules related mainly to patients’ HbA1c level. Because physicians base their treatment decisions on additional considerations besides clinical guidelines, we use an instrumental variable approach that is robust to partial compliance. We evaluate the robustness of results by gradually narrowing down the bandwidth around the treatment threshold and thus only including patients with HbA1c levels increasingly close to the treatment threshold level. The findings of this study are expected to provide novel insights into the effectiveness of different T2DM treatment pathways in a real-life setting and can directly inform clinical practice.
Primary outcomes (all to be measured over time horizons of one year, five years, ten years, and 15 years): HbA1c and BMI levels; number of severe adverse health events (stroke, heart attack, neuropathy, retinopathy, nephropathy, lower limb amputations – each event type evaluated separately); all-cause mortality. Bandwidth of the running variable (HbA1c) around the threshold will be narrow i.e. derived from data so that treatment and control groups, as per RDD design, will be comparable and will differ only by the treatment intervention. We will use the first HbA1c measurement of the patient for the comparison group assignment and future HbA1c measurements (for time horizons as stated above) as the outcome, making HbA1c measurement a sensible outcome.
Secondary outcomes (all to be measured over time horizons of one year, five years, ten years, and 15 years): probability of at least one all-cause emergency hospitalization; number of all-cause hospitalizations; probability of at least one all-cause hospitalization; probability of at least one severe adverse health event.
Till Bärnighausen - Chief Investigator - University of Heidelberg
Hrvoje Curis - Corresponding Applicant - University of Heidelberg
Anant Jani - Collaborator - University of Surrey
Christian Bommer - Collaborator - University of Heidelberg
Duy Do - Collaborator - University of Heidelberg
Julia Lemp - Collaborator - University of Heidelberg
Justine Davies - Collaborator - University of Birmingham
Michaela Theilmann - Collaborator - University of Heidelberg
Pascal Geldsetzer - Collaborator - University of Heidelberg