BACKGROUND: A clinical trial, entitled "FREEDOM", showed that denosumab, a new anti-osteoporosis drug, was efficient in reducing the risk of having a fracture in postmenopausal women suffering from osteoporosis.
PURPOSE: We will conduct a similar study as the "FREEDOM" trial based on data from de-identified primary care records. In the trial, participants were randomly allocated to receive denosumab or placebo. Therefore, the risk for having a fracture at the beginning of the study was the same in both groups. In clinical practice, however, patients who receive denosumab may have different characteristics compared to those who are not treated for their osteoporosis. Our study will assess how good advanced statistical methods can balance these differences between groups and whether the results from our real world data study are similar to the "FREEDOM" trial.
DESIGN/METHODS: We will include patients in our study who have - to the most possible extent - the same characteristics as the "FREEDOM" participants. We will calculate the risk of having a fracture for patients with and without use of denosumab. Patient characteristics that may alter fracture risk will be assessed and 4 different statistical methods will be used to make treated and non-treated groups more comparable. We will then compare our results to the results of the "FREEDOM" trial using a set of pre-defined criteria for agreement.
POTENTIAL IMPORTANCE: Our results will inform which statistical methods should be used to balance patient characteristics in future studies when the effectiveness of different drugs shall be compared.
OBJECTIVES:
This study will aim to assess whether available analytical methods could obtain comparable findings to those from the FREEDOM Randomized Clinical Trial (RCT) in primary care real world data (RWD) using CPRD. For each method, we will specifically analyze the association between denosumab use (compared to no anti-osteoporosis drug therapy) and fracture risk amongst postmenopausal osteoporotic women meeting FREEDOM eligibility criteria.
METHODS:
Data source: CPRD-GOLD and CPRD-AURUM, linked to HES-ONS
Source Population: postmenopausal women aged 60-90, who at during the identification period (01/01/2011-31/12/2018) had evidence of osteoporosis, defined as either a history of (non-traumatic) fracture or a diagnosis of osteoporosis and were registered in CPRD for 1+ years in up-to-standard practices. Exclusion criteria will be applied to reflect FREEDOM eligibility criteria to the most possible extent.
Exposure: Denosumab 60mg for s.c. administration
Outcomes: Non-vertebral (incl. hip) fracture (primary outcome), overall fracture, clinical vertebral fracture and hip fracture (secondary outcomes).
ANALYSES:
This study will employ a cohort design. Risk for non-vertebral (incl. hip) fractures will be calculated via Cox-proportional hazard model over the course of a maximum of 36 months. The following methods to account for measures/unmeasured confounding will be applied:
1. Expert-based propensity score (PS)
o Matching
o Stratification
o Inverse probability weighting (IPW)
2. High-dimensional propensity score (HDPS)
o Matching
o Stratification
o Inverse probability weighting (IPW)
3. Disease risk score
4. Instrumental variables
For each of the methods, we will report its outcome analysis finding and compare it with the FREEDOM trial findings individually. Results from the individual datasets will then be pooled in meta-analyses, and the pooled estimate compared with the FREEDOM RCT results using a set of pre-specified criteria for agreement.
Sensitivity analyses including negative control outcomes will be conducted.
Non-vertebral fracture (incl. hip fractures); overall fracture; clinical vertebral fracture; hip fracture
Daniel Prieto-Alhambra - Chief Investigator - University of Oxford
Victoria Y Strauss - Corresponding Applicant - University of Oxford
Annika Jodicke - Collaborator - University of Oxford
Antonella Delmestri - Collaborator - University of Oxford
Eng Hooi Tan - Collaborator - University of Oxford
Joe Maskell - Collaborator - Amgen Ltd
HES Admitted Patient Care;ONS Death Registration Data;Patient Level Index of Multiple Deprivation