Time-related biases in studies distort allopurinol efficacy for reducing mortality in gout
Click Here to Manage Email Alerts
Although multiple observational studies in gout have reported significantly diminished mortality with allopurinol, these cannot be used as evidence due to time-related biases that exaggerate the drug’s benefits, according to data.
Meanwhile, the studies that avoided these biases found a null effect of allopurinol on mortality, the researchers added.
Data derived from Suissa S, et al. Arthritis Rheumatol. 2021;doi:10.1002/art.41710.
“Gout, the most common type of inflammatory arthritis, has been associated with increased risks of cardiovascular events and related mortality,” Samy Suissa, PhD, of the Lady Davis Institute at the Jewish General Hospital, in Montreal, and colleagues wrote in Arthritis & Rheumatology. “Treatment with allopurinol is highly effective at reducing urate levels and the frequency of flares in patients with gout. Consequently, the hypothesis that allopurinol could, via this effectiveness, also reduce the higher mortality of patients with gout has received considerable attention.”
“A meta-analysis of randomized controlled trials reporting on the comparison of allopurinol or oxypurinol with placebo found an OR of all-cause mortality of 0.94 (95% CI: 0.62-1.44), though it included trials of patients not only with gout but with other indications,” they added. “On the other hand, several observational studies have assessed this association, with hazard ratios of all-cause mortality with allopurinol use compared with non-use ranging widely between 0.39 to 1.46. The inconsistent findings of the effect of allopurinol on mortality among observational studies and with the meta-analyses call for a methodological examination of these data.”
To assess the potential for time-related biases that may explain these differences, Suissa and colleagues conducted a review of observational studies that examined the effect of allopurinol on mortality in patients with gout. The researchers initiated their literature search on Jan. 2, 2021, using the keywords “allopurinol,” “mortality,” “cohort” and “observational,” with no restriction on publication date or time.
All observational studies that analyzed allopurinol monotherapy as an exposure and all-cause mortality as an outcome were included. Meanwhile, studies that compared allopurinol with other urate lowering therapies were excluded. The researchers ultimately identified 178 potential papers. After excluding reviews, editorials, opinion pieces, meta-analyses and studies that compared allopurinol to other therapies or failed to examine allopurinol alone, 12 observational studies remained for inclusion.
According to the researchers, among the 12 studies included in the analysis, three were affected by immortal time bias while another three were affected by immeasurable time bias. The remaining six studies avoided time-related biases.
“Immortal time bias arises in cohort studies from misclassifying as ‘exposed,’ rather than ‘unexposed,’ a period of follow-up in which, by design, the study outcome cannot occur,” Suissa and colleagues wrote. “Immortal time is typically introduced when the patient’s exposure/treatment status is determined after the start of follow-up. Immeasurable time bias results from periods of time in cohort or case-control studies during which a subject cannot be recognized as being exposed to a drug because prescription records were not available.”
“For example, if exposure is based on outpatient prescription records, it will be missing during hospitalizations, thus making the patient appear unexposed,” they added. “These time-related biases tend to exaggerate the apparent benefit of drugs.”
Studies with immortal time-related biases reported reductions in all-cause mortality associated with allopurinol use, with a pooled hazard ratio of 0.71 (95% CI, 0.5-1.01) for death associated with the drug. In those with immeasurable time bias, the pooled hazard ratio was 0.62 (95% CI, 0.56-0.67). Meanwhile, studies that avoided these biases reported a null effect of allopurinol on mortality (pooled HR = 0.99; 95% CI, 0.87‐1.11).
“These observational studies cannot be used as evidence of decreased mortality with allopurinol because of known biases that tend to greatly exaggerate the potential benefit of treatments,” Suissa told Healio Rheumatology. “Other studies that avoided these biases found no such reduction in mortality with allopurinol. Allopurinol is very effective at reducing urate levels and the frequency of flares in patients with gout, but it would be misleading at this time to suggest that this treatment also improves survival.”
Perspective
Back to Top
Gout has been associated with increases in the risk of all-cause mortality, with increasing mortality, and with higher serum urate levels of between 10 to 237%. The hypothesis that allopurinol could reduce the higher mortality of patients with gout by lowering urate levels has received considerable attention. Several observational studies have assessed this association with hazard ratios of all-cause mortality with allopurinol use compared with non-use varying between 0.39 to 1.46.
Time-related biases, such as immortal and immeasurable time biases have been shown to affect many observational studies conducted using computerized health care databases. These time- related biases tend to exaggerate the apparent benefit of drugs.
In this study, the researchers evaluated 12 observational studies of the effect of allopurinol on all-cause mortality. They found that six studies were subject to major time-related biases from immortal and immeasurable time, two biases that tend to overestimate the benefit of a drug. The six studies that avoided these two biases in their design of data analysis found a null effect of allopurinol on mortality.
While observational studies are important to assess the real-world effects of medications on major outcomes, proper design and analysis are essential to minimize bias.
Perspective
Back to Top
David A. McLain, MD, MACR, FACP, FRCP
Big data through observational studies has produced numerous papers in many fields showing the benefits of specific medications on mortality, cancer and other variables not directly tied with use of the drug. For example, the use of metformin has been associated in large database studies with a lower incidence of cancer in the users.
A fault of most of these observations has been time bias. One of these is the immortal-time bias, a bias introduced with time-fixed cohort analyses that misclassify unexposed time as exposed —the treatment group has a fixed timeline and the control group has no time limit of exposure and therefore are “immortal”. There are other time biases, including time-window bias and time-lag bias, which all tend to greatly exaggerate the benefits observed with a drug. Some have even resulted in a drug being studied for a new indication, such as metformin for cancer. When these studies are recalculated eliminating the time bias, the results are often minimal or not significant.
Collapse
Read more about
Click Here to Manage Email Alerts