COLCORONA was a double-blinded randomised controlled trial of over 4000 patients comparing a 30 day course of colchicine with placebo for people with COVID-19 but not admitted to hospital. The primary endpoint was death or hospitalization for COVID-19. The trial was stopped before it had completed its planned recruitment: “This decision was made due to logistical issues related to maintaining the central study call center active 24 hours per day for a prolonged period of time, as well as the need to provide healthcare systems with study results in a timely fashion given the state of the COVID-19 pandemic.”
In the colchicine arm this primary endpoint occurred in 4.7% of patients and in the placebo arm it occurred in 5.8% of patients – odds ratio, 0.79; 95.1% confidence interval (CI), 0.61 to 1.03; P=0.08. This was driven by a reduction in hospitalizations (odds ratio, 0.75; 95% CI, 0.57 to 0.99) rather than an effect on death (odds ratio, 0.56; 95% CI, 0.19 to 1.67). A pre-specified analysis of only those patients with PCR-confirmed COVID-19 and found an odds ratio, 0.75; 95% CI, 0.57 to 0.99; P=0.04 for the primary endpoint. The pre-print abstract concludes “Among non-hospitalized patients with COVID-19, colchicine reduces the composite rate of death or hospitalization”.
The COLCORONA press release indicated that colchicine reduced deaths by 44%.
RECOVERY is an open-label randomised controlled trial platform testing multiple therapies for COVID-19. Recruitment to the colchicine arm was stopped because “the DMC saw no convincing evidence that further recruitment would provide conclusive proof of worthwhile mortality benefit either overall or in any pre-specified subgroup”. Over 11000 inpatients were randomised to colchicine or usual care. No difference in 28-day mortality was seen between the two groups – risk ratio, 1.02; 95% CI, 0.94-1.11; P=0.63.
What is driving the disparity between the trials?
One obvious difference is in the patient population. COLCORONA recruited people who were not in hospital while RECOVERY recruited those who were. Sanjay Kaul remarked: “Is it possible that interventions that don’t reduce mortality in hospitalized patients might reduce it in non-hospitalized patients? UNLIKELY!”
The more likely reason for the differing results is numbers, Kaul says, – “there is instability of data at low numbers”. In RECOVERY there were a total of 2178 deaths while in COLCORONA there were only 14 deaths. When there are so few events, there is a higher risk that differences are due to chance rather than a treatment effect.
Milton Packer explains the two trials were designed to answer different questions:
“COLCORONA studied outpatients, and its goal was to determine if colchicine would reduce hospitalizations. They recorded few deaths, since most outpatients do not die. If they had studied four times as many outpatients (or had completed the trial as planned), they would still not be able to record a meaningful number of deaths or answer a mortality question. Their primary endpoint was equivalent to a hospitalization endpoint; the upper bound of the 95% CI for hospitalization alone by ITT was 1.03. No one should refer to such a finding as a “neutral result”.
“RECOVERY studied hospitalized patients, who were sicker and at higher risk for mortality. The trial was not designed to determine if colchicine prevented hospitalizations, since all of the patients were inpatients.”
What Kaul points out is that the COLCORONA investigators “implied in the Press Release there was a possible mortality benefit that ‘approaches statistical significance’. I don’t know if ‘approaches statistical significance’ is a thing, but to draw such an inference on the basis of what can be best described as an inconclusive result is wrong. However, to do so on the basis of only 14 deaths is doubly wrong!”
Milton agrees, “Based on 14 deaths and the distribution of deaths, no reliable conclusions could be drawn. I do not understand why the COLCORONA investigators concluded that colchicine reduced mortality by 44%. That conclusion is not evidence-based”