Table 1.
Demographic and clinical characteristics of the study population before propensity-score weighting.
Fig 1.
Cumulative incidence of death since hospital admission (inpatients) or diagnosis (outpatients) in the two treatment groups. The 21-day crude cumulative incidence of death was 7.5% In the colchicine group and 28.5% in the control group (P = 0.006; adjusted hazard ratio of death: 0.24 [95%CI: 0.09 to 0.67]). The cumulative incidence curves and number at risk at the bottom of the figure refer to the cohort after inverse probability of treatment weighting (numbers at risk are rounded to the nearest integer).
Fig 2.
Exploratory subgroup analyses.
Exploratory subgroup performed by Cox regression analyses adjusted by inverse probability of treatment weighting. The relative hazard reduction associated with colchicine was evident across different patient categories. However, in some categories the hazard ratio could not be reliably estimated because of sparse data (i.e. patients with cancer, with CKD, and receiving tocilizumab), and of no deaths in the stratum (outpatients). Solid circles represent the hazard ratio of death. Vertical bars represent 95 percent confidence intervals.
Fig 3.
A-B. Linear trajectories since hospital admission of the mean of C-reactive protein and of lymphocyte count. Linear trajectories since hospital admission of the mean of C-reactive protein (Panel A), and of lymphocyte count (Panel B), in patients who received colchicine (red) or did not receive colchicine (blue). Trajectories express the average linear trends predicted from the propensity score-adjusted random coefficients model (see text). C-reactive protein is expressed as logarithm base 2, therefore1-unitdecrease implies halving of C-reactive protein levels. Log2 C-reactive protein had a sharper decrease in the colchicine group compared to the control group (P = 0.009); lymphocyte count showed a sharper increase in lymphocyte count in the colchicine group compared to the control group (P = 0.018). Bands represent 95% confidence intervals. Data values are indicated by circles the diameter of which is proportional to the inverse probability of treatment weight based on the propensity score that was used in the all the analyses.
Fig 4.
Mean trajectory since hospital admission of log IL-6 serum levels (Panel A) in hospitalized patients who received colchicine. The dotted line is estimated based on random-coefficients mixed model with time included as polynomial variable.IL-6 serum level is expressed as logarithm base 2, therefore 1-unit decrease implie shalving of IL-6 serum levels. Bands represent 95% confidence intervals. Data values are indicated by circles the diameter of which is proportional to the inverse probability of treatment weight based on propensity score which was used in the analysis for comparison between treatment groups. The linear decreasing trend was statistically significant (P<0.001).