Abstract
Objective
The objective was to identify factors determining acute arthritis resolution and safety with colchicine and prednisone in acute calcium pyrophosphate (CPP) crystal arthritis.
Methods
We conducted a post hoc analysis of the COLCHICORT trial, which compared colchicine and prednisone for the treatment of acute CPP crystal arthritis, using a composite outcome of secondary endpoints of the primary analysis. Factors associated with the sustained arthritis resolution with prednisone or colchicine treatment on day three, and the occurrence of gastrointestinal adverse events (AEs) with colchicine, were examined. Two different machine‐learning approaches were used to identify independent factors associated with the efficacy outcome: multiple logistic regressions and a decision tree were used.
Results
In total, 43 of 89 (48.3%) of participants were considered definite responders. In univariable analysis, definitive responders were more often hospitalized for stroke (9 of 43; P = 0.04), had an age ≥80 years old (39 of 43; P = 0.045), and were dyslipidemic (25 of 43; P = 0.03), whereas poor responders were more commonly hospitalized to manage the acute arthritis episode (22 of 43; P < 0.001) In multiple logistic regression, acute arthritis of the wrists (odds ratio [OR] 4.06, 95% confidence interval [CI] 1.21–15.50.85) were associated with arthritis resolution on day three, whereas randomization in the colchicine arm (OR 0.31, 95% CI 0.11–0.83) and diuretic use (OR 0.23, 95% CI 0.097–0.95) were associated with a poor treatment response. Hospital admission for acute arthritis, C‐reactive protein levels, and estimated glomerular filtration rate were decision tree nodes selected as crucial for predicting definitive flare resolution. Three candidate variables were identified in the multiple logistic regression model explaining the occurrence of gastrointestinal AEs with colchicine: male sex (OR 0.33, 95% CI 0.07–1.29) and diabetes (OR 0.24, 95% CI 0.030–1.24) seemed protective, whereas statin use (OR 3.54, 95% CI 0.83–18.82) seemed associated with their occurrence.
Conclusion
This study identified factors associated with the definitive response to colchicine and prednisone treatment in acute CPP crystal arthritis. Colchicine was associated with poorer efficacy and was impacted by a dual effect on its safety and efficacy by medications and associated conditions.
INTRODUCTION
Calcium pyrophosphate deposition (CPPD) disease occurs as a consequence of the inflammatory response and cartilage damage because of the pathologic presence of calcium pyrophosphate (CPP) crystals inside joints. 1 , 2 , 3 , 4 , 5 , 6 , 7 It is an umbrella term for different acute and chronic inflammatory and noninflammatory phenotypes, which often coexist. 7 , 8 , 9 Acute CPP crystal arthritis is the most frequently recognized phenotype and is characterized by severe joint pain and swelling, like gout flares, commonly affecting the knees, wrists, and ankles, and lasting for several days to weeks. 10 , 11 , 12 , 13 Acute CPP crystal arthritis is the most common cause of acute arthritis in older adults, and frequently occurs during in‐hospital stays, and its management is at the cross‐roads of internal medicine, geriatrics, rheumatology, and general practice. 1 , 12 The objective of the management of these acute flares of CPP crystal arthritis is to obtain rapid and sustained pain relief, mainly by controlling joint inflammation, without producing adverse events (AEs) in the fragile target population of older people. 1 Anti‐inflammatory strategies for the treatment of acute CPP crystal arthritis can rely on colchicine, prednisone, nonsteroidal anti‐inflammatory drugs, intraarticular glucocorticoid injections, and, in some difficult cases, targeted therapies, mainly interleukin‐1 (IL‐1) or IL‐6 inhibitors. 1 No guidance has yet been provided on which drug to choose among these options for the treatment of acute CPP crystal arthritis, mainly because of a lack of comparative evidence of efficacy and safety but also because no factors predicting treatment response or safety have been clearly identified so far. 1 , 3 , 14 In particular, how the specifics of older people, including decreased renal function, age, and comedications, impact the efficacy and safety of drugs targeting inflammation, remains unknown and adds to the difficulties of managing acute CPP crystal arthritis for physicians.
SIGNIFICANCE & INNOVATIONS.
This post hoc analysis of the first trial performed in acute CPP crystal arthritis identified factors (demographic, acute arthritis characteristics, associated conditions [including underlying osteoarthritis], and medications) predictive of arthritis resolution with prednisone and colchicine.
Using for this post hoc analysis a different primary endpoint than the one used for the primary analysis of the COLCHICORT trial, prednisone provides more acute CPP crystal arthritis resolution after two days of treatment than colchicine.
Associated conditions and medications have a dual inverse effect on colchicine's efficacy and safety.
The COLCHICORT trial was the first‐ever completed randomized controlled trial studying acute CPP crystal arthritis and compared the efficacy after the first day of treatment and safety of a two‐day regimen of colchicine or prednisone. 11 The drugs provided equivalent pain relief on day two, with a safety profile favoring prednisone because of the occurrence of gastrointestinal AEs, mainly diarrhea, in 22% of participants in the colchicine arm. The first analyses provided hints that colchicine had better response rates when administered within 12 hours of arthritis onset (odds ratio [OR] 4.0, 95% confidence interval [CI] 0.7–21.0), that statin use may increase the risk of colchicine‐induced diarrhea (OR 3.0, 95% CI 0.9–10.6), and that prednisone use may require less additional treatments to obtain definitive flare resolution. 11 Pain relief after the second day of treatment was a secondary outcome and did not significantly differ between treatment groups. Patients with colchicine seemed to require more additional anti‐inflammatory drugs in the follow‐up than those randomized in the prednisone group. The objective of this post hoc analysis was to identify factors associated with acute CPP crystal arthritis resolution sustained after day three with treatments with colchicine and prednisone and their safety, to help choose between available drugs.
PATIENTS AND METHODS
Participants and design
We conducted a post hoc analysis of the COLCHICORT equivalence randomized controlled trial, which compared colchicine and prednisone for the treatment of acute CPP crystal arthritis, and showed equivalence between the two drugs in obtaining pain relief at 24 hours (day two) (NCT03128905). 11 Detailed methods and results of the full trial have been reported previously. 11 In short, this was a 7‐day, open‐label equivalence trial with participants randomized 1:1 to colchicine 1.5 mg on day one and 1 mg on day two or prednisone 30 mg on day one and two. Systematic analgesia was given in the first 24 hours, with 50 mg tramadol and 1 g paracetamol every 8 hours. After the second dose of the experimental drugs on day two, investigators were free to decide how to manage the episode until the last assessment on day eight, including using additional doses of colchicine, prednisone, and intraarticular glucocorticoid injections. The primary endpoint was the change in visual analog scale (VAS) pain (0–100) from baseline on day two (24 hours). Ninety‐five patients were included in the analysis of the per protocol (PP) population. Secondary endpoints included the change in VAS pain from baseline to day three (48 hours) and the need for additional analgesics and anti‐inflammatory drugs (additional colchicine/prednisone doses or intraarticular glucocorticoid injections).
All participants presented with acute CPP crystal arthritis with a symptom duration of <36 hours and pain intensity at ≥40 mm (/100 mm) on a VAS for the most painful joint. Diagnosis was ascertained by identifying CPP crystals on synovial fluid aspiration or with imaging evidence of CPPD, 15 , 16 and all of them retrospectively met the newly developed American College of Rheumatology/EULAR classification criteria for CPPD disease. 10
The trial was sponsored by the 2016 Northwest GIRCI French Inter‐regional Hospital Clinical Research Programme (PHRC no. 16.31). The trial protocol was reviewed and approved by a French national institutional review board (CPP Nord‐Ouest number 17/43, registration number 2016, ANSM number 170356A‐21). All patients provided written informed consent to participate in the study and the study complied with the French Data Protection Authority (CNIL)‘s MR001 reference methodology.
Outcomes of treatment response and safety
In this post hoc analysis, the main efficacy outcome was a composite outcome of secondary endpoints of the primary analysis and was defined as a change on the pain VAS from baseline of ≥50% and/or a pain VAS <40/100 by day three (after the full two days of the experimental phase) and no further use of colchicine, prednisone, or any other systemic anti‐inflammatory agent and no intraarticular glucocorticoid injection.
The safety outcome examined was the occurrence of any gastrointestinal AE (diarrhea, abdominal pain, nausea, or vomiting) in participants randomized in the colchicine arm. No other AEs were sufficiently frequent to examine risk factors.
Variables
In the current analysis, a number of baseline variables were examined, including (1) general demographic and clinical variables such as age, sex, body mass index, blood pressure, presence of comorbidities, and estimated glomerular filtration rate (eGFR); (2) clinical and biologic characteristics of the acute CPP crystal arthritis (affected joint(s), pain VAS, duration since onset, cause of hospital admission, arthrocentesis performed or not, prior episode of acute arthritis, C‐reactive protein (CRP) level, neutrophil and monocyte counts); (3) conventional radiography evidence of CPPD and osteoarthritis (OA) in specific and symptomatic sites; (4) current drugs; and (5) randomized group (colchicine vs prednisone).
Statistical analysis
In the absence of any prior data orienting the selection of baseline characteristics that may or may not have been associated with treatment response, we chose to use an unoriented automatic selection of variables, which is otherwise avoided. 17 These post hoc analyses had not been prespecified in the original protocol. Data exploration and statistical analysis were conducted using the R programming language (R version 4.3.0).
Descriptive analysis provided means and standard deviations for normally distributed quantitative variables, and medians and interquartile ranges (IQRs) for nonnormally distributed quantitative variables. The normality of the data was verified using the Shapiro‐Wilk test. Counts and frequencies are presented for qualitative variables. Some continuous variables were dichotomized. The best threshold value for these variables was chosen from the receiver operating characteristic curve according to the Youden index (maximum sum of sensitivities [Se] + specificities [Sp]). Univariable analyses for the efficacy outcome, first combining both treatments and then for each treatment arm individually, and for the safety outcome of colchicine were performed using the chi‐squared or Fisher exact tests for qualitative data according to the Cochran rule and with Student's t tests or Wilcoxon‐Mann‐Whitney tests for quantitative data according to data normality using the Shapiro‐Wilk test. Unadjusted ORs and their 95% CIs were calculated. These univariable analyses were performed to inform the multivariable analysis and were therefore not adjusted for multiple comparisons. Two complementary multivariable models, a multiple logistic regression model with stepwise backward elimination and a decision classification and regression tree (CART) model, were fitted for the efficacy outcome of the treatments combined. An insufficient sample size in the colchicine group alone prevented a consistent decision tree for the occurrence of any gastrointestinal AE with colchicine and multivariable models for the efficacy outcome of individual treatments.
Variables with a between‐group difference associated with a P value of <0.20 or with a ≥10‐point difference, and with <10% missing values, and with each binary condition present in >10% of the sample, were entered into a multiple logistic regression model to which a stepwise backward elimination based on Akaike information criterion (AIC) was applied to obtain a reduced model. Multivariable ORs and their 95% CI of the reduced model were calculated.
Additionally, the predictive performance of the “traditional” logistic regression model was supplemented by a decision tree approach using the CART algorithm with the rpart package in R, based on the Gini impurity index. 18 The choice of that alternative is justified by many advantages, namely nonparametric modeling, interpretability, and visualization. The CART algorithm was used to search for the split on each variable to partition the data into two groups depending on a binary response (treatment response and no treatment response). The CART method iteratively split data into subgroups (nodes) using a hierarchical strategy. That process yields a tree, and the subgroups in the final partition (the leaves) are associated with predictions given by the majority response within each leaf. The CART method identifies the best overall split (partition) optimizing a loss function based on the impurity criterion and under certain constraints (eg minimal size of leaves, etc). Similarly to the AIC for variable selection in logistic regression, pruning was then applied to generate a parsimonious tree model using the “cost complexity” method (ie, eliminating leaves that do not significantly increase accuracy in order to prevent overfitting). Within each (intermediate and final) node, the tree plot provides classes of patients, labeled with the dominant response modality, the percentage of participants classified as good responders, and the percentage of the sample they represent.
For both multivariable methods, all patients were considered in the training set without test set, and unbiased logistic regression and tree performances were calculated with leave‐one‐out cross validation (LOOCV); the error rate (misclassification) and Se and Sp (based on a 50% probability cutoff) were calculated. Statistical significance was set at P < 0.05.
Patient consent for publication
All patients provided written informed consent to participate in the study and the study complied with the French Data Protection Authority (CNIL)‘s MR001 reference methodology.
Ethics approval
The trial protocol was reviewed and approved by a French national IRB (CPP Nord‐Ouest number 17/43, registration number 2016, ANSM number 170356A‐21).
RESULTS
Participants
The PP population of the COLCHICORT trial included 95 patients, 46 of whom were treated with prednisone and 49 of whom were treated with colchicine. A total of 89 participants had pain assessment on day two and were included in this analysis, 43 from the prednisone group and 46 from the colchicine group. The characteristics of the participants are described in Table 1. Briefly, the median age was 88 years (IQR 82–91; 26 participants (27.4%) were male, 24 participants (25%) had diabetes, and 76 participants (80%) had hypertension. The median eGFR was 62.1 mL/min per 1.73 m2 (IQR 45.2–81.5). Acute CCP crystal arthritis with a first‐ever flare was reported in 70 participants (73.7%), and the acute arthritis affected mainly the knees in 46 participants (48%), the wrists in 19 participants (20%), and the ankles in 12 participants (13%) (Table 1).
Table 1.
Patients’ characteristics according to the combined response to treatment with colchicine and prednisone (change in pain VAS from baseline of ≥50% and/or pain VAS <40/100 by day three (48 hours) and no further use of anti‐inflammatory agents)*
| Baseline characteristics | Per protocol population (n = 95) | No definitive flare resolution (48 h) (n = 46) | Definitive flare resolution (48 h) (n = 43) | P value |
|---|---|---|---|---|
| Anthropometry | ||||
| Age, median (IQR), years | 88 (82–91) | 88 (79–91) | 87 (84.5–90.5) | 0.63 |
| Age, >80 years, n (%) | 72 (80.9) | 33 (71.7) | 39 (90.7) | 0.045 |
| Male, n (%) | 26 (27.4) | 9 (19.6) | 13 (30.2) | 0.36 |
| Body mass index, mean (SD) | 25.7 (5.47) | 26.4 (6.3) | 25 (4.6) | 0.24 |
| Body mass index >25, n (%) | 48 (50.5) | 25 (54.3) | 21 (48.8) | 0.76 |
| SAP >124 mm Hg, n (%) | 58 (61.7) | 31 (68.9) | 23 (53.5) | 0.21 |
| Comorbidities | ||||
| High blood pressure, n (%) | 76 (80) | 37 (80.4) | 35 (81.4) | 1 |
| Type 2 diabetes, n (%) | 24 (25.3) | 12 (26.1) | 12 (27.9) | 1 |
| Dyslipidemia, n (%) | 44 (46.3) | 15 (32.6) | 25 (58.1) | 0.027 |
| Current treatments, n (%) | ||||
| Statins | 28 (29.5) | 12 (26.1) | 15 (34.9) | 0.5 |
| Antidiabetics | 16 (16.8) | 9 (19.6) | 7 (16.3) | 0.9 |
| Diuretics | 28 (29.5) | 17 (37) | 10 (23.3) | 0.24 |
| ACE/RAA2 inhibitors | 26 (27.4) | 11 (23.9) | 12 (27.9) | 0.85 |
| Beta blockers | 48 (50.5) | 23 (50) | 25 (58.1) | 0.58 |
| Calcium channel inhibitors | 21 (22.1) | 8 (17.4) | 12 (27.9) | 0.35 |
| Proton pump inhibitors | 32 (33.7) | 15 (32.6) | 16 (37.2) | 0.82 |
| Anticoagulants | 65 (68.4) | 30 (65.22) | 32 (74.42) | 0.48 |
| Characteristics of the acute CPP crystal arthritis episode | ||||
| Initial cause of hospital admission, n (%) | ||||
| Acute arthritis | 26 (27.4) | 22 (47.8) | 4 (9.3) | 0.00017 |
| Trauma | 28 (29.5) | 12 (26.1) | 14 (32.6) | 0.66 |
| Stroke | 11 (11.6) | 2 (4.3) | 9 (20.9) | 0.04 |
| General deterioration in health | 12 (12.6) | 4 (8.7) | 6 (14) | 0.51 |
| Infection dyspnea | 7 (7.4) | 3 (6.5) | 3 (7) | 1 |
| Prior episode of acute CPP crystal arthritis, n (%) | 25 (26.3) | 10 (21.7) | 13 (30.2) | 0.5 |
| Painful joint, n (%) | ||||
| Shoulder | 7 (7.4) | 3 (6.5) | 3 (7) | 1 |
| Elbow | 4 (4.2) | 3 (6.5) | 1 (2.3) | 0.62 |
| Hip | 0 (0) | 0 (0) | 0 (0) | — |
| Wrist | 21 (22.1) | 6 (13) | 13 (30.2) | 0.086 |
| Knee | 51 (53.7) | 29 (63) | 19 (44.2) | 0.12 |
| Ankle | 20 (21.1) | 12 (26.1) | 7 (16.3) | 0.38 |
| Cervical spine (CDS) | 8 (8.4) | 5 (10.9) | 3 (7) | 0.71 |
| Metacarpophalangeal joint | 5 (5.3) | 2 (4.3) | 3 (7) | 0.67 |
| Pubic symphysis | 0 (0) | 0 (0) | 0 (0) | — |
| Other | 3 (3.2) | 2 (4.3) | 1 (2.3) | 1 |
| Baseline pain VAS >70/100 (mm), n (%) | 41 (46.1) | 22 (47.8) | 19 (44.2) | 0.89 |
| More than one painful joint, n (%) | 18 (18.9) | 10 (21.7) | 7 (16.3) | 0.7 |
| Duration of the acute arthritis episode <12 hours, n (%) | 32 (34) | 14 (31.1) | 14 (32.6) | 1 |
| Arthrocentesis performed, n (%) | 43 (45.3) | 22 (47.8) | 19 (44.2) | 0.9 |
| Randomization arm, n (%) | ||||
| Prednisone | 46 (48.4) | 18 (39.1) | 25 (58.1) |
0.11 |
| Colchicine | 49 (51.6) | 28 (60.9) | 18 (41.9) | |
| Laboratory results | ||||
| C‐reactive protein, median (IQR), mg/L | 80 (39–143.5] | 91.5 (56.8–148.2) | 75 (32–132) | 0.23 |
| eGFR < 60 mL/min/1.73 m2, n (%) | 41 (48.8) | 20 (45.5) | 21 (52.5) | 0.67 |
| Hypomagnesemia | 6 (9.5) | 1 (3.4) | 5 (17.2) | 0.19 |
| Neutrophils >10,000/mm3, n (%) | 14 (17.1) | 6 (14) | 8 (20.5) | 0.62 |
| Monocytes >1,000/mm3, n (%) | 39 (47.6) | 19 (44.2) | 20 (51.3) | 0,67 |
| Imaging, n (%) | ||||
| CPPD in the index joint | 84 (94.4) | 42 (95.5) | 36 (92.3) | 0.66 |
| Wrist CPPD | 58 (73.4) | 29 (72.5) | 28 (80) | 0.63 |
| Knee CPPD | 49 (65.3) | 21 (61.8) | 25 (69.4) | 0.67 |
| Pubic symphysis CPPD | 34 (41.5) | 14 (35.9) | 18 (48.6) | 0.37 |
| Osteoarthritis in the index joint | 51 (60.7) | 25 (59.5) | 23 (63.9) | 0.87 |
| Metacarpophalangeal joint osteoarthritis | 26 (32.9) | 14 (35) | 12 (34.3) | 1 |
| Scaphotrapeziotrapezoidal joint osteoarthritis | 28 (35.4) | 15 (37.5) | 12 (34.3) | 0.96 |
| Wrist osteoarthritis | 29 (36.7) | 12 (30) | 17 (48.6) | 0.16 |
| Knee osteoarthritis | 34 (45.3) | 14 (41.2) | 20 (55.6) | 0.34 |
Univariable comparisons between responders and nonresponders were performed. ACE, angiotensin‐converting enzyme; CDS, crowned dens syndrome; CI, confidence interval; CPP, calcium pyrophosphate; CPPD, CPP deposition; eGFR, estimated glomerular filtration rate; IQR, interquartile range; RAA2, renin‐angiotensin antagonist; SAP, systolic arterial pressure; VAS, visual analog scale.
Predictive factors of sustained arthritis resolution after day three with colchicine or prednisone
A total of 31 participants (72.1%) in the prednisone group and 30 participants (65.2%) in the colchicine group were considered good treatment responders on day three (P = 0.64). Overall, 51 of 89 patients (57.3%) did not require any additional anti‐inflammatory treatments (colchicine/prednisone/intraarticular glucocorticoid injections) after 48 hours, and the flare was considered definitively resolved on day three. In total, 43 of 89 (48.3%) of participants were considered definite responders. In univariable analysis, definitive responders were more often hospitalized for stroke (9 of 43; P = 0.04), had an age ≥80 years old (39 of 43; P = 0.045) and were dyslipidemic (25 of 43; P = 0.03), whereas poor responders were more commonly hospitalized to manage the acute arthritis episode (22 of 43; P < 0.001; Table 1).
The 89 participants with complete data were included in the logistic regression model, which examined all relevant variables in the univariable analysis except eight variables, which had >10% of missing data: magnesium levels, calcium levels, transferrin levels, OA of the symptomatic joint, and conventional radiography evidence of CPPD and OA at the knees and wrists. History of stroke, age ≥80 years old, and admission for acute arthritis, all significantly associated with treatment response in univariable analysis, could not be included in the model because one binary condition represented <10% of the population. The model retained five explanatory variables, three of which were significantly associated with treatment response: acute arthritis affecting wrists (OR 4.06, 95% CI 1.21–15.50) was associated with definitive flare resolution at 48 hours in the reduced model, whereas randomization in the colchicine arm (OR 0.31, 95% CI 0.11–0.83) and diuretic use (OR 0.32, 95% CI 0.097–0.95) were negatively associated with definitive flare resolution (Figure 1). The error rate was 36% in the training set and 41.6% in the LOOCV, associated with an Se of 48.8% and Sp of 67.4%.
Figure 1.
Forest plot of independent factors selected in the reduced multiple logistic regression explaining definitive flare resolution by day three defined by a change in pain VAS from baseline of ≥50% and/or pain VAS <40/100 by day three (48 hours) and no further use of anti‐inflammatory agents. Odds ratios and their 95% CI of variables selected in the reduced model are presented. CI, confidence interval; VAS, visual analog scale.
The pruned decision tree obtained with the CART algorithm integrated threshold values for CRP and eGFR and being hospitalized for the acute CPP arthritis or not (Figure 2). The best‐case scenario was for participants hospitalized for other reasons than the acute CPP arthritis episode, with CRP levels of ≥88 mg/L and eGFR <43 mL/min/1.73 m2, which represented 9% of patients, and 86% had a definitive flare resolution after 48 hours of treatment. The worst‐case scenario was for participants hospitalized because of their acute CPP arthritis episode. They represented 31% of participants and were rapid responders in 17% of cases. The rate of participants wrongly classified by the tree was 21.8% in the training set and 26.9% in the LOOCV, associated with an Se of 64.9% and Sp of 80.5%. This tree correctly classifies two of three definitive responses. The predictive performance of the decision tree was better than that of the regression model, although a little less sensitive.
Figure 2.
Illustration of the decision tree model generated by CART algorithm for definitive flare resolution by day three defined by a change in pain VAS from baseline of ≥50% and/or pain VAS <40/100 by day three (48 hours) and no further use of anti‐inflammatory agents. The percentage of treatment responders and the percentage of patients from the overall sample are shown in each box representing a tree node, after the name of the dominant class in each node. At each split, an observation goes to the left branch if the condition is satisfied and goes on the right otherwise. The rate of participants wrongly classified by the tree was 21.8% in the training set and 26.9% in the leave‐one‐out crossvalidation. CART, classification and regression tree; CRP, C‐reactive protein; eGFR, estimated glomerular filtration rate; VAS, visual analog scale.
Factors associated with sustained arthritis resolution after day three with colchicine and prednisone individually
The unadjusted univariable analysis of response to each treatment individually is presented in Supplementary Table 1. The context of the hospital admission for the acute arthritis was associated with a poor response to both drugs (OR 0.05, 95% CI 0.00–0.30, for colchicine; and OR 0.21, 95% CI 0.04–0.93, for prednisone), whereas hospital admission for stroke was associated with a good treatment response in all patients treated with prednisone but did not affect response to colchicine (OR 2.60, 95% CI 0.39–21.48). Coprescribed diuretics tended to be associated with a lower response to both colchicine (OR 0.18, 95% CI 0.01–1.13) and prednisone, to a lesser extent, (OR 0.45, 95% CI 0.13–1.53) but did not reach statistical significance. Participants treated with antidiabetic drugs seemed less likely to be responders to prednisone (OR 0.35, 95% CI 0.06–1.68) than those treated with colchicine (OR 1.71, 95% CI 0.36–8.33), and a diagnosis of type 2 diabetes tended to favor a response to colchicine rather than to prednisone. Calcium channel blockers were associated with a good response to colchicine (OR 3.68, 95% CI 0.99–15.01) but did not affect response to prednisone (OR 0.95, 95% CI 0.18–5.43). Responders to colchicine had a lower eGFR compared to nonresponders, (53.3 mL/min/1.73 m2 [IQR 36.8–82] vs 73.7 mL/min/1.73 m2 [IQR 56.2–86.4]; P = 0.07), whereas renal function was consistent between responders and nonresponders of the prednisone group (P = 0.54). Acute CPP crystal arthritis in the knees and ankles had lower response rates with colchicine (OR 0.28, 95% CI 0.08–0.94 and OR 0.18, 95% CI 0.01–1.13, respectively), whereas the efficacy of prednisone did not seem to be affected (OR 0.69, 95% CI 0.20–2.34 and OR 0.82, 95% CI 0.20–3.39, respectively). Although the presence of OA in any joints did not affect the response to colchicine, OA in the knees (OR 20.43, 95% CI 3.05–415.00) and wrist (OR 6.50, 95% CI 1.52–35.60) was associated with a good treatment response to prednisone. Responders in the prednisone group more often had OA in the index joint (64% vs 42%, non significant [NS]) contrary to the colchicine group (64% vs 72%, NS). Imaging evidence of CPPD in the index joint or any other joint did not affect treatment for either of the two drugs.
Because of an insufficient sample size for each treatment taken individually, no multivariable analysis was performed to examine separate associations with treatment response.
Factors associated with gastrointestinal AEs with colchicine
Among the participants randomized in the colchicine arm, 20 of 55 (36.4%) experienced at least one gastrointestinal AE. In the univariable analysis, already having digestive disorders at baseline (3 of 20; P = 0.04) and treatment with diuretics (7 of 20; P = 0.03) were the only variables significantly associated with gastrointestinal AEs (Table 2). The following variables were included in the multiple logistic regression model, which included 49 participants with complete data: sex, diabetes, dyslipidemia, statin use, diuretic use, beta‐blocker use, and anticoagulant use. Three explanatory variables were eventually retained in the model, none of which were statistically significantly associated with the occurrence gastrointestinal AEs related to colchicine: male sex (OR 0.33, 95% CI 0.07–1.29) and diabetes (OR 0.24, 95% CI 0.030–1.24) seemed protective, whereas statin use (OR 3.54, 95% CI 0.83–18.82) seemed to be associated with the AE (Figure 3).
Table 2.
Patients’ characteristics according to the occurrence of gastrointestinal adverse events in the colchicine group*
| Baseline characteristics | Colchicine arm (n = 55) | No gastrointestinal adverse event (n = 35) | Gastrointestinal adverse events (n = 20) | P value |
|---|---|---|---|---|
| Anthropometry | ||||
| Age >80, median (IQR), years | 41 (74.6) | 25 (71.4) | 16 (80) | 0.70 |
| Male, n (%) | 17 (30.9) | 14 (40) | 3 (15) | 0.10 |
| Body mass index ≥25, n (%) | 26 (47.3) | 17 (48.6) | 9 (45) | 1 |
| Comorbidities | ||||
| High blood pressure, n (%) | 45 (81.8) | 26 (74.3) | 19 (95) | 0.075 |
| Type 2 diabetes, n (%) | 13 (23.6) | 10 (28.6) | 3 (15) | 0.33 |
| Dyslipidemia, n (%) | 24 (43.6) | 14 (40) | 10 (50) | 0.66 |
| Current treatments, n (%) | ||||
| Statins | 16 (29.1) | 8 (22.9) | 8 (40) | 0.3 |
| Anti‐diabetics | 8 (14.5) | 5 (14.3) | 3 (15) | 1 |
| Diuretics | 10 (18.2) | 3 (8.6) | 7 (35) | 0.026 |
| ACE/RAA2 inhibitors | 15 (27.3) | 10 (28.6) | 5 (25) | 1 |
| Beta blockers | 27 (49.1) | 19 (54.3) | 8 (40) | 0.46 |
| Calcium channel inhibitors | 14 (25.5) | 9 (25.7) | 5 (25) | 1 |
| Proton pump inhibitors | 15 (27.3) | 10 (28.6) | 5 (25) | 1 |
| Antibiotics | 9 (16.4) | 6 (17.1) | 3 (15) | 1 |
| Laxatives | 17 (30.9) | 10 (28.6) | 7 (35) | 0.85 |
| Antinausea | 3 (5.5) | 2 (5.7) | 1 (5) | 1 |
| Anticoagulants | 42 (76.4) | 25 (71.4) | 17 (85) | 0.33 |
| Characteristics of the acute CPP crystal arthritis episode | ||||
| Initial cause of hospital admission, n (%) | ||||
| Acute arthritis | 16 (29.1) | 11 (31.4) | 5 (25) | 0.84 |
| Trauma | 12 (21.8) | 8 (22.9) | 4 (20) | 1 |
| Stroke | 5 (9.1) | 4 (11.4) | 1 (5) | 0.64 |
| General deterioration in health | 5 (9.1) | 2 (5.7) | 3 (15) | 0.34 |
| Infection, dyspnea | 4 (7.3) | 2 (5.7) | 2 (10) | 0.62 |
| Prior episode of acute CPP crystal arthritis | 15 (27.3) | 9 (25.7) | 6 (30) | 0.98 |
| Baseline symptoms | ||||
| Digestive disorders | 3 (5.5) | 0 (0) | 3 (15) | 0.043 |
| Anxiety/insomnia/delirium | 7 (12.7) | 5 (14.3) | 2 (10) | 1 |
| Baseline pain VAS ≥70/100, mm | 24 (43.6) | 16 (45.7) | 8 (40) | 0.90 |
| Imaging, CPPD in the index joint, n (%) | 47 (95.9) | 29 (93.5) | 18 (100) | 0.53 |
| Laboratory results | ||||
| eGFR ≤‐à mL/min/1.73 m2, n (%) | 22 (43.1) | 12 (35.3) | 10 (58.8) | 0.19 |
| Serum magnesium <0.66 mmol/L, n (%) | 3 (8.3) | 2 (8.3) | 1 (8.3) | 1 |
Univariable comparisons between responders and nonresponders were performed. ACE, angiotensin‐converting enzyme; CPP, calcium pyrophosphate; CPPD, CPP deposition; eGFR, estimated glomerular filtration rate; IQR, interquartile range; RAA2, renin‐angiotensin antagonist; VAS, visual analog scale.
Figure 3.
Forest plot of independent factors selected in the reduced multiple logistic regression explaining the occurrence of at least one gastrointestinal adverse event (diarrhea, abdominal pain, nausea, or vomiting) with colchicine. ORs and their 95% CIs of variables selected in the reduced model are presented. CI, confidence interval; OR, odds ratio.
DISCUSSION
This post hoc analysis provides new factors associated with a new definition of treatment response after a two‐day regimen of colchicine or prednisone for acute CPP crystal arthritis. These factors include characteristics of the acute flare (site and inflammatory biomarkers), of the patients (age, gender, and renal function), evidence of OA, and current drugs. Some of these factors were shared by both treatments, and some were treatment‐specific. Contrary to the analysis of the primary efficacy outcome of rapid pain relief on day two in the COLCHICORT trial, 11 the treatment arm did influence sustained arthritis resolution after day three, favoring prednisone in the logistic regression model. The study also provides new factors associated with the onset of gastrointestinal AEs with colchicine treatment. In most cases, these factors were also those associated with a good response.
The study identified new factors associated with sustained treatment response to colchicine and prednisone for the treatment of acute CPP crystal arthritis. The first analyses of the COLCHICORT trial showed that treatment administered within 12 hours after flare onset was associated with a better response to colchicine, similarly as it had been shown in the AGREE trial in gout flares but did not impact the response to prednisone. 11 , 19 However, the rapidity of colchicine administration did not seem to impact flare resolution on day three in this new analysis. This post hoc analysis provides new profiles of patients more likely to be treatment responders. Overall, patients with nonsevere CPP crystal arthritis affecting the wrist rather than the knee, with lower levels of markers of inflammation and pain intensity and impaired renal function, had a higher chance of having their pain sufficiently relieved by day three without any need for additional anti‐inflammatory agents. Flares affecting knees and ankles seemed to be less responsive to colchicine, whereas the type of joint involved did not affect prednisone efficacy. In addition, the presence of OA (in wrists and knees) was highly associated with a good treatment response to prednisone but did not affect response to colchicine. This differential response between the two drugs in case of underlying OA is consistent with the negative results of colchicine in chronic symptoms of OA, whereas prednisolone has been shown to relieve symptoms of hand OA. 20 , 21 , 22 , 23 Patients treated with calcium channel blockers seemed to be better responders, whereas those treated with diuretics more rarely had a sustained flare resolution by day three. Patients hospitalized to manage their acute CPP crystal arthritis seemed to have more difficult‐to‐treat episodes, whichever treatment was used. Although the COLCHICORT trial demonstrated the equivalence between colchicine and prednisone in achieving rapid pain relief by day two, 11 this post hoc analysis suggests that prednisone more often provides pain relief after the 2‐day treatment regimen, without further need for other anti‐inflammatory agents. This finding was suggested by the multiple logistic regression model, but the colchicine arm was not selected by the decision tree as being significant, imposing caution with this interpretation. Combined with the relatively better safety profile identified in the trial analyses first published, 11 these elements suggesting prednisone's superior efficacy to colchicine further supports its position as the first‐line option for the treatment of acute CPP crystal arthritis, which can be adjusted according to the predictive factors that we identified, particularly in case of type 2 diabetes, in which case colchicine seems a better option.
Decision trees are not commonly presented for decision‐making from trial results but provide intuitive guidance for personalized medicine according to patient profile clusters. 24 , 25 , 26 , 27 In our analyses, the CART analysis provided better results of sensitivity and specificity in crossvalidation than the logistic regression model did. The tree nodes determined which factors were decisive in each subpopulation obtained by the previous nodes to predict the treatment response. The multiple discretization is one of the main features of classification trees because the tree determines the most discriminating thresholds for quantitative variables in each node subpopulation. The decision tree clustering profiles of treatment further illustrated the importance of CRP levels, renal function, and initial reason for hospital admission to predict treatment response. The tree's performances had decent combined Se and Sp, with a one in four chance of misclassification, but will need to be tested in other populations to confirm that these results can be extrapolated.
Drug–drug interaction is a well‐known safety issue when considering the use of colchicine, mainly mediated by colchicine being metabolized both by cytochrome P450 3A4 (CYP 3A4) and P‐glycoprotein (P‐gp) located in the enterocytes and bile ducts. 28 The only firm available data so far on the drug–drug interactions of colchicine is with the strongest inhibitors of CYP 3A4 (macrolide antibiotics and ketoconazole) and of P‐gp (cyclosporine, tacrolimus, verapamil, and diltiazem), which either mildly or strongly influence colchicine exposure in a study of pharmacokinetics. 28 , 29 Clinical evidence suggests that statins impair the safety of colchicine, 30 as was again suggested by the first safety analyses of the COLCHICORT trial, showing that participants experiencing diarrhea were treated with statins twice as often as those who did not. 11 The explanation is expected to be that statins increase exposure to colchicine by its competitive metabolism, either through competition for CYP 3A4 (competitive metabolism clearance) or for P‐gp (decreased biliary and enterocytic clearance) according to the type of statin. 29 , 31 This post hoc analysis further suggests, despite a lack of statistical power, that statins are associated with colchicine's gastrointestinal AEs. Around 20% of colchicine is estimated to be cleared through the kidneys, but evidence in the literature is contradictory on whether chronic kidney disease is associated with a higher risk of AEs with colchicine. 32 , 33 Our results indicate that even a mild decrease in eGFR could be associated with an increased risk of gastrointestinal AEs.
There seems to be a dual effect of conditions and cotreatments increasing colchicine exposure, resulting in improved efficacy but also an increased risk of gastrointestinal side effects. Prior studies have addressed the issue of factors associated with safety alone, but it was not known how they could also influence treatment response. With limitations inherent to the statistical power of the trial, this post hoc analysis suggests that the above‐mentioned factors associated positively or negatively with colchicine safety are associated with its efficacy. For example, statins and older age to some extent seemed to be associated with a better treatment response but more gastrointestinal AEs, both potentially explained by increased exposure to colchicine. Diuretics, on the other hand, seemed to have a negative effect on colchicine's efficacy but were also associated with an increased risk of gastrointestinal AEs. These findings further question the ideal dose of colchicine according to age, renal function, and other current drugs at the time of the flare to provide the best possible balance between efficacy and safety. This concept is however in contradiction with the results of the AGREE trial performed in gout flares, which showed that high doses of colchicine (up to 4.8 mg/day) did not provide better rapid (24 hours) pain relief than lower doses (1.8 mg/day), despite an obvious higher exposure to colchicine. 19 Renal function and coprescriptions were however not reported in the AGREE trial, and participants were on average almost 40 years younger than those in the COLCHICORT trial, suggesting that these combined minor modulators of colchicine exposure may have a greater impact on older poly‐medicated people.
The study has inherent limitations. First, sample size impacted the statistical power of the study and required a selection of variables of interest and limited the analysis of the drugs’ individual efficacy, which was deliberately limited to an unadjusted univariable analysis often failing to reach statistical significance because of insufficient power, to avoid building invalid multivariable models. However, the robustness of the variables and outcomes prospectively collected in this first‐ever trial in acute CPPD disease provided a strong offset for these drawbacks, allowing clinically relevant hypotheses to be drawn from these observations, most of which were nonstatistically significant. Second, the prognostic value of the machine‐learning generated classification needs to be considered with caution given the substantial number of misclassifications in crossvalidation. These results need to be validated in other cohorts or trials to confirm that they can be extrapolated in all contexts of acute CPP crystal arthritis.
This post hoc analysis of the COLCHICORT trial identified factors associated with a sustained response to colchicine and prednisone treatment for acute CPP crystal arthritis after a 2‐day treatment regimen. Prednisone appeared to perform better in achieving pain relief in acute CPP crystal arthritis after two days of treatment, without requiring any further use of anti‐inflammatory agents, and should be the first‐line option for efficacy and safety reasons, except in case of type 2 diabetes. Renal function and comorbidities, inflammatory biomarkers, arthritis site, presence of underlying OA, and other current drugs are important determinants of treatment response and safety and should further guide the choice between them. Statins and age seem to have a dual effect on colchicine's efficacy and safety profiles, and further dose‐finding studies adjusting on these factors are required.
AUTHOR CONTRIBUTIONS
All authors contributed to at least one of the following manuscript preparation roles: conceptualization AND/OR methodology, software, investigation, formal analysis, data curation, visualization, and validation AND drafting or reviewing/editing the final draft. As corresponding author, Dr Pascart confirms that all authors have provided the final approval of the version to be published and takes responsibility for the affirmations regarding article submission (eg, not under consideration by another journal), the integrity of the data presented, and the statements regarding compliance with institutional review board/Declaration of Helsinki requirements.
Supporting information
Disclosure form.
Supplementary Table 1: Patients’ characteristics according to individual response to colchicine or prednisone (change in pain VAS from baseline of 50% or more and/or pain VAS <40/100 by day 3 (48 hours) and no further use of anti‐inflammatory agents). Univariable comparisons between responders and non‐responders were performed, with significant differences appearing in bold.
Supported by the French Inter‐regional Hospital Clinical Research Programme (2016).
Additional supplementary information cited in this article can be found online in the Supporting Information section (http://onlinelibrary.wiley.com/doi/10.1002/acr.25642).
Author disclosures are available at https://onlinelibrary.wiley.com/doi/10.1002/acr.25642.
References
- 1. Pascart T, Filippou G, Lioté F, et al. Calcium pyrophosphate deposition disease. Lancet Rheumatol 2024;6(11):e791–e804. [DOI] [PubMed] [Google Scholar]
- 2. Abhishek A, Neogi T, Choi H, et al. Review: unmet needs and the path forward in joint disease associated with calcium pyrophosphate crystal deposition. Arthritis Rheumatol 2018;70(8):1182–1191. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3. Rosenthal AK, Ryan LM. Calcium pyrophosphate deposition disease. N Engl J Med 2016;374(26):2575–2584. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4. Zhang W, Doherty M, Bardin T, et al. European League Against Rheumatism recommendations for calcium pyrophosphate deposition. Part I: terminology and diagnosis. Ann Rheum Dis 2011;70(4):563–570. [DOI] [PubMed] [Google Scholar]
- 5. Latourte A, Ea H‐K, Richette P. Calcium pyrophosphate and basic calcium phosphate crystal arthritis: 2023 in review. Gout Urate Cryst Depos Dis 2024;2(2):101–107. [Google Scholar]
- 6. Sirotti S, Scanu A, Pascart T, et al. Calcium pyrophosphate crystal formation and deposition: Where do we stand and what does the future hold? Curr Rheumatol Rep 2024;26(10):354–365. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7. Tedeschi SK. A New Era for Calcium Pyrophosphate Deposition Disease Research: The First‐Ever Calcium Pyrophosphate Deposition Disease Classification Criteria and Considerations for Measuring Outcomes in Calcium Pyrophosphate Deposition Disease. Gout Urate Cryst Depos Dis 2024;2(1):52–59. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8. McCarthy GM, Dunne A. Calcium crystal deposition diseases ‐ beyond gout. Nat Rev Rheumatol 2018;14(10):592–602. [DOI] [PubMed] [Google Scholar]
- 9. Pascart T, Latourte A, Tedeschi SK, et al. Features associated with different inflammatory phenotypes of calcium pyrophosphate deposition disease: A study using data from the International American College of Rheumatology/EULAR Calcium Pyrophosphate Deposition Classification Criteria Cohort. Arthritis Rheumatol 2024;76(12):1780–1788. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10. Abhishek A, Tedeschi SK, Pascart T, et al. The 2023 ACR/EULAR classification criteria for calcium pyrophosphate deposition disease. Arthritis Rheumatol 2023;75(10): 1703–1713. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11. Pascart T, Robinet P, Ottaviani S, et al. Evaluating the safety and short‐term equivalence of colchicine versus prednisone in older patients with acute calcium pyrophosphate crystal arthritis (COLCHICORT): an open‐label, multicentre, randomised trial. Lancet Rheumatol 2023;5(9):e523–e531. [DOI] [PubMed] [Google Scholar]
- 12. Maravic M, Ea HK. Hospital burden of gout, pseudogout and other crystal arthropathies in France. Joint Bone Spine 2015;82(5):326–329. [DOI] [PubMed] [Google Scholar]
- 13. Ramonda R, Oliviero F, Galozzi P, et al. Molecular mechanisms of pain in crystal‐induced arthritis. Best Pract Res Clin Rheumatol 2015;29(1):98–110. [DOI] [PubMed] [Google Scholar]
- 14. Zhang W, Doherty M, Pascual E, et al. EULAR recommendations for calcium pyrophosphate deposition. Part II: management. Ann Rheum Dis 2011;70(4):571–575. [DOI] [PubMed] [Google Scholar]
- 15. Mandl P, D'Agostino MA, Navarro‐Compán V, et al. 2023 EULAR recommendations on imaging in diagnosis and management of crystal‐induced arthropathies in clinical practice. Ann Rheum Dis 2024;83(6):752–759. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 16. Tedeschi SK, Becce F, Pascart T, et al. Imaging features of calcium pyrophosphate deposition disease: Consensus definitions from an international multidisciplinary working group. Arthritis Care Res (Hoboken) 2023;75(4):825–834. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17. Lydersen S. Statistical review: frequently given comments updated. Ann Rheum Dis 2025;84(5):660–663. [DOI] [PubMed] [Google Scholar]
- 18. Neufeld AC, Gao LL, Witten DM. Tree‐values: selective inference for regression trees. J Mach Learn Res 2022;23:23. [PMC free article] [PubMed] [Google Scholar]
- 19. Terkeltaub RA, Furst DE, Bennett K, et al. High versus low dosing of oral colchicine for early acute gout flare: twenty‐four‐hour outcome of the first multicenter, randomized, double‐blind, placebo‐controlled, parallel‐group, dose‐comparison colchicine study. Arthritis Rheum 2010;62(4):1060–1068. [DOI] [PubMed] [Google Scholar]
- 20. Leung YY, Haaland B, Huebner JL, et al. Colchicine lack of effectiveness in symptom and inflammation modification in knee osteoarthritis (COLKOA): a randomized controlled trial. Osteoarthritis Cartilage 2018;26(5):631–640. [DOI] [PubMed] [Google Scholar]
- 21. Døssing A, Henriksen M, Ellegaard K, et al. Colchicine twice a day for hand osteoarthritis (COLOR): a double‐blind, randomized, placebo‐controlled trial. Lancet Rheumatol 2023;5(5):e254–e262. [DOI] [PubMed] [Google Scholar]
- 22. Singh A, Molina‐Garcia P, Hussain S, et al. Efficacy and safety of colchicine for the treatment of osteoarthritis: a systematic review and meta‐analysis of intervention trials. Clin Rheumatol 2023;42(3):889–902. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23. Kroon FPB, Kortekaas MC, Boonen A, et al. Results of a 6‐week treatment with 10 mg prednisolone in patients with hand osteoarthritis (HOPE): a double‐blind, randomised, placebo‐controlled trial. Lancet 2019;394(10213):1993–2001. [DOI] [PubMed] [Google Scholar]
- 24. Sharma L, Kwoh K, Lee JJ, et al. Development and validation of risk stratification trees for incident slow gait speed in persons at high risk for knee osteoarthritis. Ann Rheum Dis 2019;78(10):1412–1419. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25. Jatuworapruk K, Grainger R, Dalbeth N, et al. Development of a prediction model for inpatient gout flares in people with comorbid gout. Ann Rheum Dis 2020;79(3):418–423. [DOI] [PubMed] [Google Scholar]
- 26. Rehberg M, Giegerich C, Praestgaard A, et al; MOBILITY, MONARCH, TARGET, and ASCERTAIN investigators . Identification of a rule to predict response to sarilumab in patients with rheumatoid arthritis using machine learning and clinical trial data. Rheumatol Ther 2021;8(4):1661–1675. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27. Bolognesi MM, Capitoli G, Galimberti S, et al. Dissecting the histological features of lupus nephritis highlights new common patterns of injury in class III/IV. Ann Rheum Dis 2022;81(12):1704–1711. [DOI] [PubMed] [Google Scholar]
- 28. Pascart T, Richette P. Colchicine in gout: an update. Curr Pharm Des 2018;24(6):684–689. [DOI] [PubMed] [Google Scholar]
- 29. Terkeltaub RA, Furst DE, Digiacinto JL, et al. Novel evidence‐based colchicine dose‐reduction algorithm to predict and prevent colchicine toxicity in the presence of cytochrome P450 3A4/P‐glycoprotein inhibitors. Arthritis Rheum 2011;63(8):2226–2237. [DOI] [PubMed] [Google Scholar]
- 30. Schwier NC, Cornelio CK, Boylan PM. A systematic review of the drug‐drug interaction between statins and colchicine: patient characteristics, etiologies, and clinical management strategies. Pharmacotherapy 2022;42(4):320–333. [DOI] [PubMed] [Google Scholar]
- 31. Kwon OC, Hong S, Ghang B, et al. Risk of colchicine‐associated myopathy in gout: influence of concomitant use of statin. Am J Med 2017;130(5):583–587. [DOI] [PubMed] [Google Scholar]
- 32. Bausson J, Keller N, Von Hunolstein JJ, et al. Safety and efficacy of colchicine in crystal‐induced arthritis flare in 54 patients with severe chronic kidney disease. RMD Open 2024;10(1):e003872. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33. Bajpai R, Partington R, Muller S, et al. Prognostic factors for colchicine prophylaxis‐related adverse events when initiating allopurinol for gout: retrospective cohort study. Rheumatology (Oxford). 2025;64(3):1147–1154. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Disclosure form.
Supplementary Table 1: Patients’ characteristics according to individual response to colchicine or prednisone (change in pain VAS from baseline of 50% or more and/or pain VAS <40/100 by day 3 (48 hours) and no further use of anti‐inflammatory agents). Univariable comparisons between responders and non‐responders were performed, with significant differences appearing in bold.