The Influence of Patient Factors on the Population Pharmacokinetics of Colchicine: Implications for Safe and Effective Dosing

Colchicine is an alkaloid found in the plants Colchicum autumnale (autumn crocus) and Gloriosa superba (glory lily). It has been used for several centuries as an anti-inflammatory, particularly in the acute treatment of gout flares. Colchicine is also recommended for the prevention of gout following the initiation of urate-lowering therapy [1]. The ability of colchicine to dampen the innate immune response has been exploited for the management of several other chronic conditions, including familial Mediterranean fever, Behcet’s disease and pericarditis. More recently, colchicine has shown promise in the secondary prevention of cardiovascular disease [2].

Colchicine has a narrow margin of safety. Toxicities, including treatment-limiting gastrointestinal effects, were common under traditional dosing strategies for the control of inflammation during gout flares. Patients were advised to take a loading dose of 1.2 mg followed by repeated doses of 0.6 mg every 1–2 h until either the pain subsided, gastrointestinal symptoms appeared or a maximum dose of 6 mg was achieved [3]. Since gastrointestinal symptoms may precede other serious adverse effects like myopathies, nerve damage and bone marrow suppression [4], the recommended dose has been reduced to 1.2 mg followed by 0.6 mg 1 h later [3]. Fatal overdoses have been rarely reported after single ingestions as low as 7 mg [4].

The pharmacokinetics of colchicine have well been characterised in healthy volunteers [5‐10] but have not been extensively studied in target patient populations (although see [11, 12] for studies in those with kidney impairment and [13, 14] for pharmacokinetic analyses in those with familial Mediterranean fever). Available data suggest that orally administered colchicine is subject to a zero-order absorption and variable bioavailability in the order of 24–88% [6, 7, 9]. Colchicine has been found to be a substrate for the ABCB1 efflux transport (P-glycoprotein) and CYP3A4 [15, 16] expressed in the gut wall and liver. Peak plasma concentrations (steady-state maximum concentration [C_max,ss]) can occur between 30 min and 3 h after administration of single doses [17], while secondary peaks at about 6 h in some individuals suggest evidence of entero-hepatic recycling [7]. Elimination is mainly through the bile, while an average of 10–25% has been reported to be eliminated by the kidneys [5, 12]. Systemic clearance (apparent clearance [CL/F]) values in the order of 10–40 L/h have been reported. Volume of distribution estimates of 6–15 L/kg support the accumulation of colchicine in peripheral compartments, particularly in leukocytes.

Despite a long history of clinical use, and an expanding array of therapeutic indications, the rational dosing of colchicine remains under-researched. Current dose recommendations appear to have been established largely by consensus and trial and error. Low dosages of 0.5–0.6 mg once or twice daily are commonly used to prevent gout flares when starting urate-lowering therapy and for the management of pericarditis. Dosages of 1–1.5 mg once or twice daily are recommended for familial Mediterranean fever. While ≤ 0.6 mg daily is purported to be safe in most patients [18], little is known about whether this is also the case in special populations such as those with impaired kidney or liver function or those taking interacting drugs. Dosing in obese people, where the efficacy of low-dose regimens may be compromised, has not been extensively explored. The impact of these patient factors on steady-state colchicine plasma concentrations and the implications for efficacy and safety are not clear. A therapeutic range for steady-state colchicine plasma concentrations of 0.5–3 mcg/L has been proposed [19] but this is not supported by a clear evidence base for efficacy and safety. The aims of this study were (1) to develop a population pharmacokinetic model for colchicine, (2) to determine the influence of patient factors on colchicine pharmacokinetics and (3) to predict the probability of achieving steady-state plasma concentrations within the proposed therapeutic range of 0.5–3 ng/mL for colchicine under different doses and across a range of patient characteristics.

Data from 80 people with gout from the NZ Gout Study, including 160 colchicine plasma concentrations, were available for analysis. Three samples below the LLOQ were excluded as well as four where the sample time and dose could not be reconciled. There were 78 individuals and 153 colchicine plasma concentrations from the NZ Gout Study available for analysis. An additional 211 samples from 13 people were extracted from published studies. In total, the final dataset, therefore, included 91 individuals and 364 colchicine plasma concentrations. Demographic and clinical details from the NZ Gout study are summarised in Table 1.

A two-compartment pharmacokinetic model with zero-order absorption and first-order elimination provided the best fit to the colchicine data. Plots showing the apparent zero-order absorption in the first 4 h after the dose using data extracted from two studies [5, 7] are presented in Figure S1 (see the electronic supplementary material). A first-order absorption model was tested but did provide a better model fit (i.e. was not inferior); therefore, the zero-order absorption model was retained to align with previous findings from pharmacokinetic studies in the literature [5, 7, 13, 17]. Bioavailability was fixed to 0.47 in the final model. The data did not support the estimation of between-subject variance for the volume parameters, and these were fixed to values used by Karatze et al. [22]. An absorption time lag for colchicine has been reported [17] but was found to create model instability during the model building. Previous work suggests that the absorption time lag may follow a bimodal distribution – a maximum concentration of about 30 min post-dose in some individuals and a maximum concentration of about  2 h in others [17]. In the NZ Gout Study, 17 individuals were noted to have C_max,ss values less than or equal to the C_min,ss, which was assumed to represent a delay in absorption. Therefore, a parameter to capture the delay (t_lag) was fixed at 1.3 h in these individuals (the longest time post-dose recorded in the NZ Gout Study) and to zero otherwise. This resulted in a stable model. Note that a ‘mixture model’ ($MIX in NONMEM) to account for the bimodal t_lag was not supported. Finally, a formulation parameter was estimated on bioavailability (F1) to account for the differences in tablet formulation between the NZ Gout Study and the data extracted from the literature.

Total body weight (TBW) and statin use were found to be significant covariates on colchicine clearance parameters, while TBW and sex were found to predict volume parameters. The final covariate models for colchicine clearance and volume were described by:where \(\text{CL}/F\) is apparent oral clearance, \(V1/F\) is the apparent volume for the central compartment, \(V2/F\) is the apparent volume for the peripheral compartment, \(Q/F\) is the apparent intercompartmental clearance, \(WT\) is total body weight, \({\theta }_{statin}\) is the fractional effect of statin use and \({\theta }_{\text{SEX}}\) is the fractional effect of female sex. Parameter estimates for the final colchicine pharmacokinetic model are provided in Table 2. The sampling-importance-resampling runs produced parameter estimates close to the final model, suggesting a stable model. Statin use decreased colchicine clearance by ~ 30%, while female sex reduced colchicine volume by about 50%.

No major model misspecifications are obvious in the diagnostic plots presented in Fig. 2. pcVPCs for the oral data and for the NZ Gout Study data are provided in Fig. 3. The 5th, 50th and 95th percentiles of the model predictions are a good reflection of the observed data, suggesting an acceptable model fit. pcVPCs stratified by body weight, statin use and sex are presented in Figure S2 (see the electronic supplementary material).

Bild vergrößern

Bild vergrößern

The NONMEM code for the final colchicine pharmacokinetic model is provided in the Electronic Supplementary Information.

The probability of achieving steady-state colchicine plasma concentrations within the nominal therapeutic range of 0.5–3 ng/mL as well as the median and 5th and 95th percentiles of the predicted concentrations for dosages of 0.5–3 mg daily is presented in Table 3. The same metrics under different weight quartiles, statin use and for males and females are presented in Fig. 4(a–f) and Tables S3–8 (see the electronic supplementary material). For a typical gout cohort, the model predicts that about 90% of the C_min,ss values will fall between 0.16 ng/mL and 6.11 ng/mL, between 0.77 and 9.96 ng/mL for C_max,ss, and between 0.31 and 6.67 ng/mL for C_av,ss (Table 3). C_av,ss values for colchicine dosages ≤ 1.5 mg a day were predicted to fall below the upper limit of the therapeutic range > 80% of the time, while 0.5 mg daily produced C_av,ss values below 0.5 ng/mL in those with a body weight of > 80 kg (Table S3). Colchicine dosages of 1 mg twice daily were predicted to produce C_av,ss values > 3 ng/mL in about 30–50% of patients with a body weight of < 97 kg and in those taking statins (Table S7). Dosages of 1.5 mg twice daily were expected to exceed 3 ng/mL in > 70% of patients with a body weight of < 97 kg and in about 80% of those taking statins (Table S8).

Bild vergrößern

The MATLAB code used for the colchicine plasma concentration simulations is provided in the Electronic Supplementary Information.

In this study, we found that total body weight, statin use and sex significantly influenced the pharmacokinetics of colchicine. Colchicine CL/F was reduced by about 30% in those taking statins, and female sex was associated with a 50% reduction in colchicine apparent volume. Simulations from the model predict that colchicine dosages of ≤ 1.5 mg daily are expected to have a reasonable probability of safety in most patients, although 0.5 mg daily carries a risk of reduced efficacy in those with a body weight of > 80 kg. This assumes that the lower limit of the therapeutic range is reasonable, which is yet to be established. At dosages of > 1.5 mg daily, the model predicts that there is an increased risk of exceeding the upper limit of the therapeutic range in patients with low body weight (< 80 kg) who are taking statins. Reassuringly, long-term clinical trials in patients taking colchicine 0.5 mg daily have not reported evidence of increased risk of myotoxicity in those taking concomitant statins [30].

There are few previously published colchicine pharmacokinetic models and none to our knowledge using an industry-standard, non-linear mixed effects methodology. A study by Berkin et al. [13] summarised a colchicine population analysis in paediatric and adult patients with familial Mediterranean fever but reported only the descriptive non-compartmental outputs. Other empirical models by Thomas et al. [5], Ferron et al. [7] and Rochdi et al. [9] were developed by individually fitting intensively sampled plasma concentrations from healthy volunteers and finding the mean parameter values using a two-stage approach. No formal model development was conducted in the model reported by Karatze et al. [22]. The authors used a previously published two-stage model [5] combined with colchicine clearance estimates from a study in volunteers with different levels from kidney function [12] to explore possible colchicine doses for the treatment of coronavirus disease 2019 (COVID-19). The same model has since been used to simulate colchicine plasma concentrations in support of a consensus statement about low-dose colchicine safety and efficacy in the management of gout and cardiovascular disease [18]. Our work adds to this literature by presenting a model developed using data from a target population for colchicine use (gout patients). The plasma concentrations simulated here used a robust Monte-Carlo methodology, sampling covariates from a simulated clinical trial database of gout patients, and provide evidence of the impact that patient factors will have on plasma colchicine exposure.

A fractional reduction in colchicine clearance of about 25% was observed in individuals taking concomitant ACEIs, but this was not included in the final model. ACEIs are not reported to be P-glycoprotein or CYP3A4 substrates, inhibitors or inducers. The ACEI cohort were older than those not taking ACEIs (73 vs 55 years, respectively, p = 0.0006) and had lower kidney function (CLcr 53 vs 86 mL/min, respectively, p < 0.0001). We explored the possibility that the observed ACEI effect was driven by a correlation with statin use, but no difference in the frequency of concomitant statin use between those taking and not taking ACEIs (35% vs 20%, p = 0.2014, Fishers Exact test) was found. It was noted that only six of the 16 participants taking an ACEI were also taking a statin, which does not support correlation. In addition, we found no evidence to support the idea that the ACEI effect in the model might be capturing the impact of reduced kidney function. The range of renal function values in the data were limited and a correlation between ACEI and kidney function as a cause of the ACEI effect on colchicine clearance could not be ruled out. On the grounds that the ACEI effect has no mechanistic basis and might be misleading, we deleted this covariate from the final model until further data are available to substantiate this finding.

The authors acknowledge that the widely reported therapeutic range for colchicine (0.5–3 ng/mL) is somewhat speculative and indeed of unknown origin. For the purposes of our research, we have made the assumption that 0.5–3 ng/mL is the acceptable range for predicting colchicine safety and efficacy. The first reference in the literature appears to be a review paper by Molad [19] in 2002. An early study by Katz et al. [31] looked at colchicine plasma concentrations and response in patients with familial Mediterranean fever. The authors reported that responders had mean colchicine steady-state plasma concentrations of 2.5 ng/mL while non-responders had a mean of 0.9 ng/mL. Most patients were taking 1 mg daily doses. The timing of the plasma concentrations relative to the dose were not standardised and ranged from 2 h post-dose to 24 h, with most samples taken between 2 and 16 h post-dose. Importantly, the authors did not propose the quoted therapeutic range of 0.5–3 ng/mL. It remains unclear whether this reference range refers to a C_min,ss, C_max,ss or C_av,ss, so we have reported all three metrics here.

Our work should be viewed in light of some additional limitations. The pharmacometric analyses conducted were not adequately powered to detect the impact of kidney function on colchicine pharmacokinetics, if one exists. While only 10–25% of colchicine is reported to be excreted in the urine [5.12], colchicine clearance is reported to be impacted by kidney function in several sources [12]. The NZ Gout Study excluded participants with an estimated glomerular filtration rate (eGFR) of < 30 mL/min, and as a result, there were only nine individuals with a CLcr of < 50 mL/min in the modelling dataset and only four with a CLcr of < 40 mL/min. Finally, our simulations sampled patient factors from a virtual population of people with gout. While we might expect this population to share much in common with cardiovascular patients and other chronic disease groups, it is not clear whether the virtual gout patients represent other uses of colchicine, such as patients with familial Mediterranean fever.

In this pharmacometric analysis, we found that body weight, statin use and sex were significant predictors of colchicine pharmacokinetics. Together, these factors reduced the unexplained variability in colchicine clearance by almost 40%. The model predicts that low-dose regimes of ≤ 1.5 mg daily will produce colchicine plasma concentrations below the proposed upper threshold for safety (3 ng/mL) in most patients, providing reassurance to prescribers about safety concerns. The commonly used dosage of 0.5 mg daily produced plasma concentrations below the lower limit of the therapeutic range in those > 80 kg, possibly signalling a risk of reduced efficacy. Reassuringly, the positive outcome evidence at this dosage in pericarditis, gout and cardiovascular disease suggests that the lower boundary of the therapeutic range may be different from that suggested in the literature. Higher colchicine dosages of > 1.5 mg daily may exceed the proposed upper limit of safety in many individuals, particularly those with low body weight who are taking statins. This research represents an important step towards a better understanding of colchicine exposure response and provides the basis for further work on the safe and effective dosing of colchicine in clinical practice.