We performed an exploratory post hoc analysis of data from a phase II randomised, double-blind, placebo-controlled trial of siltuximab in patients with iMCD (9). In the phase II study, CRP levels were measured in patients receiving siltuximab (11 mg/kg every 3 weeks; n=53) or placebo (n=26) weekly for the first 3 weeks, then once every 3 weeks for a total period of 48 weeks (9). Patients received best supportive care, which included management of effusions, use of antipyretic, antipruritic, antihistamine and pain drugs, management of infections, transfusions, and standard management of infusion-related reactions as specified in institutional guidelines. Use of erythropoietin-stimulating agents, anti-tumour treatments, biological treatments, or an increase from baseline or a new course of corticosteroids were not allowed (9).

We classified patients as having low or high CRP levels at baseline using a threshold of 40 mg/l (corresponding to approximately 40 pg/ml of IL-6; selected based on our clinical experience and the mean [standard deviation] baseline CRP level in the siltuximab arm of 43.18 [53.63] mg/l). To evaluate CRP inhibition, one-way repeated measures of analysis of variance followed by Tukey’s post hoc correction were performed using data from patients with iMCD treated with siltuximab (high and low baseline CRP groups) for CRP measured on days 1, 8 and 15 (all CRP measurements taken in the first treatment cycle of the study). A p-value of ≤ 0.05 was considered significant.

In contrast to iMCD, where siltuximab is the only drug tested in a randomised trial, there are many reports in the literature evaluating therapies in patients with COVID-19. In June 2022, we searched ClinicalTrials.gov and the published literature (PubMed) to identify studies evaluating the efficacy, or assessing the potential role, of anti-IL-6 therapy in patients with severe COVID-19. We found 53 studies with tocilizumab, two with siltuximab and nine with sarilumab, plus a prospective meta-analysis of 27 randomised trials that included 10,930 patients hospitalised for COVID-19. The meta-analysis concluded that administration of IL-6 antagonists, compared with usual care or placebo, was associated with significantly lower 28-day all-cause mortality. Surprisingly, only one study, by Luo et al. (2020), reported the dynamics of CRP reduction for individual patients throughout therapy (20).

We analysed the previously published data of Luo et al. (20) showing CRP inhibition in patients with COVID-19 treated with tocilizumab. We plotted data for patients with multiple CRP data points against a hypothetical curve showing optimal CRP inhibition. This curve was based on the half-life of CRP and showed CRP levels that would reflect a complete block of IL-6.

We developed a model by expanding and modifying the basic structure of a previously developed model (10, 15) to accommodate observations from the experimental literature and our own experimental data (Supplementary Table 1) (10, 11, 15, 20). The model was built using the SimBiology package in MATLAB 2019b. It comprised 17 biological elements (e.g. IL-6, IL-6R, gp130) with 20 biological reactions to represent their time-dependent concentrations (Supplementary Table 1).

IL-6 binds to its membrane-bound receptor (IL-6R) or plasma-soluble receptor (sIL-6R). Both IL-6–IL-6R and IL-6–sIL-6R complexes bind to a membrane glycoprotein transducer (gp130), initiating signal transduction (Figure 2) (1–4). Our algorithm was developed to take into account these components of IL-6 signalling to model the formation of IL-6–(IL-6R/sIL-6R)–gp130 complexes and then of CRP production, which serves as a proxy for IL-6 bioactivity. The algorithm included reactions in a general plasma compartment and a local immobile compartment (lymph nodes for iMCD and bronchoalveolar fluid [BAF] for COVID-19).

The model was fitted to experimental data for IL-6, glycoprotein (gp)130 and CRP simultaneously with a proportional error model. This was chosen based on the large order of magnitude variation in the data across the experiments.

where y_exp is biological species, y is experimental response, y_sim is biological species y simulation value and b is the proportional error parameter. IL-6R and gp130 in the solution form were tagged ‘s’; IL-6 and mAb could only be in solution form.

The reaction system was divided into a general plasma compartment and an immobile compartment present at the cellular membrane. One of the simplest approaches to model these biological reactions is to use a ‘curve fitting’ to construct a curve (mathematical model) that has the best fit to series of known data. The algorithm of stiff ode15s solver was used with settings as shown in Supplementary Table 1. A tolerance of 1e–5 was chosen for termination of the data-fitting procedure for the step change in estimated parameters and the function value, and 1e–6 for the first-order optimality. This fitting was performed in an iterative fashion as the model contained numerous parameters to allow efficient simultaneous estimation.

A molar equivalent-order kinetic equation was used for the binding of receptor and mAb in the solution. The kinetic equation of surface-immobilised ligand and protein was used for interactions on the membrane. Catabolism and appeared clearance were also described by mass/molar action. The degradation rate or clearance rate for non-structural proteins was estimated using the half-life (t_1/2). For example, if the value for the plasma half-life cycle of sIL-6R was 2 hours, κ would be 0.277. This estimation of degradation rates was found to be acceptable for our model after fitting, and aided in reducing the degrees of freedom in the fitting process. In addition, degradation rates for all biological species on the membrane were estimated to be zero based on previous studies (10, 15), showing complete recovery in the hepatocyte compartment (e.g. gp130). In plasma, different cases of steady-state IL-6 and sIL-6R levels were simulated with a constant generation rate. The steady-state level of sIL-6R is 75 ng/ml and IL-6 levels are in the range of 1–10,000 pg/ml. The program performed the simulation for 15 days (360 hours) and the observed biological species had to reach a stable level before applying treatment. A simulation can be reproduced using this software and the above settings.

The ‘Sensitivity Analysis’ task of the SimBiology package was used to estimate parameter sensitivities for IL-6, IL-6R and CRP production, as shown in Supplementary Figure 5. This task uses the complex-step derivative approximation, which is often used for metabolic systems. Full dedimensionalisation of the analysis was used to permit comparisons of parameters with widely different orders of magnitude. The calculation for complex-step derivative sensitivities can be represented by the following equations for each parameter:

where f = is the first derivative of the simulation function with respect to a given reaction rate k _reaction and t is time.

The iMCD scenario was modelled with persistent high IL-6 secretion, at 1 ng/ml – the median concentration reported for patients with iMCD (9). The severe COVID-19 scenario (with a cytokine storm and massive IL-6 production) used 3 ng/ml IL-6 in BAF – the median concentration reported for patients with severe COVID-19 (13).

The diffusion of mAbs into bronchoalveolar lavage fluid is unknown but is expected to be low (~0.1–0.3%) according to the diffusion data for an anti-respiratory syncytial virus mAb into bronchoalveolar lavage fluid in animal models (21). In contrast, the diffusion of mAbs into lymph nodes occurs more freely through the sinusoidal clefts, which allows mAb movement and biodistribution, based on rituximab-mediated B-cell depletion in lymph nodes, and was estimated to be 8.46% of the plasma concentration (22). For this reason, we modelled primarily with local mAb concentrations at 10% of those in plasma. To account for the uncertainty (particularly in the scenario of COVID-19), we also modelled various situations using concentrations from 1% to 100% of those in plasma, and present these results as Supplementary Data.

Using the mathematical model described above, we modelled the ability of tocilizumab and siltuximab, or a combination of both, to inhibit IL-6 bioactivity (i.e. formation of IL-6/IL-6R–gp130 and IL-6/sIL-6R–gp130 complexes). The inhibition rate (%) versus pre-treatment levels was defined as the ratio of gp130 complexes – IL-6/IL-6R–gp130 and IL-6/sIL-6R–gp130 – after applying mAb treatments to reach their stable level. We also computed CRP levels; we considered a 1-day delay for induction of CRP production by IL-6. Results show the concentration of CRP in the circulating compartment. Daily IL-6 production was assumed to be constant throughout treatments.

Doses of siltuximab and tocilizumab used in the model were selected based on clinical experience. Siltuximab 700 mg was used, as this represents the median dose of siltuximab for a person weighing 75–80 kg. A higher dose of tocilizumab (800 mg) was used because some patients with COVID-19 have been administered two injections of 400 mg each (Supplementary Table 2). In the modelled scenarios, siltuximab (700 mg) or tocilizumab (800 mg) was either injected once at day 0 (D0) or given as repeated daily injections (D0, D1, D2, D3, —, Dx). A further scenario modelled alternating daily injections of siltuximab and tocilizumab (siltuximab: D0, D2, —, Dx; tocilizumab: D1, D3, —, Dx+1).

Given previous estimates indicating weak diffusion of plasma mAb into lymph nodes and interstitial lung tissue (8.46% and 14.9% of that in plasma, respectively) (21, 22) and the established pharmacokinetics/pharmacodynamics of siltuximab (23) and tocilizumab (24), we assumed the mAb concentration in lymph nodes (iMCD scenario) or BAF (COVID-19 scenario) to be 10% of that in plasma. Because of uncertainty around mAb biodistribution, particularly into BAF (unknown but expected to be low: ~0.1–0.3% of that in plasma; see Supplementary Methods), we also include results with mAb penetration into lymph nodes and BAF of 100% and 1% of the concentration in plasma as Supplementary Data.

Substantial variability was observed in the baseline CRP levels of patients with iMCD who participated in the phase II trial of siltuximab (9). Across the 79 patients with iMCD (siltuximab, n=53; placebo, n=26), baseline CRP levels were 0.39–181 mg/l (Figure 3A). To assess the impact of baseline CRP levels on response to siltuximab, patients were divided into low and high CRP groups based on a threshold CRP value of 40 mg/l (corresponding to approximately 40 pg/ml of IL-6 (25)).

In patients with low baseline CRP levels (≤40 mg/l) who were treated with siltuximab (n=35), CRP levels were significantly reduced versus baseline on day 8 (n=26; p=0.0003) and day 15 (n=22; p=0.0043) post dosing (Figure 3B). In patients with high baseline CRP levels (>40 mg/l) who were treated with siltuximab (n=17), CRP levels were significantly reduced versus baseline on day 8 (n=17; p < 0.0001) and day 15 (n=15; p < 0.0001) post dosing (Figure 3C). However, a significant increase in CRP levels from day 8 to day 15 was observed in this group (p=0.0120); this increase was not observed in patients with low baseline CRP levels (p=0.243). Moreover, for some patients with high baseline CRP, CRP levels remained above 40 mg/l after siltuximab treatment. For these patients, a single dose of siltuximab did not efficiently inhibit IL-6 activity.

To our knowledge, only one study (20) has reported the dynamics of CRP reduction for individual patients with severe COVID-19 treated with anti-IL-6 therapy. To evaluate the completeness of IL-6 activity inhibition in patients with severe COVID-19 treated with the anti-IL-6R therapy tocilizumab in the study by Luo et al. (20), we examined CRP levels in patients who died or had disease aggravation (n=5) and those with clinical stabilisation or improvement (n=10) following tocilizumab administration. In general, incomplete CRP reduction was observed in patients who died or experienced disease aggravation following treatment with tocilizumab (Figure 4A). The three patients who died on, or before, day 7 had CRP levels of 52, 18 and 13 mg/l at the last reported measurement. One patient who survived beyond day 7 but experienced disease aggravation had a CRP level of 94 mg/l on day 7. Another patient whose CRP levels decreased to 6 mg/l at day 7 also experienced disease aggravation despite the reduction in CRP. In patients with clinical stabilisation or improvement following treatment with tocilizumab, CRP levels were more effectively suppressed (Figure 4B), with all CRP levels ≤5 mg/l at last reported measurement (median 2.3, range 0.5–5).

To further evaluate this trend, we selected patients with several CRP measurements to assess the dynamics of the relative reduction (%) in CRP serum values. We compared the reduction in CRP to a hypothetical curve of CRP reduction required for complete block of IL-6 (taking into account the 18-hour half-life of CRP). The two patients who died had early CRP reduction rates (at 24 hours post tocilizumab treatment) above the theoretical curve representing a complete block of IL-6; conversely, at 24 hours post tocilizumab treatment, the three patients who stabilised had CRP reduction rates below the theoretical curve (Figure 4C).

We modelled a situation representative of iMCD, with persistent high IL-6 secretion, at 1 ng/ml IL-6, the median concentration reported for patients with iMCD (9). We also modelled a situation representative of severe COVID-19, with a cytokine storm and massive IL-6 production (3 ng/ml IL-6 in BAF – the median concentration reported for patients with severe COVID-19 (13)). Assuming a mAb concentration in lymph nodes or BAF of 10% that in plasma (see Materials and Methods), we generated theoretical curves showing the efficacy of siltuximab, tocilizumab or a combination of the two mAbs in reducing the formation of gp130 complexes (a surrogate measure for IL-6 bioactivity) and plasma CRP levels produced in the liver.

Applying the model to the scenario representative of iMCD (with persistent high IL-6 secretion at 1 ng/ml and mAb penetration into lymph nodes at 10% of the plasma concentration), a single administration of siltuximab suppressed IL-6 bioactivity by 80%. Similar to the clinical data in patients with iMCD with high baseline CRP levels (Figure 3C), IL-6 bioactivity increased over the next 14 days to 90% of pre-treatment levels (Figure 5A). Repeated injections of siltuximab every 24 hours produced more sustained inhibition than a single injection, with IL-6 bioactivity plateauing at 82% inhibition (Figure 5B). Tocilizumab inhibited IL-6 bioactivity to a lesser extent than siltuximab; repeated daily injections of tocilizumab caused IL-6 bioactivity to plateau at 60% inhibition (Figure 5C). Theoretical administration of alternating daily injections of siltuximab and tocilizumab had a marginally greater effect in suppressing IL-6 bioactivity than siltuximab monotherapy (Figure 5D).

Because CRP is measured in clinical practice and used as a surrogate measure for IL-6 bioactivity when administering anti-IL-6 therapies, we repeated the simulations described above to determine the inhibition of CRP in response to siltuximab, tocilizumab or both (Figure 6). CRP inhibition reflected the results observed for inhibition of gp130 complex formation.

In a situation representative of severe COVID-19 (3 ng/ml IL-6 in BAF; mAb penetration into BAF 10% of that in plasma), a single injection of siltuximab inhibited IL-6 bioactivity by 90%, with an immediate resumption of IL-6 bioactivity within 24 hours (Supplementary Figure 2A). Due to the continuous daily production of IL-6 (which was factored into the model), repeated daily dosing was required to maintain IL-6 bioactivity inhibition. Repeated injections of siltuximab every 24 hours caused IL-6 bioactivity to plateau with 65% inhibition (Supplementary Figure 2A and Supplementary Table 3). A single tocilizumab injection inhibited IL-6 bioactivity by 52%, with immediate restoration of IL-6 bioactivity within 24 hours. Repeated daily injections of tocilizumab caused IL-6 bioactivity to plateau with 38% inhibition (Supplementary Figure 2A and Supplementary Table 3). Alternating daily injections of siltuximab and tocilizumab produced maximal inhibition of IL-6 bioactivity, with a brief inhibition of 94% followed by a sustained plateau of inhibition at 85% (Supplementary Figure 2A and Supplementary Table 3). Similar results were obtained when CRP was used as a biomarker for IL-6 bioactivity (Supplementary Figure 2B and Supplementary Table 3).

To account for uncertainty around the diffusion of mAbs into BAF and lymph nodes, particularly when inflammation is present, we replicated the simulations described above with various diffusion abilities of the mAbs. In both iMCD and COVID-19, increasing the local mAb concentration from 10% to 100% increased inhibition of IL-6 bioactivity, with a sustained full blockade of IL-6 bioactivity with combination treatment, even in the scenario of severe COVID-19 with a cytokine storm (Supplementary Figure 3). In the iMCD scenario with mAb concentrations in lymph nodes at 1% of those in plasma, repeated daily administration of siltuximab caused IL-6 bioactivity to plateau at 35% inhibition, with a similar plateau observed for alternating daily injections of siltuximab and tocilizumab (Supplementary Figure 4). Repeated daily administration of tocilizumab caused IL-6 bioactivity to plateau at 20% inhibition. In the COVID-19 scenario with mAb concentrations at 1% of those in plasma, IL-6 bioactivity was minimally inhibited even with repeated daily injections of anti-IL-6 or anti-IL-6R mAb, either alone or in combination (Supplementary Figure 4). The plateaus of gp130 inhibition with repeated administration of siltuximab, tocilizumab or both mAbs at various diffusion abilities of the mAbs (1%, 10%, 20%, 40% and 100% of that in plasma) are shown in Supplementary Table 3.