We identified six short-term randomised trials (12–52 week duration) evaluating 4 mg/kg or 8 mg/kg tocilizumab in 2891 patients with rheumatoid arthritis (weighted mean age 52·3 years; 19% male; appendix p 28
C-reactive protein was the most widely reported inflammation marker and its weighted mean concentration at baseline was 28·2 (SD 1·9) mg/L. Tocilizumab treatment (4, 8, and 16 mg/kg in randomised or observational studies every 4 weeks) was associated with a dose-dependent reduction in C-reactive protein concentration (appendix p 29
). The 8 kg/mg dose reduced fibrinogen, increased interleukin 6 and soluble IL6R, and increased LDL and HDL cholesterol ().
Summary effects of tocilizumab (8 mg/kg) and the IL6R rs7529229 variant on inflammatory, lipid, hepatic, and haematological biomarkers
40 studies contributed genotype and phenotype data for the de-novo genetic analysis of IL6R
SNPs from a total of 133 449 individuals with mean age at recruitment of 59 (range 26–75) years, of whom 49% were women. Additional characteristics of study participants are described in appendix pp 11–16, 30–33
. 12 of the 42 SNPs in the region of the IL6R
locus on the HumanCVD BeadChip met chip-wide significance (p<1×10−5
) for their association with circulating interleukin 6 in the Whitehall II study (appendix pp 20–21
). We selected three SNPs (rs7529229, rs4845371, rs12740969) based on MAF greater than 0·3, β coefficient greater than 0·9 log interleukin-6 concentration per allele, previously reported associations, and low-redundancy LD structure (appendix p 27
). The rs7529229 variant was in strong LD (r2
=0·92 in the Whitehall II study) with a non-synonymous variant (rs8192284, also annotated as rs2228145, which did not meet our initial selection criteria) previously reported to be associated with increased proteolytic cleavage of the soluble IL6R from its membrane-bound form30,31
(see appendix p 26
for mechanistic details) and became our lead SNP for the analysis. Where rs7529229 was not genotyped, a proxy SNP was used (defined on the basis of r2
≥0·90 with rs7529229 in individuals of European ancestry; appendix pp 18–19
). Information about rs4845371 and rs12740969 is reported in subsidiary analyses. 40 studies (133 449 participants) provided data for rs7529229, 18 studies (52 475 participants) for rs4845371, and 19 studies (59 126 participants) for rs12740969. All studies met the prespecified quality control threshold criteria for call rate, HWE, and MAF (appendix pp 18–19
rs7529229 SNP displayed additive associations with circulating concentrations of interleukin 6, C-reactive protein, and fibrinogen (, ). Circulating interleukin-6 concentration increased with each additional copy of the minor allele at rs7529229 (relative increase in geometric mean log interleukin-6 concentration per allele 9·45%, 95% CI 8·34–10·57; p=8·41×10−68
), whereas the concentrations of C-reactive protein and fibrinogen decreased per minor allele (relative decrease in geometric mean log C-reactive protein 8·35%, 95% CI 7·31–9·38, and fibrinogen 0·85%, 0·60–1·10, per minor allele). The associations with interleukin 6 and C-reactive protein were consistent across study-specific subgroups (appendix pp 34–37
) with no evidence of genotype-by-subgroup interaction (p>0·05 for all analyses). Concentration of soluble IL6R increased per minor allele (). The functional rs8192284 variant showed associations with interleukin 6, C-reactive protein, and fibrinogen that were directionally concordant with those of rs7529229 in the Whitehall II study (appendix p 24
). No significant association was noted between the rs7529229 SNP and concentration of total, LDL, and HDL cholesterol or triglycerides in analyses including up to 114 615 individuals ().
Association of the IL6R rs7529229 variant with (A) interleukin 6, (B) C-reactive protein, and (C) fibrinogen concentration
The blood markers interleukin 6, soluble IL6R, C-reactive protein, fibrinogen, and total, LDL, and HDL cholesterol were available in both genetic studies and tocilizumab treatment trials allowing a direct comparison of IL6R genotype and IL6R blockade (). The minor allele of rs7529229 and treatment with tocilizumab showed directionally concordant effects; each was associated with reduced C-reactive protein and fibrinogen and increased interleukin 6 and soluble IL6R (, ). Tocilizumab treatment increased circulating total, HDL, and LDL cholesterol, and triglycerides, but the IL6R rs7529229 SNP, by contrast, showed no significant association with any of these lipid fractions (, ). In randomised trials, tocilizumab increased concentrations of albumin and haemoglobin and decreased erythrocyte sedimentation rate (ESR), platelet count, and serum amyloid A (). The effect of rs7529229 was directionally concordant with that of tocilizumab on albumin, haemoglobin, and platelet count (, ). Data for ESR were unavailable in the genetic studies, but plasma viscosity (reflected by ESR) was lower in carriers of the rs7529229 minor allele (mean difference per allele −2·16×10−3 mPa.s, 95% CI −3·86×10−4 to −3·94×10−3, p=0·02; five studies, 15 589 individuals). Absence of data for serum amyloid A in the genetic analysis precluded comparison with tocilizumab treatment. In comparison of tocilizumab treatment with the rs7529229 variant, the direction of effect was concordant for nine of the ten biomarkers (, ), and greater than expected under the null hypothesis of no concordance (binomial test, p=0·01).
Associations of the minor allele of the IL6R SNP rs7529229 and tocilizumab (8 mg/kg) versus placebo with commonly reported biomarkers
We also examined the association of IL6R
variants with coronary heart disease. In a meta-analysis of 34 studies (25 458 coronary heart disease cases, 100 740 controls) the OR for the primary outcome (all fatal and non-fatal coronary heart disease events; appendix pp 15–16
) per minor allele at rs7529229 was 0·95 (95% CI 0·93–0·97, p=1·53×10−5
). There was low heterogeneity between studies (I2
=10%, 95% CI 0–41) and the effect estimates were consistent in prospective and case-control studies, including previously published data32
(). In a subset of 97 300 individuals (27 studies) for whom relevant data were available, the association of rs7529229 with the primary outcome (14 360 cases and 82 940 controls) was consistent in stratified analyses (appendix pp 1–7
) with no evidence for effect modification by any of these subgroups (appendix pp 40–41
Association of IL6R rs7529229 with risk of fatal and non-fatal coronary heart disease
Associations of rs7529229 with risk of fatal or non-fatal stroke (OR 0·98, 95% CI 0·95–1·02, p=0·30) in 6904 cases and 90 512 controls (27 studies) and with fatal or non-fatal cardiovascular disease events combined (OR 0·98, 95% CI 0·95–1·00, p=0·05) in 17 595 cases and 76 321 controls (26 studies) were suggestive but not compelling (). Up to three of six randomised trials of tocilizumab reported the incidence of cardiac or vascular events, or both, with median follow-up of 24 weeks (appendix pp 22–23
). However, imprecise endpoint definition and the small number of events prevented comparison with genetic studies.
Association of IL6R rs7529229 with secondary and safety endpoints
In an analysis of safety endpoints in tocilizumab trials, data suggested an increased risk of infection (OR 1·30, 95% CI 1·07–1·58) and increased concentrations of hepatic enzymes alanine transaminase and aspartate transaminase () with tocilizumab treatment compared with placebo (appendix pp 22–23
). By contrast with evidence for tocilizumab, genetic analyses (although in a relatively small subset) did not reveal any association with concentrations of aspartate transaminase () or in log γ-glutamyl transferase (relative difference in geometric mean per allele −0·64%, 95% CI −1·95 to 0·69, p=0·34; seven studies, 15 641 individuals). Our genetic experiment did not include infection as an outcome and published evidence for the IL6R
rs7529229 variant was scarce.33
Genome-wide association studies of tuberculosis34
and meningococcal disease35
have not reported associations of variants in IL6R
with risk of those outcomes.
Neither the evidence from tocilizumab trials nor the genetic studies to date have suggested an association of IL6R blockade with increased risk of cancer. The pooled OR for development of any cancer was 0·42 (95% CI 0·06–2·88; four cases and 1196 controls) for tocilizumab treatment in randomised trials, and was 0·98 (95% CI 0·93–1·03; 5376 cases and 57 123 controls) for the IL6R
rs7529229 variant. In published genome-wide association studies and new look-ups, the IL6R
rs7529229 variant showed no association with breast cancer (OR 1·01, 95% CI 0·94–1·10) or colorectal cancer (OR 1·03, 95% CI 0·96–1·12; ; appendix p 17
The IL6R rs7529229 variant was associated with lowered systolic blood pressure (per-allele β coefficient −0·21 mm Hg, 95% CI −0·37 to −0·05, p=0·01) and diastolic blood pressure (per-allele β coefficient −0·11 mm Hg, 95% CI −0·20 to −0·02, p=0·02) in 33 studies (112 979 individuals). There was suggestive evidence that the rs7529229 variant was associated with reduced risk of type 2 diabetes (OR 0·97, 95% CI 0·94–1·00, p=0·06) in 12 859 cases and 86 807 controls (), although this exploratory finding needs further investigation.