Background
Rheumatoid arthritis (RA) is a chronic inflammatory disease, characteristic of persistent synovitis and destruction of bone and cartilage in multiple joints [
1]. Although its etiological causes are still unclear, the efficacy of specific proinflammatory cytokine targets clearly identifies these cytokines as relevant in RA pathogenesis. Since the first report of the tumor necrosis factor-α inhibitor (TNF-i) infliximab as the first biological disease-modifying antirheumatic drug (bDMARD) for treatment of RA in 1993 [
2], bDMARDs, including several TNF-i [
2‐
5], tocilizumab (TCZ) as an interleukin 6 (IL-6) inhibitor [
6,
7], anakinra as an IL-1 inhibitor [
8], abatacept (ABT) as a T cell co-stimulator inhibitor [
9], and rituximab as a B cell depleting monoclonal antibody [
10], have become available, and innovative progress in RA treatment has been made due to their potent clinical efficacy.
Currently, although several categories of bDMARDs are available, the preferable bDMARD for each RA case is not obvious. Therefore, TNF-i, TCZ, and ABT are basically treated as equivalent treatments in the recent disease management recommendations [
11,
12]. However, it is occasionally observed that some patients who do not respond to the first biological treatment have a more obvious response to other bDMARDs. To date, many studies that focused on switching bDMARDs have been reported [
5,
9,
13‐
16], and some authors recently suggested that switching to a bDMARD with a different mode of action may be more efficacious than switching to one that targets the same molecule [
15,
16]. These studies of bDMARD switching and the aforementioned occasional cases suggest that the dominant inflammatory cytokine that should be targeted may be different in each RA patient.
For the purpose of achieving early remission and efficient medical economics, it is desirable to predict the dominant cytokine that should be inhibited before starting anticytokine therapy. The objective of this study was to develop a scoring system based on common laboratory indices that could discriminate between individuals more likely to respond to either TCZ or a TNF-i.
Methods
Patients
First, 45 newly diagnosed Japanese RA patients were included to estimate the relative expression of IL-6 and TNF-α mRNA in the peripheral blood before therapeutic intervention. Then, another 98 Japanese RA patients receiving TCZ or a TNF-i from 2005 to 2010 at Iizuka Hospital were studied to develop a scoring system that could discriminate between individuals more likely to respond to either TCZ or a TNF-i. We also studied an additional 228 Japanese RA patients for the validation study. They were treated with TCZ or a TNF-i at Kyushu University Hospital, Nagasaki University Hospital, Sasebo-chuo Hospital, Japanese Red Cross Okayama Hospital, and Iizuka Hospital from 2011 to 2014, and the number of patients in each setting was 31, 18, 73, 17, and 89, respectively. All of the patients fulfilled the 1987 classification criteria of the American College of Rheumatology for RA. This study was approved by the ethics committee of Kyushu University Hospital, Nagasaki University Hospital, Sasebo-chuo Hospital, Japanese Red Cross Okayama Hospital, and Iizuka Hospital, and the principles of the Helsinki Declaration were followed throughout the study. The gene expression study in 45 newly diagnosed RA patients was also approved by the ethics committee of Iizuka Hospital and was performed in accordance with the “Ethical Guidelines for Human Genome/Gene Research” published by the Japanese Ministry of Education, Culture, Sports, Science and Technology (MEXT). Informed consent was obtained from all participants.
Analysis of gene expression level
In order to clarify the signature of gene expression at RA pathogenesis, peripheral whole blood was drawn from the 45 patients who were newly diagnosed and exposed neither to steroids nor to anti-rheumatic drugs. The samples were prepared and subjected to RNA extraction using the PAXgene system (QIAGEN, Germantown, MD, USA). Messenger RNA levels were then measured using Agilent whole human genome 60 K (Agilent Technologies, Santa Clara, CA, USA) and the log-transformed raw intensity data were normalized with a quantile algorithm. In order to evaluate the expression balance between IL-6 and TNF-α, correlation was tested between the relative mRNA levels of each gene.
Development of a scoring system that predicts the efficacy of TCZ
Twenty-seven RA patients who were administered TCZ and 71 RA patients who were administered a TNF-i (etanercept (ETN): 36 patients, infliximab (IFX): 25 patients, adalimumab (ADA): 10 patients) at Iizuka Hospital from 2005 to 2010 were retrospectively analyzed to investigate the correlation between the efficacy of treatment and the laboratory parameters immediately before commencing treatment. Patients administered conventional DMARDs or patients with previous use of other bDMARDs were included in this analysis. In each group (TCZ group or TNF-i group), disease activity score in 28 joints using erythrocyte sedimentation rate (DAS-ESR) at week 24 and at baseline was used for calculation of the DAS ratio (= DAS-ESR at week 24/DAS-ESR at baseline) as the measure of drug efficacy. Then, correlation was tested between the DAS ratio and each laboratory test result at baseline. In this process, patients with liver dysfunction or renal dysfunction were excluded, and patients with DAS remission at baseline were also excluded. Male hemoglobin (Hb) values were multiplied by 0.88 to correct for the difference between male and female Hb levels, based on the average ratio of male and female Hb levels in similarly aged Japanese people [
17]. Several items that were significantly correlated with the DAS ratio in the TCZ group but not in the TNF-i group were selected, and the cut-off values for each item were defined using receiver operating characteristic (ROC) analysis to separate the good responders to TCZ from the moderate responders and non-responders to TCZ. Then, a scoring system to predict the efficacy of TCZ was developed using these cut-off values. The validity of this scoring system was tested by applying this system to the clinical data on the aforementioned 98 patients at Iizuka Hospital.
The validation study of the scoring system in a second set of patients
Another test was performed to validate this scoring system by using the clinical data on 228 patients (TCZ: 129 patients, TNF-i: 99 patients (ETN: 31 patients , IFX: 25 patients, ADA: 26 patients, golimumab (GLM): 13 patients, certolizumab pegol (CZP): 4 patients)). These patients were treated with TCZ or a TNF-i at Kyushu University Hospital, Nagasaki University Hospital, Sasebo-chuo Hospital, Japanese Red Cross Okayama Hospital, and Iizuka Hospital from 2011 to 2014, and the number of patients in each setting was 31, 18, 73, 17, and 89, respectively. Their laboratory tests at baseline were scored by the aforementioned scoring system, and the association between the score and the efficacy of TCZ or a TNF-i was examined.
Statistics
The differences between two groups were analyzed using Student’s t test. If there were a significant difference between the variances of the two samples, the Wilcoxon rank-sum test was applied. Correlation between the IL-6 and TNF-α mRNA expression levels and the DAS ratio and other continuous variables was tested using Spearman’s rank correlation. The differences between the good responder and non-responder rates in the TCZ or TNF-i groups were examined using the chi-squared (χ2) test. All analyses were performed by JMP statistical software (SAS Institute). P < 0.05 was considered statistically significant.
Discussion
Currently, several types of bDMARDs are available; however, there are no guidelines on the appropriate treatment for individual RA patients according to the treatment mode of action, although there are some suggestions with respect to drug safety [
20]. Therefore, TNF-inhibitors, TCZ, and ABT are treated in the same way in recent disease management recommendations [
11,
12]. On the other hand, many studies focusing on switching bDMARDs showed that switching to a bDMARD with a different mode of action may be more efficacious than switching to one targeting the same molecule [
15,
16], suggesting that the dominant inflammatory cytokine that should be targeted may be different in each RA patient.
However, several studies have failed to identify the dominant cytokine in a specific individual with RA [
21]. In this study, we found that the relative mRNA levels of
IL-6 and
TNF-α were inversely correlated in blood samples from 45 newly diagnosed RA patients. As the half-life of mRNA is known to be five times shorter than for proteins [
22], it is possible that evaluation of the relative expression level of mRNAs may reveal timely power-balance between
IL-6 and
TNF-α in RA. It may not be universal in each clinical situation, because the blood samples were drawn from newly diagnosed patients before any therapeutic intervention. However, it is still possible that it might be one of the explanations for the better efficacy of TCZ after inadequate response to TNF-i in some patients. The idea led us to the following study to find key laboratory indices that may reflect the power-balance between
IL-6 and
TNF-α in RA.
Then, we found the values of Plt, iron, Hb, AST, and ALT were significantly correlated with the efficacy of TCZ treatment, and these correlations may reflect the predominant cytokine in RA pathogenesis, as described subsequently. For example, inflammatory thrombocytosis is a well-known phenomenon, and it is dependent on hepatic thrombopoietin (TPO) production stimulated by IL-6 [
23,
24]. The TPO produced mediates proliferative and anti-apoptotic signals via the janus kinase (JAK) and signal transducer and activator of the transcription (STAT) pathway by binding to the cellular homolog of the myeloproliferative leukemia virus oncogene (c-MPL, CD110) receptors on the surface of megakaryocytes, leading to megakaryocyte proliferation and thrombocytopoiesis [
25,
26]. Moreover, it was reported that administration of IL-6 to humans is associated with an increase in circulating platelet counts [
27,
28].
IL-6 is also known to induce inflammatory anemia by the production of hepcidin. Hepcidin is a peptide hormone synthesized mainly by hepatocytes and secreted in the plasma [
29]. It acts as a negative iron regulator that inhibits iron absorption from the duodenum and iron release from macrophages, leading to multiple iron metabolism disorders, including iron deficiency anemia [
30,
31]. IL-6 activates the JAK/STAT signaling pathway, which activates the hepcidin promoter [
31,
32]. In contrast, hepcidin production is not stimulated by TNF-α [
33].
It is also a known fact that IL-6 has a hepatoprotective effect. For example, IL-6 promotes hepatic survival in a variety of liver injury models, including Fas-mediated injury and toxic damage [
34‐
36]. It has also been shown that IL-6 is essential for liver regeneration after partial hepatectomy [
37,
38]. This hepatoprotective effect of IL-6 is mediated by gp130 and JAKs, leading to the activation of STATs, which mediates anti-apoptotic and anti-necrotic signals to hepatocytes [
39‐
42]. In addition, IL-6 is associated with liver regeneration via a mitogen-activated protein kinase (MAPK) pathway, which promotes hepatocyte proliferation [
42,
43]. Moreover, it was also reported that liver regeneration was mediated by platelets [
44], the production of which is promoted by IL-6 stimuli as described. Thus, IL-6 promotes hepatoprotection by various pathways. In contrast, TNF-α is a well-known, predominant mediator of hepatocyte apoptosis and liver injury by promoting caspase-8 activation and the mitochondrial death pathway via c-Jun N-terminal kinase 2 (JNK2) [
45‐
47]. Actually in the drug information sheets, mild increased serum AST and ALT was common after TCZ treatment (≤22% for AST, ≤36% for ALT), while it was uncommon after TNF-i treatment (generally ≤5%).
While Plt, Hb, AST, and ALT were correlated with TCZ efficacy, it was an interesting finding that C-reactive protein (CRP) was not correlated with the efficacy of TCZ, and this tendency was similar in the additional 228 patients (data not shown). Recently, it was reported that baseline serum concentration of CRP might not be predictive of clinical outcomes after TCZ treatment [
21]. Although the reasons for this phenomenon are unclear, such acute inflammatory markers might not be suitable for prediction because of their dispersion.
In this study, higher Plt and lower Hb, AST, and ALT levels were associated with better TCZ efficacy, and these results may be due to the aforementioned physiological functions of IL-6. Because these mechanisms are unique to IL-6, each value of Plt, Hb, AST, and ALT may reflect the cytokine balance of IL-6 and TNF-α. Recently, the efficacy of alternative IL-6 inhibitors other than TCZ has been reported [
48‐
50]. Our scoring system was developed with the result of the efficacy of TCZ; however, the aforementioned four items depend on the physiological function of IL-6 in general, so the efficacy of other IL-6 inhibitors may be predictable using our scoring system.
A verification test of the scoring system based on the original 98 patients and an additional 228 patients revealed that patients scoring 2 or more points had a significantly higher rate of good response and a lower rate of no response to TCZ compared with a TNF-i, and this tendency became more obvious if the score was 3 or more. These results suggest TCZ could have greater efficacy in patients with a score of 2 or more (approximately 50% of all patients), and especially those with a score of 3 or more (approximately 25% of all patients), compared to those with a score less than 2. These “TCZ-sensitive” patients could be distinguished in a very simple way using our scoring system, which is easily calculated using common laboratory findings.
This study has two limitations. First, it had a relatively small sample size. It would be better to have a large sample size to develop this type of scoring system. To overcome this disadvantage, we analyzed a second set of patients. Second, this was a retrospective study, and it was difficult to secure an adequate “wash-out” period when the treatment was switched from other bDMARDs. If the score was applied after an adequate wash-out period, it is possible that the predictive value of the scoring system might be more accurate. The execution of a prospective study is desirable in the future.
Competing interests
YK has received speaking fees from Chugai, BMS, Takeda, Ono, Mitsubishi Tanabe, Abbvie, Ayumi, and Nihon, and research funding from Eli-Lilly, Astellas, Daiichi Sankyo, and Mochida. AK has received speaking fees from Chugai, Abbvie, Eisai, Mitsubishi Tanabe, Pfizer, Jansen, Takeda, and Astellas, and research funding from Chugai, Abbvie, Eisai, Mitsubishi Tanabe, Pfizer, Jansen, Takeda, and Astellas. HT has received speaking fees from Astellas, Asahikasei, Janssen, Santen, Mitsubishi Tanabe, Nipponkayaku, Thermofisher diagnostics, Eisai, Eli-Lilly, Kissei, Takeda, Daiichi Sankyo, BMS, Chugai, Pfizer, and Abbvie, and research funding from Mitsubishi Tanabe, Eisai, Takeda, Daiichi Sankyo, Nipponkayaku, Teijin, Eli Lilly, Abbvie, and Chugai. TH has received speaking fees from Mitsubishi Tanabe, Chugai, Abbvie, Astellas, BMS, Takeda, Janssen, Pfizer, and Actelion, and research funding from Mitsubishi Tanabe, Chugai, Astellas, BMS, Takeda, Eli Lilly, CSL Behring, Asahi-kasei, Santen, Teijin, and Daiichi Sankyo. All other authors have declared no competing interest.