Analysis of Trial Data for Infliximab and Golimumab: Baseline C-Reactive Protein Level and Prediction of Therapeutic Response in Patients With Psoriatic Arthritis
Anti–tumor necrosis factor medications have demonstrated good efficacy in treating psoriatic arthritis (PsA). Clinical responses at the primary end point in recent clinical trials of golimumab in PsA subjects yielded lower American College of Rheumatology 20% criteria for improvement (ACR20) responses than those seen in the Infliximab Multinational Psoriatic Arthritis Controlled Trial 2 infliximab study. However, baseline C-reactive protein (CRP) levels of PsA subjects enrolled in these trials differed significantly. We hypothesized that baseline CRP levels predict the observed differing clinical response rates at the primary end point.
Combining the data from the infliximab and golimumab trials, we stratified the data by baseline CRP levels and then correlated these cohorts with response to treatment, specifically the Psoriasis Area and Severity Index and the ACR20, ACR50, and ACR70 responses.
The mean baseline CRP level in the infliximab trial was 1.9 mg/dl versus 1.4 mg/dl in the golimumab trial. Only 23% of golimumab subjects had a CRP level ≥1.7 mg/dl versus 35% of infliximab subjects (P = 0.0023). All subjects with a CRP level ≥1.7 mg/dl at baseline achieved higher ACR20, ACR50, ACR70 response rates (56%, 36%, and 18%, respectively) at the primary end point of 14 weeks in the infliximab and golimumab clinical trials than any subjects with a CRP level <1.7 mg/dl (48%, 28%, and 13%, respectively [P = 0.045, P = 0.027, and P = 0.089, respectively]).
These results demonstrate that an increased baseline CRP level predicted improved therapeutic response at the primary end point for subjects treated with both golimumab and infliximab. To allow for comparison between future drug development studies and meta-analyses, baseline CRP levels should be factored into the multivariate analysis.
Psoriasis (PsO) and psoriatic arthritis (PsA) are chronic inflammatory diseases that affect multiple organs, notably the skin and joints. PsA is associated with PsO and is accompanied by chronic arthritis, enthesopathy, seronegativity, and dactylitis. Although it has been debated by some that these diseases may be distinct, most evidence supports that PsO and PsA are closely related, or a disease within a disease (). Approximately 6–42% of patients with PsO also have PsA (see references in &lsqbr;&rsqbr;).
In the past, treatment for PsA was limited to nonsteroidal antiinflammatory drugs, regular physiotherapy, and traditional disease-modifying antirheumatic agents. The development of tumor necrosis factor (TNF)–blocking agents offered tremendous improvement in disease control of both joint and skin symptoms. TNF inhibitors have been shown in the previously described Infliximab Multinational Psoriatic Arthritis Controlled Trial 2 (IMPACT 2; infliximab) and Golimumab in Active Psoriatic Arthritis Despite Prior Conventional Therapy (GO-REVEAL; golimumab) clinical trials to be effective in treating both PsO and PsA ([2, 3]). Usage of these TNF inhibitors has increased rapidly, but there are few comparative studies that describe efficacies of the different TNF inhibitors in the treatment of PsA ([4, 5]), and available analyses do not correct for baseline patient demographics.
Differences in PsA response rates were noted in the GO-REVEAL study and IMPACT 2, with infliximab showing an improved effect as compared to golimumab. Because baseline C-reactive protein (CRP) levels have been shown to correlate with PsA improvement rates (), the differences in clinical response in recent clinical trials may plausibly reflect the baseline CRP levels in PsA patients.
We hypothesized that lower baseline CRP levels in the GO-REVEAL trial compared with the baseline CRP levels in the IMPACT 2 may predict the different primary end point clinical responses to golimumab and infliximab in patients with PsA.
Box 1. Significance & Innovations
- Clinical response in psoriatic arthritis clinical trials of infliximab and golimumab can be confounded by differing baseline C-reactive protein (CRP) levels.
- Subjects with low baseline CRP levels had a lower clinical response than subjects with higher CRP levels.
- The landmark clinical trial of infliximab (Infliximab Multinational Psoriatic Arthritis Controlled Trial 2) contained a higher percentage of subjects with higher CRP levels than in the landmark golimumab trial (Golimumab in Active Psoriatic Arthritis Despite Prior Conventional Therapy).
- Future drug development studies and meta-analyses should record and factor baseline CRP levels into the multivariate analysis.
Patients and methods
Data from these trials have already been published ([2, 3]). In these trials, adult patients with active PsA, despite previous therapy, were enrolled and assigned to receive either placebo or golimumab (GO-REVEAL) or infliximab (IMPACT 2) at the trial dose. Data for patients who had received treatment were requested from the sponsor of the IMPACT 2 and GO-REVEAL trial ([2, 3]). Primary end point data for these clinical trials were collected at week 0 and week 14.
Combining data from the infliximab and golimumab trials, we stratified the data by the baseline CRP levels. Data were divided into tertiles by the sponsor upon request, because direct access to the data was not permitted. We then correlated these stratified cohorts according to response to treatment, specifically the Psoriasis Area and Severity Index and the American College of Rheumatology 70% criteria for improvement (ACR70), 50% criteria for improvement (ACR50), and 20% criteria for improvement (ACR20) (data provided by Janssen) ().
Significance was calculated by studying the changes in the previously described response measures from baseline to week 14 using the Cochran-Mantel-Haenszel chi-square test for categorical variables, Student's t-test, and binomial distribution test. The level of significance was chosen to be a P value less than 0.05.
The inclusion criteria and handling of missing data were managed as previously described ([2, 3]). Demographic information for the previous studies are shown in Table 1. Statins and other concomitant medications were not considered in the analysis, and cohort sizes were sufficiently large to produce significant results.
Table 1. Characteristics of patients at baseline in the GO-REVEAL trial and IMPACT 2*
|Men, no. (%)||174 (60)||71 (71)|
|Age, mean ± SD years||47.0 ± 10.8||47.1 ± 12.8|
|Swollen joints, mean ± SD (range 0–66)||13.0 ± 9.9||13.9 ± 7.9|
|Tender joints, mean ± SD (range 0–68)||23.2 ± 16.4||24.6 ± 14.1|
|C-reactive protein level, mean ± SD mg/dl||1.4 ± 1.7||1.9 ± 2.1|
|PsA subtype, no. (%)|| || |
|Arthritis involving distal interphalangeal joints||46 (15.8)||26 (26)|
|Arthritis mutilans||3 (1.03)||1 (1)|
|Asymmetric peripheral arthritis||93 (32.1)||18 (18)|
|Polyarticular arthritis||118 (40.7)||53 (53)|
|Spondylitis with peripheral arthritis||43 (11.0)||2 (2)|
|>3% BSA affected by psoriasis, no (%)||217 (74.8)||83 (83)|
|PASI score, mean ± SD (range 0–72)||10.5 ± 8.0||11.4 ± 12.7|
|Patients taking MTX, no. (%)||140 (48.3)||47 (47)|
|Patients taking oral corticosteroids, no. (%)||46 (15.9)||15 (15)|
|Patients taking NSAIDs, no. (%)||220 (75.9)||71 (71)|
In total, 290 subjects from the GO-REVEAL trial (golimumab) and 100 subjects from the IMPACT 2 (infliximab) were examined. The baseline demographics between the 2 trials were similar (Table 1), with the exception of an increased percentage of men in the IMPACT 2 (71% versus 60% in the GO-REVEAL trial), which was statistically significant (P = 0.03).
Data for the GO-REVEAL study were divided into tertiles by the study sponsor, with CRP levels of >1.2 mg/dl (94 subjects), between 0.3 mg/dl and 1.2 mg/dl (96 subjects), and <0.3 mg/dl (100 subjects). The IMPACT 2 study sponsor also divided the data into even tertiles, with CRP levels of ≥1.7 mg/dl (35 subjects), between 0.65 mg/dl and <1.7 mg/dl (29 subjects), and <0.65 mg/dl (36 subjects). To evaluate both trials simultaneously, the CRP levels in the IMPACT 2 were used to split data from the GO-REVEAL study. Incorporating subject data from both trials resulted in a total of 102 subjects with a CRP level ≥1.7 mg/dl, 199 subjects with a CRP level between 0.65 mg/dl and 1.7 mg/dl, and 89 subjects with a CRP level <0.65 mg/dl. An analysis was performed for CRP level thresholds at 0.65 mg/dl and 1.7 mg/dl, but only the data for 1.7 mg/dl are shown, for clarity.
Subjects with low baseline CRP levels had a lower clinical response than subjects with higher CRP levels. The CRP level in 90% of normal individuals is <3.0 mg/dl and does not exceed 1.7 mg/dl. The response rates for golimumab and infliximab patients with baseline CRP levels <1.7mg/dl were compared. When evaluating the ACR20, ACR50, and ACR70 response rates for golimumab versus infliximab, there was no significant difference in response rates at these low CRP levels (P > 0.05 for all) (Table 2). Thus, for subjects with low baseline CRP levels (<1.7 mg/dl), infliximab and golimumab demonstrated nearly the same efficacy.
Table 2. Patient response rates in the GO-REVEAL trial versus the IMPACT 2 at week 14*
|PASI75||108/223 (48.4)||25/67 (37.3)||38/65 (58.5)||23/35 (65.7)|
|ACR20||106/223 (47.5)||33/67 (49.3)||32/65 (49.2)||24/35 (68.6)|
|ACR50||63/223 (28.3)||22/67 (32.8)||19/65 (29.2)||15/35 (42.9)|
|ACR70||30/223 (13.5)||13/67 (19.4)||8/65 (12.3)||5/35 (14.3)|
In contrast, subjects with higher CRP levels at baseline achieved higher ACR20, ACR50, ACR70 response rates, irrespective of the study drug. After combining both infliximab and golimumab data, the response rates for subjects with high CRP levels (≥1.7 mg/dl) were 56% (ACR20), 36% (ACR50), and 18% (ACR70) at the primary end point of 14 weeks compared to those with CRP levels <1.7 mg/dl (48% [ACR20], 28% [ACR50], and 14% [ACR70]; P = 0.045, P = 0.027, and P = 0.089, respectively) (Table 3). Regardless of the TNF blocker used, patients with higher baseline CRP levels had greater ACR20, ACR50, and ACR70 response rates.
Table 3. Percentage of subjects who achieved at least a PASI75, ACR20, ACR50, or ACR70 response rate in the GO-REVEAL trial and IMPACT 2 at week 14*
|PASI75|| || ||146/288 (50.6)|| || ||48/102 (47.1)|
|ACR20||80/174 (46.0)||58/114 (50.9)||138/288 (47.9)||43/71 (60.6)||14/31 (45.2)||57/102 (55.9)|
|ACR50||50/174 (28.7)||32/114 (28.1)||82/288 (28.5)||26/71 (36.7)||11/31 (35.5)||37/102 (36.3)|
|ACR70||29/174 (16.7)||9/114 (7.89)||38/288 (13.2)||14/71 (19.7)||4/31 (12.9)||18/102 (17.6)|
There was a higher percentage of subjects in the infliximab study with a higher CRP level. Within the infliximab study, the mean CRP level was 1.9 mg/dl. The mean baseline CRP level in the golimumab trial was lower, at 1.4 mg/dl. In the golimumab trial, 23% (67 of 290 subjects) had CRP levels ≥1.7 mg/dl (mean CRP level 3.79 mg/dl), and in the infliximab trial, 35% (35 of 100 subjects) had CRP levels ≥1.7 mg/dl (mean CRP level 4.07 mg/dl), with a statistically significant difference in means (P = 0.002). Overall, there were more severely affected subjects, as measured by CRP level, in the infliximab trial than in the golimumab trial.
A multivariate analysis was performed with male sex and CRP level. The proportion of men was 60% in the GO-REVEAL trial and 71% in the IMPACT 2. The response rates for men versus response rates for women were statistically similar (Table 3), except for the ACR70 response rate for subjects with CRP levels <1.7 mg/dl, in which only 7.89% of women achieved this level versus 16.7% of men (P = 0.003).
There is growing evidence that biomarkers, such as high-sensitivity CRP (hsCRP), help predict disease progression and therapeutic response. In particular, hsCRP has been shown to be independently associated with PsA (), and high CRP levels were shown to protect against treatment termination in both PsA () and rheumatoid arthritis trials (see references in &lsqbr;&rsqbr;). These studies have shown that clinical response and continuation of patients with PsA to treatment are uniformly correlated to the baseline CRP level and that increased systemic inflammation indicates a greater potential for treatment response (). While there is no clear evidence indicating that CRP level is directly correlated to PsO or PsA disease severity, some studies have suggested a possible relationship. Interestingly, some studies have demonstrated a correlation between CRP level and response rates of PsA to treatment ().
In comparing the infliximab trial results to those of the golimumab trial, we observed a higher percentage of subjects with a higher baseline inflammatory burden as measured by higher CRP levels. By pooling the golimumab and infliximab trial data, it was similarly evident that higher baseline CRP levels predicted higher ACR20, ACR50, and ACR70 responses for CRP levels ≥1.7 mg/dl.
Our study demonstrated that clinical trial subjects in 2 phase III studies receiving different TNF blockers had significantly different baseline CRP levels. This difference was shown to affect treatment response. Although it is tempting to compare clinical trials of drugs with similar pharmacotherapy targets, baseline subject demographics must be controlled for in a comparative analysis. A correlation between baseline CRP levels and response to TNF blockers has been previously observed in ankylosing spondylitis () and Crohn's disease (). Interestingly, one large observational study showed that increased CRP levels at baseline are sometimes associated with good treatment response and drug survival, that is, the length of time a drug is used before termination (). However, a similar analysis of the GO-REVEAL trial alone demonstrated a correlation between baseline CRP levels and PsA response to treatment (). Studies have shown that the CRP level decreases with clinical improvement (), as do other inflammatory biomarkers, such as complement (). Wagner et al () analyzed 92 biomarkers, including CRP level, adiponectin, apolipoprotein C-III, serum glutamic oxaloacetic transaminase, and TNFα, and found that CRP level was able to significantly predict clinical response to golimumab, but that combinations of biomarkers were also effective at predicting response. Biomarkers may serve as useful predictive measures in future comparative analyses of trials of therapy with biologic agents. To our knowledge, no clinical trials currently control for baseline CRP levels and their effect on the subject response. Existing published meta-analyses comparing biologic agents thus far did not correct for different baseline CRP levels ([4, 5]).
Some studies have shown a relationship between therapeutic response and baseline demographics. Demographic information was collected and showed that the GO-REVEAL and IMPACT 2 subjects had similar baseline characteristics, except for sex and CRP level. Some studies have shown that male sex increased drug survival times, or the length of time a drug would be used before termination, in PsA (). In contrast, other studies have shown that female sex improved drug survival or that there was no effect (). In our study, a higher percentage of men than women were enrolled, but a sex breakdown of response rates (Table 3) showed no statistically significant difference in response rates, except for achievement of the ACR70 response in subjects with CRP levels <1.7 mg/dl.
Furthermore, it should be noted that the IMPACT 2 preceded the GO-REVEAL trial. Because of the increased availability of effective therapies since the IMPACT 2, subjects entering more recent clinical trials may have a decreased disease burden and consequently lower baseline inflammatory burden (CRP level).
In conclusion, these results demonstrate that clinical response in PsA clinical trials can be confounded by baseline CRP levels. The results from different studies cannot be compared unless their baseline CRP levels are factored into the comparison, such as by stratification of response by baseline CRP levels. Given the effect of CRP levels on treatment response, future drug development studies or meta-analyses should account for baseline CRP levels to ensure consistency and to facilitate comparison.
All authors were involved in drafting the article or revising it critically for important intellectual content, and all authors approved the final version to be published. Dr. Au had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Study conception and design. Au, Gottlieb.
Acquisition of data. Au, Gottlieb.
Analysis and interpretation of data. Au, Ramirez-Fort, Gottlieb.
ROLE OF THE STUDY SPONSOR
The authors requested data from Janssen Inc, and Janssen Inc reviewed the manuscript before submission. Janssen Inc had no role in the study design or in the analysis or interpretation of the data, the writing of the manuscript, or the decision to submit the manuscript for publication.