Safety of Tocilizumab vs. TNF Inhibitors in RA
In this study, we conducted a direct comparison of the safety of TCZ with TNFIs in clinical practice, using a prospective, multi-center cohort with the largest possible number of patients. We demonstrated that the unadjusted IR of SAEs was not significantly higher in the TCZ group compared with the TNFI group, whereas the unadjusted IR of SIs of the TCZ group was 3.5-fold higher than the TNFI group. However, after adjusting for covariates, the use of TCZ compared to the use of TNFIs was not significantly associated with the development of SAEs or SIs.
Some studies have investigated the safety of TCZ in RA patients. It has been reported that the IR of SAEs was 20 to 30/100PY and that the most frequent SAE was infection (5 to 9/100PY). In the present study, the IRs of SAEs (21.36/100PY) and SIs (10.68/100PY) were similar to those of previous reports. The most frequently reported category of SAE in our study was infection and the incidence rate of non-pulmonary infection in the TCZ group was conspicuously higher compared to the TNFI group (7.57/100PY versus 1.73/100PY). Among non-pulmonary infections, skin and bone and joints were common sites in the TCZ group. Previous studies also reported that skin infections, as well as pulmonary infections, were frequently observed in patients treated with TCZ. Although the reasons for the high incidence rates of these types of infections in patients given TCZ have not been explained, special attention should be paid, not only to pulmonary infections, but also to skin infections in TCZ users.
We found no increased risk for the use of TCZ compared to the use of TNFIs for the development of SIs after adjusting for covariates at baseline. It is notable that the unadjusted IR of SIs in the TCZ group (10.7 (7.02 to 15.6)) was significantly increased compared to the TNFI group (3.53 95%CI, 1.52 to 8.18); this can be explained by several factors. The multivariate analysis indicated that the differences in clinical characteristics of the patients between the two groups influenced the difference in IRs of SIs (Table 4). The use of oral corticosteroids (prednisolone-equivalent dose) ≥5 mg/day was a significant risk factor for SIs in our study. Previous studies have reported that use of oral corticosteroids significantly increased the risk of SIs in patients undergoing treatment with biologics. Patients in the TCZ group of our study used concomitant corticosteroids more frequently. It has been shown that the presence of comorbidities increased the risk of SIs in RA patients. Although the HR of comorbidities was 2.20 in our study, it did not achieve statistical significance (P = 0.067). Relatively more patients in the TCZ group than in the TNFI group had at least one comorbidity (34.1% for the TCZ group, 27.3% for the TNFI group, P = 0.069). These data indicate that patients in the TCZ group may be more predisposed to infections than those in the TNFI group.
The low IR of SIs in the TNFI group apparently contributed to the increased IRR of SIs. The IR of SIs in the TNFI group in our study (3.03/100 PY) was lower than in previous studies (5 to 6/100PY), resulting in an increased IRR when comparing TCZ and TNFIs. We previously reported a significant decrease over time of the risk for SIs with TNFI treatment, possibly explained by evidence-based risk management of RA patients given TNFIs. In the present study, patients in the TNFI group started TNFIs in or after 2008, five years after the approval of IFX for RA in Japan. Information about risk of SIs in patients given TNFIs from observational studies has been extensively shared among Japanese rheumatologists, leading to improved risk management and, in consequence, lowered IRs for SIs in the TNFI group. To accurately compare the outcome between a new drug and an existing one, differences in the calendar year of drug approval should be considered. Therefore, in our study, we compared the use of two biologics, TCZ and TNFIs, in clinical practice during the same time period.
There are potential limitations of this study. First, we have to mention the possibility of selection bias. The patients in this study were enrolled from university hospitals or referral hospitals that are dedicated to the treatments of RA, which may indicate unidentified selection bias. However, because almost all patients who were registered from the participating hospitals to the all-cases post-marketing surveillance programs for each biological DMARD were enrolled in the REAL, selection bias was substantially low. Second, although there is concern about information bias, such as recall bias and reporting bias, in epidemiological studies in general, we collected patient data using the same case report form prospectively, which should overcome the misclassification and underestimation of SAEs derived from these types of bias. Third, clinical practice is always accompanied by the indication bias occurring when a drug is preferentially prescribed to patients with different baseline characteristics. In this study, it was notable that the difference in the percentage of patients who were given MTX at baseline between the two groups was significant, which would have affected the results of our study. To address this possibility, we estimated the risk of SAEs and SIs in patients with concomitant MTX in addition to the whole study population, and found them to be similar. Fourth, we did not investigate the comparison of effectiveness between the two groups due to incomplete data about disease activity in some patients.
Discussion
In this study, we conducted a direct comparison of the safety of TCZ with TNFIs in clinical practice, using a prospective, multi-center cohort with the largest possible number of patients. We demonstrated that the unadjusted IR of SAEs was not significantly higher in the TCZ group compared with the TNFI group, whereas the unadjusted IR of SIs of the TCZ group was 3.5-fold higher than the TNFI group. However, after adjusting for covariates, the use of TCZ compared to the use of TNFIs was not significantly associated with the development of SAEs or SIs.
Some studies have investigated the safety of TCZ in RA patients. It has been reported that the IR of SAEs was 20 to 30/100PY and that the most frequent SAE was infection (5 to 9/100PY). In the present study, the IRs of SAEs (21.36/100PY) and SIs (10.68/100PY) were similar to those of previous reports. The most frequently reported category of SAE in our study was infection and the incidence rate of non-pulmonary infection in the TCZ group was conspicuously higher compared to the TNFI group (7.57/100PY versus 1.73/100PY). Among non-pulmonary infections, skin and bone and joints were common sites in the TCZ group. Previous studies also reported that skin infections, as well as pulmonary infections, were frequently observed in patients treated with TCZ. Although the reasons for the high incidence rates of these types of infections in patients given TCZ have not been explained, special attention should be paid, not only to pulmonary infections, but also to skin infections in TCZ users.
We found no increased risk for the use of TCZ compared to the use of TNFIs for the development of SIs after adjusting for covariates at baseline. It is notable that the unadjusted IR of SIs in the TCZ group (10.7 (7.02 to 15.6)) was significantly increased compared to the TNFI group (3.53 95%CI, 1.52 to 8.18); this can be explained by several factors. The multivariate analysis indicated that the differences in clinical characteristics of the patients between the two groups influenced the difference in IRs of SIs (Table 4). The use of oral corticosteroids (prednisolone-equivalent dose) ≥5 mg/day was a significant risk factor for SIs in our study. Previous studies have reported that use of oral corticosteroids significantly increased the risk of SIs in patients undergoing treatment with biologics. Patients in the TCZ group of our study used concomitant corticosteroids more frequently. It has been shown that the presence of comorbidities increased the risk of SIs in RA patients. Although the HR of comorbidities was 2.20 in our study, it did not achieve statistical significance (P = 0.067). Relatively more patients in the TCZ group than in the TNFI group had at least one comorbidity (34.1% for the TCZ group, 27.3% for the TNFI group, P = 0.069). These data indicate that patients in the TCZ group may be more predisposed to infections than those in the TNFI group.
The low IR of SIs in the TNFI group apparently contributed to the increased IRR of SIs. The IR of SIs in the TNFI group in our study (3.03/100 PY) was lower than in previous studies (5 to 6/100PY), resulting in an increased IRR when comparing TCZ and TNFIs. We previously reported a significant decrease over time of the risk for SIs with TNFI treatment, possibly explained by evidence-based risk management of RA patients given TNFIs. In the present study, patients in the TNFI group started TNFIs in or after 2008, five years after the approval of IFX for RA in Japan. Information about risk of SIs in patients given TNFIs from observational studies has been extensively shared among Japanese rheumatologists, leading to improved risk management and, in consequence, lowered IRs for SIs in the TNFI group. To accurately compare the outcome between a new drug and an existing one, differences in the calendar year of drug approval should be considered. Therefore, in our study, we compared the use of two biologics, TCZ and TNFIs, in clinical practice during the same time period.
There are potential limitations of this study. First, we have to mention the possibility of selection bias. The patients in this study were enrolled from university hospitals or referral hospitals that are dedicated to the treatments of RA, which may indicate unidentified selection bias. However, because almost all patients who were registered from the participating hospitals to the all-cases post-marketing surveillance programs for each biological DMARD were enrolled in the REAL, selection bias was substantially low. Second, although there is concern about information bias, such as recall bias and reporting bias, in epidemiological studies in general, we collected patient data using the same case report form prospectively, which should overcome the misclassification and underestimation of SAEs derived from these types of bias. Third, clinical practice is always accompanied by the indication bias occurring when a drug is preferentially prescribed to patients with different baseline characteristics. In this study, it was notable that the difference in the percentage of patients who were given MTX at baseline between the two groups was significant, which would have affected the results of our study. To address this possibility, we estimated the risk of SAEs and SIs in patients with concomitant MTX in addition to the whole study population, and found them to be similar. Fourth, we did not investigate the comparison of effectiveness between the two groups due to incomplete data about disease activity in some patients.