To describe the use of the LUNDEX, a new index for comparing the long-term efficacy and tolerability of biologic therapies in rheumatoid arthritis (RA) patients treated in clinical practice.
To describe the use of the LUNDEX, a new index for comparing the long-term efficacy and tolerability of biologic therapies in rheumatoid arthritis (RA) patients treated in clinical practice.
Patients (n = 949) with active RA that had not responded to at least 2 disease-modifying antirheumatic drugs (DMARDs) including methotrexate, in whom biologic therapy was being initiated, were included in a structured clinical followup protocol. The protocol included collection of data on diagnosis, disease duration, previous and ongoing DMARD treatment, and dates on which biologic treatment was started and terminated. In addition, data on efficacy measures used for calculating validated response criteria, i.e., the European League Against Rheumatism and American College of Rheumatology response criteria, were collected at fixed time points. Data were prospectively registered from March 1999 through January 2004. The LUNDEX, a new index combining the proportion of patients fulfilling a selected response criteria set with the proportion of patients adhering to a particular therapy, was designed to compare the efficacy of the different therapies.
Etanercept had higher overall LUNDEX values compared with infliximab, mostly because of a lower rate of adherence to therapy with infliximab. The relationship between the drugs was consistent irrespective of the response criteria used.
The LUNDEX is a valuable tool for evaluating drug efficacy in observational studies. It has the advantage of integrating clinical response as well as adherence to therapy in a composite value. Moreover, the LUNDEX has a practical and potentially universal application independent of diagnosis and response criteria.
Treatment of rheumatoid arthritis (RA) has undergone remarkable changes over the last few years, following introduction of biologic therapies, such as tumor necrosis factor (TNF) blockers and interleukin-1 receptor antagonists. Several randomized controlled clinical trials (RCTs) have provided documentation of the effectiveness of these drugs (1–14). However, these trials have limitations compared with observational studies. Due to the strict inclusion criteria frequently used, RA patients enrolled in RCTs are often limited to those with disease of moderate severity and without any significant comorbidity. Thus, results obtained in RCTs cannot be uncritically applied to clinical practice, since RA patients are heterogeneous regarding disease severity, duration, and comorbidity prior to therapy initiation (15–17). RCTs therefore have limitations in terms of external validity.
Moreover, RCTs are often restricted in duration and in the number of patients included, which in turn reduces the power for detecting long-term efficacy and tolerability. In addition, rare or comorbidity-associated side effects are difficult to detect. Finally, comparisons of different biologic therapies are mainly indirect, since RCTs often compare new drugs with conventional therapy.
Conversely, by using open observational studies according to a clinical protocol, it is possible to include patients continuously and without limits regarding number or comorbidity. Furthermore, observational studies allow the inclusion of different treatments in heterogeneous patient groups independent of industry support.
Previously, the concept of “adherence to therapy,” i.e., the number of patients continuing treatment with a drug, has been used to compare different drugs in observational studies (18–20). However, the adherence-to-therapy fraction provides information only about the proportion of patients receiving a drug, regardless of clinical response. Thus, only the subgroup of patients adhering to a treatment actually experiences a considerable clinical effect. To address these limitations, we developed the LUNDEX, which is an index of the proportion of patients who not only remain on a particular therapeutic regimen but also fulfill certain response criteria, such as the American College of Rheumatology 20% response criteria (ACR20) (21).
This study was observational in design and used a structured clinical protocol developed by the South Swedish Arthritis Treatment Group for monitoring new biologic therapies in RA (19). Herein we present the LUNDEX and describe its application in the evaluation of long-term efficacy and tolerability of etanercept and infliximab in RA patients treated in clinical practice.
The structured clinical protocol was developed from previous nationwide protocols for monitoring of Swedish patients with early RA, but was modified and extended to make it more suitable for monitoring of treatment outcomes. The inherent element of quality control characterizing the protocol met the legislative documentation required in Sweden; therefore, no formal ethics committee approval was necessary.
Patients eligible for the study had been diagnosed as having RA according to the clinical judgment of the treating physician. In a systematic review of 150 patients with a clinical diagnosis, we found that 98% fulfilled the American College of Rheumatology (formerly, the American Rheumatism Association) 1987 classification criteria for RA (22) (Kristensen LE, et al: unpublished observations).
The patients were treated at 8 centers in southern Sweden serving a population of ∼1.3 million individuals, during the period March 1999 through January 2004. Patients eligible for biologic therapy were selected by physicians based on disease activity and/or lack of success with glucocorticoid treatment. No specific level of disease activity was required; however, patients had to have received at least 2 disease-modifying antirheumatic drugs (DMARDs) including methotrexate without satisfactory response. The selection of particular treatment depended primarily on drug availability. Patients who had previously received biologic therapy were excluded from this study.
The dosage of the different drugs was based on the recommendations of the manufacturers. Etanercept 25 mg subcutaneously was administered twice weekly, while infliximab was infused at 3 mg/kg at 0, 2, 6, and 12 weeks and then once every 8 weeks. In the event of primary treatment failure (i.e., lack of response to treatment) or secondary treatment failure (initial clinical response followed by lack of response), the dosage of infliximab could be increased in steps of 100 mg to a maximum of 500 mg administered at 4–8-week intervals. At study enrollment, the following data were recorded: primary diagnosis, other rheumatic disease diagnoses, previous and concomitant DMARD treatment, and systemic prednisolone dosage.
Clinical data were prospectively collected at 0, 3, 6, and 12 months, and subsequently every 3–6 months. No patients were excluded from the study due to unavailability of data at any particular followup time. Initially and at each followup visit, the following data were recorded: Health Assessment Questionnaire (HAQ) score (23), patient-scored visual analog scale for pain and general health, physician's global assessment of disease activity on a 5-point scale (Evalglobal) (24), 28-joint tender and swollen joint counts, erythrocyte sedimentation rate, and C-reactive protein (CRP) level.
Any withdrawal from treatment was registered prospectively and the cause classified by the treating physician as either adverse event, lack of response/treatment failure, or miscellaneous. There were no predefined criteria for inefficacy: the decision was based on the judgment of the treating physician. The category “miscellaneous” mainly consisted of instances of poor compliance or patients moving away from southern Sweden. In cases in which the cause of treatment withdrawal was registered as both treatment failure and adverse event, the reason for withdrawal was classified as an adverse event.
To compare the efficacy of the different therapies we designed the LUNDEX, a new index that takes into account the proportion of patients fulfilling a set of criteria (e.g., the ACR20) and the proportion of patients adhering to a particular therapy. The LUNDEX is calculated as the fraction of patients adhering to therapy multiplied by the fraction of patients fulfilling the selected response criterion at a given time (Figure 1).
Adherence to therapy was calculated using life-table analysis. Improvement in the ACR20, the ACR50, and the ACR70 (21, 25) and European League Against Rheumatism (EULAR) responses (26) obtained using the 28-joint Disease Activity Score (DAS28) (27) were calculated at given followup times (28).
Demographic and baseline clinical characteristics were analyzed by Mann-Whitney U test for comparison of continuous variables and by Pearson's chi-square test for discrete or ordinal variables. Values are reported as the mean ± SD except where stated otherwise. Adherence to therapy was estimated using Kaplan-Meier plots and analyzed with a life-table technique using log rank statistics for comparing different treatments. ACR and DAS28 responses were analyzed by Pearson's chi-square test. P values less than 0.05 were considered significant.
During the observation period, 949 patients began receiving etanercept treatment (n = 309) or infliximab treatment (n = 640) for the first time. Fewer than 4% of the patients in each treatment group had incomplete clinical data at entry; these patients were excluded from efficacy calculations but remained in the Kaplan-Meier analyses for adherence. Demographic and clinical characteristics of the patients are summarized in Table 1. At baseline, several significant differences were found between treatment groups. The group receiving infliximab had a significantly lower mean HAQ score, CRP level, and DAS28 score compared with those receiving etanercept. Also, the patients receiving infliximab had significantly shorter disease duration and had received fewer previous DMARDs compared with the etanercept group. Furthermore, patients in the infliximab group were treated with concomitant methotrexate more frequently than patients in the etanercept group. Both groups consisted of patients with longstanding, treatment-resistant, severe RA.
|Etanercept (n = 309)||Infliximab (n = 640)||P|
|Female, no. (%)||253 (82)||481 (75)||0.021|
|Age, years||55.1 ± 13||56.2 ± 14|
|Disease duration, years||14.7 ± 10.1||12.7 ± 10.0||<0.001|
|No. of previous DMARDs including MTX||4.2 ± 2.05||3.6 ± 1.98||<0.001|
|MTX at entry, no. (%)||96 (31)||467 (73)||<0.001|
|Weekly MTX dosage among patients receiving MTX, mg||15.7 ± 5.1||14.0 ± 5.92||NS|
|DAS28, 0–10||5.9 ± 1.06 (n = 297)†||5.6 ± 1.20 (n = 615)†||<0.001|
|HAQ, 0–3||1.6 ± 0.64 (n = 301)†||1.4 ± 0.62 (n = 622)†||0.002|
|C-reactive protein, mg/dl (normal <0.8)||4.0 ± 3.66 (n = 298)†||3.6 ± 3.73 (n = 628)†||0.044|
The LUNDEX, using ACR20 response, ACR50 response, ACR70 response, EULAR good response, and EULAR moderate response, was calculated in each treatment group at 3, 6, 12, 24, and 36 months. Data on ACR20 and ACR50 response and on EULAR good and EULAR moderate response are presented in Figures 2A–D. For comparison, Figure 2 also shows the proportion of responders at each followup calculated using the per-protocol technique, i.e., the proportion of responders of those actually evaluated. Etanercept had the highest overall LUNDEX values, with nearly 55% of patients who started etanercept treatment fulfilling ACR20 response criteria during the first year. This fraction declined to ∼40% after 3 years of followup. In contrast, ∼45% of patients who started infliximab treatment fulfilled ACR20 response criteria at 12 months, and this dropped to ∼30% after 3 years of followup.
Kaplan-Meier curves for estimated adherence to each therapy are shown in Figure 3A. Infliximab had a significantly lower level of adherence compared with etanercept (P < 0.001).
Figures 3B and C display the proportion of patients who withdrew from each treatment group due to adverse events (Figure 3B) or treatment failure (Figure 3C). The main reason for withdrawal from treatment with infliximab was adverse events (P < 0.001), but there was also a significantly higher rate of withdrawal due to treatment failure compared with etanercept (P = 0.018). The reason for withdrawal from etanercept therapy was equally distributed between treatment failure and adverse events.
There were no significant differences between the treatment groups in the frequency of treatment withdrawal for the reason “miscellaneous.” The proportions of patients withdrawing for this reason were 1.3% and 4.3% in the etanercept and infliximab groups, respectively.
The proportion of patients fulfilling ACR20, ACR50, ACR70, and EULAR DAS28 responses is shown in Table 2. As judged by ACR20 response, etanercept showed significantly higher response rates compared with infliximab at 3, 6, and 12 months (P < 0.001, P = 0.002, and P = 0.001, respectively). Etanercept also showed a significantly higher response rate at 12 months (P = 0.011). A statistically significant difference in response to etanercept versus infliximab as judged by EULAR moderate response was also found at 3 months of followup (P = 0.033).
|Criteria, followup time||Etanercept||Infliximab||P|
There were no significant differences between treatment groups with regard to change in prednisolone dosage during the observation period (data not shown). Patients with missing efficacy data at certain followup points did not exhibit differences in response rates compared with patients whose followup records were complete.
This report introduces the LUNDEX as a suitable index for comparing biologic therapies in observational studies. As illustrated in Figures 2A–D, the LUNDEX yields considerably lower values compared with the per-protocol technique. Previously, evaluation of response rates at fixed times of followup using intent-to-treat analyses with last observation carried forward (LOCF) has been used to compare biologic treatments (1–10). Both LOCF and completer (per-protocol) analyses inflate the apparent proportions of responses in clinical studies. In observational investigations, study of patients fulfilling particular response criteria, in isolation (Table 2), does not yield information about the true fraction of patients actually responding to a particular therapy, since not all patients adhere to the different therapeutic regimens. Therefore, the response rates observed in such studies reflect drug performance in selected groups of patients, but without accounting for differences in dropout among the treatment groups.
In many RCTs this problem is solved by using intent-to-treat analysis with carry forward techniques. In observational studies this type of analysis is inappropriate. Patients are continuously entering and exiting the study, and some patients switch to different treatment groups during the observation period.
In addition, clinical information necessary for calculating response criteria is sometimes missing for dropouts. To address this problem we developed the LUNDEX for assessment of drug efficacy in RA patients. The LUNDEX provides a unifying concept of the fraction of patients, among those adhering to therapy, who truly achieve a specific response criterion after a defined followup time. It is easy to utilize, and is calculated by multiplying the proportion of patients adhering to therapy by the proportion fulfilling a particular response criteria set (Figure 1). In this way, the LUNDEX can be applied without having to use intent-to-treat analysis, which thus facilitates the process of evaluating therapies in clinical observational studies in which specific treatments are continuously being initiated and discontinued in individual patients. Furthermore, the concept of the LUNDEX is not limited to RA and to ACR or EULAR response criteria. It is a universal index that can be applied to evaluate drug efficacy in other well-defined diseases with validated response criteria.
In this study, the treatment groups were not precisely matched because of the observational design. Therefore, several significant differences were noted at baseline. From a clinical perspective, however, the groups were quite similar. Both groups were dominated by patients with long disease duration, unsuccessful treatment with several previous DMARDs, and marked disability as well as high disease activity. Therefore, we believe limited comparisons of the different treatments are justifiable. The present comparison is not hampered by inclusion of patients previously treated with biologic agents, and the indications for anti-TNF therapy as well as concomitant DMARD therapy remained stable throughout the study period.
It is therefore reasonably clear from the results of this study that patients treated with etanercept have higher LUNDEX values than those treated with infliximab. The main reason for this finding is the lower level of drug adherence in patients treated with infliximab. In turn, this lower adherence in the infliximab group was explained mostly by withdrawals because of adverse events, but there also seemed to be more instances of lack of efficacy with this treatment. However, the lower degree of disease activity at baseline in the infliximab group (lower CRP level, DAS28 score, and HAQ score [Table 1]) may account, to some extent, for the lower rate of adherence to therapy and lower LUNDEX values observed among patients in this group, since their potential for improvement was lower. Conversely, the infliximab group also included a significantly higher proportion of patients receiving concomitant methotrexate at baseline, and concomitant treatment with anti-TNF therapy and methotrexate has been reported to be a more efficacious regimen compared with monotherapy (14).
In accordance with previous reports (29, 30), we did not observe any consistent differences in per-protocol response rates between the therapies. Etanercept showed significantly better responses at some followup points for ACR20, ACR50, and EULAR (moderate) response when compared with infliximab. However, the lack of consistency combined with the heterogeneity of the baseline population makes it less likely that etanercept truly yields a better clinical response than infliximab.
The open, nonrandomized nature of this study may introduce bias, both in the process of collecting data and during the selection of patients for particular treatments (17, 31). In order to minimize observational bias, all data entries were centralized, thereby ensuring uniform interpretation of information in registration forms. Possible confounding by indication cannot be excluded in this study. However, data directly comparing the different biologic drugs are sparse (31), thus giving no obvious reason for favoring prescription of one over another.
A placebo effect, improving the response to the drugs, may be expected. However, there is no reason to believe that this effect is distributed unevenly between treatment groups. In fact the rates of response to the treatments in this study were lower compared with findings of previous controlled clinical trials (1, 2, 6–10, 13, 14). This can be explained by the variation of glucocorticoid usage and the inclusion in this study of patients encountered in the clinical setting. Also, the long observation period of this study better reflects the chronic course of RA, and thus dilutes the bias that occurs in many RCTs, which focus on treatment of disease during flares (17).
We expect that the LUNDEX will become a valuable tool for evaluating results of observational studies. It is a practical measure with potentially universal application, independent of diagnosis and response criteria.
We are indebted to all colleagues and staff in the South Swedish Arthritis Treatment Group for their cooperation and provision of data.