 |
|
|
 |
Editorial
Methotrexate: Vital New Information About a Middle Aged Drug
EDWARD H. GIANNINI, MD, DrPH, MSc,
Address reprint requests to Dr. Giannini. In this issue of The Journal 3 articles contribute substantially to our understanding of methotrexate (MTX) in rheumatoid arthritis (RA), i.e., therapeutic response to MTX1; effect of concomitant hydroxychloroquine on the bioavailability of MTX2; and whether serially determined liver concentrations of MTX and its chief metabolite affect hepatic damage scores, clinical efficacy, or laboratory determined toxicity3. THERAPEUTIC RESPONSE TO MTX Ortendahl, et al introduce us to the concept of the "therapeutic segment,"1 defined as the period of time between the start of a therapy and its discontinuation, addition of an alternative agent, or initiation of a new treatment. Such an approach allows one to study, in a relatively pure state, the time to earliest response, to maximum response, to loss of response, and disability averted. The model can also be used to study other patient centered outcomes, such as pain and toxicity-unit years and cumulative dollar costs. Conventional randomized clinical trials generally do not permit such estimates of a therapeutic response profile because nearly all trials are severely "right censored." That is, the assessment of what happens further out in time [to the right in a graph of time (x axis) versus response (y axis)] is cut off due to the trial's fixed (relatively short) duration. One's ability to employ the therapeutic segment approach is contingent upon having several resources with which to work. A large, carefully monitored cohort of patients to legitimately employ the required statistical tools is mandatory. In this regard, the Arthritis, Rheumatism and Aging Medical Information System (ARAMIS) parent cohort of 4253 patients — 437 of whom had new MTX starts during the 10 year period examined — provides an unparalleled resource. Validated response measurement tools must exist for the disease under study and these tools must be given at standardized intervals during the segment. The Stanford Health Assessment Questionnaire (HAQ) used by Ortendahl, et al provides such a tool. It has been serially administered to thousands of patients, and, as a measure of disability, found to be sensitive to change and highly correlated with a variety of clinical outcomes4. In addition, the ARAMIS centers have employed standardized patient followup protocols for over 20 years, thus adding to the feasibility of doing such a study and increasing the credibility of results. Truly remarkable evidence of the rigor with which patients were followed is the fact that not one of the 2669 disability indexes (DI) required in this study was missing. Unfortunately, existing cohorts of carefully monitored patients who have been assessed at standardized intervals using validated outcome tools are lacking to conduct therapeutic segment studies for even the most common diseases seen by the rheumatologists, both adult and pediatric. Thus, the therapeutic segment approach is likely to be possible for very few agents, at least at present. Nevertheless, the concept is sound and should be considered a major addition to the methodological tools available to the field of evidence based medicine and clinical effectiveness research. Undoubtedly, MTX remains the drug of first choice after failure of nonsteroidal antiinflammatory drugs (NSAID), both in adult RA and in juvenile idiopathic arthritis1,5. Therefore, the information presented for this agent should be viewed in itself as a major advance in our understanding of the treatment of RA. As the field awaits biomedical research to elucidate the definitive genetic profiles that distinguish "complete responders" from "partial responders" from "non-responders", rheumatologists continue to grapple with the most appropriate sequence of therapy, as well as the time a patient should be given in which to respond. Importantly, statistical methodologies already well developed in theoretical mathematical statistics in the areas of sequential theory and sequential estimation may help in addressing these issues. Compounding the conundrum is the increased use of combination therapy and the rapid fire appearance on the market of new, potentially more efficacious agents. While Ortendahl, et al make no pretense about resolving these questions (e.g., should biologics be added earlier in an attempt to "ratchet down" the disease even quicker?), they do spell out in exquisite detail the shape, magnitude, and other characteristics of the pure MTX response and offer some guidance for its use. Further discussion and interpretation of the figures presented by Ortendahl, et al are necessary. Their Figure 1 shows the rise in MTX use and a corresponding fall in HAQ DI scores through the study period (1988-1998). The authors are careful to remind the reader that not all the patients whose DI appear in the graph received MTX. It is likely that the development of more effective treatment regimens, such as combination therapy and the appearance of biologics towards the end of the study period, account for at least some of the decrease in the DI through time. Their Figure 2 graphs the regression of mean DI on time and introduces generalized estimating equations (GEE) (an extension of general linear models) now being used in analysis of medical event prediction6. GEE take into account the fact that multiple measurements of the DI in the same patient through time are correlated with each other, rather than being completely independent (i.e., this is a longitudinal dataset). GEE may also take into account that patients were followed for differing periods of time (segment length differed). Finally, this is a "clustered" sample in that the data are from 8 different geographic centers and this may also have been taken into account. Yet, the authors do not specify exactly which of these facts were considered by GEE. The authors do point out the link function, distribution assumptions, and covariance structure. However, the reasons for choosing the polynomials that they did (cubic, quadratic; also true for their Figures 3 and 4) are not given. It is likely that the polynomials chosen fit the data better than higher order polynomials or other nonlinear models. The authors note that they were surprised to find that improvement in the DI continued into the fourth year of the therapeutic segment. Plateau of response did not occur until 30 to 42 months after therapy began, after which a rise in the DI was detected. But how can this be, when the general consensus in the literature is that improvement to be appreciated by MTX generally occurs up to about 6 months, at which time a plateau is reached7,8? Even the accompanying article by Carmichael, et al cites this belief2. Several explanations are possible including that MTX doses rose gradually through the observation period of the study — a segment in real time when rheumatologists were becoming more comfortable giving higher doses of the drug. Yet, the authors address this by allowing time into a regression model and found MTX dosage not to be a predictor of HAQ disability. Other possible explanations for this apparent anomaly include continued improvement in range of motion and strength that occurs after 6 months due to longterm disease suppression, interstudy differences in the patient populations, use of differing measures of therapeutic success, differing MTX dosage regimens (perhaps the more recent studies got to an ideal dose of MTX faster than was done in 1988), and differences in concurrent therapy that may have affected the bioavailability of MTX. Their Figure 4 addresses another important source of bias recognized by the investigators. "Right-censoring" (discussed above) can influence whether a drug appears more effective, or less effective, depending upon whether those who responded better or worse change treatments more quickly. In this study, patients could only be right-censored if they continued MTX until the end of the 10 year followup. Figure 4 seems to quell this concern in that the mean DI of the 2 groups (i.e., those who changed therapy vs those who were right-censored) were similar at baseline, and the DI over the period of 0 to 6 months was not statistically significant. In addition, the therapeutic response curves (i.e., the mean DI over time) appear similar. Their Figure 5 addresses the potential concern that the shape of the therapeutic response curve is related to the duration of the therapeutic segment. Because the segment may vary for each patient, this question is crucial, and Figure 5 seems to suggest one has no effect on the other. However, the curves must be interpreted with caution. The curves are not mutually exclusive in that each patient contributes to every curve of the same length or shorter than that patient's segment. The curves might look quite different if groups of patients with similar segment durations were plotted separately. In summary, the data presented by Ortendahl, et al regarding the characteristics of the pure MTX treatment segment add considerably to our knowledge of this agent. The concept of the therapeutic segment approach is a methodological advance of considerable merit, but its general usefulness in the study of other agents may be limited due to the lack of adequate patient cohorts followed carefully through time using standardized measures at predetermined intervals. Despite the statistical uncertainties mentioned above, it is likely that the overall conclusions and recommendations of the authors are valid. EFFECT OF HCQ ON MTX BIOAVAILABILITY The second article, by Carmichael, et al2, addresses the apparent increased bioavailability of MTX when prescribed concomitantly with hydroxychloroquine. The concurrent use of these 2 agents is common in adult rheumatology practices, but uncommon in pediatric rheumatology5,9. In brief, the mean area under the time-concentration curve for MTX was increased by a significant amount, as was the time to reach the maximum concentration (tmax), and a corresponding decrease in the maximum concentration (Cmax). Hydroxychloroquine's pharmacokinetic profile appeared unaffected. The authors state that these findings may account for the apparent increased potency of MTX when given concurrently with hydroxychloroquine and the lesser degree of flare of disease after MTX discontinuation. Two safety issues are raised. The decrease in Cmax may account for the decreased acute liver toxicity produced by MTX when given with hydroxychloroquine. However, since MTX is excreted through the kidneys, physicians must be aware of the increased potential for renal abnormalities. We live in the age of combination therapies, many times introduced early in the disease course10, which in some situations institute "off label" approaches (particularly in pediatric rheumatology) in which there is little or no pharmacokinetic data about the agent/s in the specific disease population. This well done study should serve as yet another warning to rheumatologists of the potential for toxic drug-drug interactions that may arise unexpectedly even when using drugs that have been on the market for years. LONGITUDINAL MEASUREMENT OF MTX LIVER CONCENTRATIONS In the third article, Fathi, et al3 present results of a 3.5 year double blind study of whether MTX and MTX polyglutamate accumulation in the liver correlates with clinical efficacy or laboratory toxicity. In addition, a new classification system (the Iowa Score) for grading abnormalities is investigated for its construct validity. The first 18 weeks were by double blind, placebo controlled, dose ranging study. After 18 weeks patients who had received placebo were blindly re-randomized to either 5 or 10 mg/m2 oral MTX once weekly. Maximum dose of MTX was 35 mg per week. Liver biopsies were done at baseline and after 1, 2, and 3.5 years after baseline. Histological abnormalities did not progress using the Roenigk score. The authors conclude that there is no correlation between MTX or MTX polyglutamate accumulation in the liver and patient demographics, liver histology, concomitant medications (NSAID and prednisone), and disease activity. Thus, MTX serum concentrations are likely to be of little value in predicting clinically important hepatotoxicity. The uniqueness and clinical relevance of this study comes from its determination of serial liver biopsies (including baseline biopsies) assessed with 2 different scoring systems, and its attempt to correlate MTX and MTX polyglutamate liver concentrations with a range of demographic, disease, and laboratory variables. Although the authors make a solid case for their conclusions based upon the data presented, use of GEE on this longitudinal dataset may reveal additional information and associations not apparent by univariate or multiple logistic regression. Nevertheless, it appears that even after years of widespread use of MTX, rheumatologists still have few tools to guide them in the monitoring of MTX and liver toxicity. Guidelines exist for performing liver biopsies based largely on liver function tests11. However, it is possible that clinical scenarios exist, such as specific clinical and laboratory profiles incorporating variables other than ALT and AST, that portend potential hepatic problems. Additionally, clinicians have very little guidance for what to do when certain liver pathologies appear in terms of adjustment to the therapeutic regimen that would still provide some chance of maintaining control of the arthritis. Further work on the development of clinical guidelines appears to be sorely needed in this area. The Iowa Score, in which various abnormalities are weighted for their importance, deserves further investigation. 1. Ortendahl M, Holmes T, Schettler JD, Fries JF. The methotrexate therapeutic response in rheumatoid arthritis. J Rheumatol 2002;29:2084-91. 2. Carmichael SJ, Beal J, Day RO, Tett SE. Combination therapy with methotrexate and hydroxychloroquine for rheumatoid arthritis increases exposure to methotrexate. J Rheumatol 2002;29:2077-83. 3. Fathi NH, Mitros F, Hoffman J, et al. Longitudinal measurement of methotrexate liver concentrations does not correlate with liver damage, clinical efficacy or toxicity during a 3.5 year, double-blind study in RA. J Rheumatol 2002;29:2092-8. 4. Hawley DJ, Wolfe F. Sensitivity to change of the health assessment questionnaire (HAQ) and other clinical and health status measures in rheumatoid arthritis: results of short-term clinical trials and observational studies versus long-term observational studies. Arthritis Care Res 1992;5:130-6. [MEDLINE] 5. Brunner HI, Kim KN, Ballinger SH, et al. Current medication choices in juvenile rheumatoid arthritis II - Update of a survey performed in 1993. J Clin Rheumatol 2001;7:295-300. 6. Wolfe F, Pincus T, O'Dell J. Evaluation and documentation of rheumatoid arthritis disease status in the clinic: which variables best predict change in therapy. J Rheumatol 2001;28:1712-7. [MEDLINE] 7. Weinblatt ME, Maier AL, Fraser PA, Coblyn JS. Longterm prospective study of methotrexate in rheumatoid arthritis: conclusion after 132 months of therapy. J Rheumatol 1998; 25:238-42. [MEDLINE] 8. O'Dell JR. Methotrexate use in rheumatoid arthritis. Rheum Dis Clin North Am 1997;23:779-96. [MEDLINE] 9. O'Dell JR. Triple therapy with methotrexate, sulfasalazine, and hydroxychloroquine in patients with rheumatoid arthritis. Rheum Dis Clin North Am 1998;24:465-77. [MEDLINE] 10. Wolfe F, Rehman Q, Lane NE, Kremer J. Starting a disease modifying antirheumatic drug or a biologic agent in rheumatoid arthritis: standards of practice for RA treatment. J Rheumatol 2001;28:1704-11. [MEDLINE] 11. Kremer JM, Alarcon GS, Lightfoot RW Jr, et al. Methotrexate for rheumatoid arthritis. Suggested guidelines for monitoring liver toxicity. American College of Rheumatology. Arthritis Rheum 1994;37:316-28. [MEDLINE] |