1706 LETTERS

apy. We would like to ask for clarification of a few points about the study’s design and outcome.

First, the authors provide very few details about the randomization procedure. They state only that the patients were randomly assigned to one of the CYC regimens. They mention that, in order to avoid referral bias from rccruiting exclusively at tertiary care centers, the study patients were recruited from a large panel of French hospitals, most of which were general hospitals. Therefore, it is important to know how and where the randomization of patients took place: locally (at each hospital), where the physicians might have been biased by knowledge of patients’ clinical status, or by blinded investiga- tors, perhaps at a central site? An unbiased randomization is critical to ensuring a fair comparison, yet not enough details are given to provide such reassurance.

Second, we note that randomization did not occur until several days after the patients had received initial treatment. All of the patients initially received a single pulse of intrave- nous CYC as well as high doses of corticosteroids for several days before randomization. Why were patients not randomized to either group A (intravenous CYC) or group B (oral CYC) at the time of study enrollment? Failure to randomize patients to one of the treatment arms at the time ol enrollment may have resulted in either decreased ability to extrapolate the study’s results to all patients with generalizcd WG or-even worse-the introduction of bias into the study. For example, it is conceivable that physicians at some of the participating hospitals might have deemed certain patients “too sick” to be enrolled in the study, who possibly did not receive the “stan- dard of care” treatment (oral CYC). As a result, some of the sickest patients eligible for the study might not have been enrolled. If so, caution should be used whcn attcmpting to generalize the study’s conclusions to WG patients who are very ill.

Furthermore, failure to assign patients to a treatment group at the time of their enrollment might have resulted in the preferential enrollment of patients with poor initial treat- ment responses into the oral daily CYC arm, i.e., resulting in the introduction of bias. In Table 1 of the article by Guillevin et al, the baseline clinical manifestations and laboratory find- ings in patients in the 2 groups appear comparable. However, no information is provided on ,some variables likely to be associated with poor outcomes, e.g., the 2 groups’ mean baseline creatinine levels or the distribution of patients with advanced renal insufficiency (2). A predominance of patients with these variables in group B might explain the slightly (although not statistically significant) inferior final outcomes among the patients who received CYC orally.

Third, given the difficulty of defining disease activity in patients with systemic vasculitis, all therapeutic trials in these disorders are potentially troubled by imprecise definitions of outcome measures. In this regard, we find it difficult to distinguish between the investigators’ definitions of complete remission and partial remission. The former is defined as improvement in the patient’s gcneral condition, absence of new manifestations of WG, a normal erythrocyte sedimenta- tion rate, and improvement (partial or total) of renal function and other parameters. Absence of ongoing treatment was apparently not a criterion for complete rcmission. Similarly, partial remission was defined as stability or attenuation of clinical and radiologic symptoms and regression or disappear-

ance of laboratory abnormalities under constant treatment. On paper, at least, distinctions between these 2 end points appear to be quite subjective. For this reason, the conclusion that the final outcomes for patients in groups A and B were equivalent (66.7% and 56.5% in remission, respectively; P not significant) must be interpreted with caution, particularly since blinding the patients and the evaluators to thc treatment assignmcnts would have been very difficult (and was not done).

Fourth, the authors note that they added trimethoprimisulfamethoxazole (TIS) to the treatment rcgi- men after detecting a high frequency of Pneumocystis carinii pneumonia in the first 12 patients. Was the T/S effective in preventing this complication in the remaining 38 patients?

The investigators’ efforts in conducting the first-ever prospective, randomized treatment trial in WG are laudable. Our comments are not intended to diminish the potential importance of this study, but rather to help understand its proper interpretation. The study’s conclusions, as discussed in the accompanying editorial (3) , offer the suggestion that the standard initial therapy for patients with generalized WG might reasonably be altered. The study’s confirmation of oral CYC‘s superiority for preventing relapses, juxtaposed against the alarming rate of treatment-related side effects experienced by patients in the oral CYC arm (e.g., the 34.7% incidence of opportunistic infections), demands a complete understanding of how this study was performed.

John H. Stone, MD, MPH David €3. Hellmann, MD Johns Hopkins Universily Baltimore, M D

Guillevin L, Cordier J-F, Lhote F, Cohen P, Jarrousse B, Royer I, et al. A prospective, multicenter, randomizcd trial comparing stcroids and pulse cyclophosphamidc vcrsus steroids and oral cyclophosphamide in the treatment of generalized Wegener’s gran- ulomatosis. Arthritis Rheum 1997;40:2187-98. Falk RJ, Hogan S, Carey TS, Jennette JC, and the Glomerular Disease Collaborative Network. Clinical course of anti-neutrophil cytoplasmic autoantibody-associated glomerulonephritis and sys- temic vasculitis. Ann Intern Med 199011 13556-63. Hoffman GS. Treatment of Wegener’s granulomatosis: timc to change the standard of care [editorial]? Arthritis Rheum 1997;40: 2099-1 04.

Pulse versus oral cyclophosphamide in the treatment of Wegener’s granulomatosis: comment on the article by Guillevin et a1

To the Editor: We read with interest the report by Guillevin et a1 on

the use of intermittent (pulse) CYC in the treatment of WG (1). The relative efficacy of pulse versus daily oral CYC in this disease has been a subject of intense controversy. While the efforts of Guillevin and colleagues to examine this in a randomized, prospective manner are to be applauded, our interpretation of their results leads us to a very different conclusion than that reached by the authors.

Consistent with the findings of previous studies (2,3), Guillevin et al have demonstrated that pulse CYC treatment does not lead to sustained remission and is associated with a

LETTERS 1707

significantly higher rate of disease relapse. In the abstract of their article they also directly state that their ”results indicatc that pulsc CYC is as effective as oral CYC in achieving initial remission of WG and is associated with fewer side effects and lower mortality” (1). These arguments are subsequently used in the Discussion section as the basis for favoring the initial choice of pulse CY C.

With regard to cfficacy, a major shortcoming of this study is that the number of patients in cach treatment arm is too small to enable detection of a significant difference in remission rate if in fact one exists (Type I1 error). By stating that the 2 treatments are “as effective,” the authors imply that the study tested for and established treatment equivalency. Equivalence studies have become an increasingly important type of trial and are particularly relevant when one is hoping to establish that a potentially less toxic treatment is as effective as an cstablished standard. It is often underrecognized that such trials have important distinctions in their planning and inter- pretation and that large sample sizes are needed in order to demonstrate the similarity of 2 treatments with any degree of confidence (4-6).

Although different approaches havc been uscd to calculate the sample size in equivalence studies, a frequently cited method is that described by Makuch and Simon, which is based on the use of confidence intervals (4). In previous studies, daily oral CYC has been found to result in complete remission in -75% of patients (7). If one uses this as the probability of response in the formula of Makuch and Simon, 233 patients would be needed in each treatment arm to demonstrate with 95% confidence ( a = 0.05) that pulse CYC is at worst 10%) inferior to daily oral CYC. This calculation is based on a power of 80%, which indicates that even with 233 patients per arm, there would still be a 20% risk of reaching the wrong conclusion. Since the trial by Guillevin et a1 included only 27 and 23 patients in the 2 respective treatment arms, one certainly cannot conclude from the data that pulse CYC is “as effective” as daily oral CYC in achieving initial remission, and for the authors to do so is frankly erroneous.

With rcgard to toxicity, according to Table 4 of Guil- levin et al’s report, the only side effect that was significantly less frequent with pulse CYC therapy was the infection rate (40.7% versus 69.6%; P < 0.05). The most frequent infection in both groups was pneumonia due to PCP. As noted by the authors, the occurrence of PCP could be drastically reduced or eliminated by the use of appropriate prophylaxis. Tf the incidence of PCP was reduced or eliminated, the 2 regimens would have a similar rate of side effects. Therefore, perhaps a more appropriate conclusion of this study would be that daily oral CYC with PCP prophylaxis is the preferred regimen becausc it is effective, has an equal degree of toxicity, and is associated with a significantly lower rate of disease relapse than pulse CYC.

Although thc question of pulse versus daily oral CYC will undoubtedly remain a matter of debate, we would urge physicians to weigh the issues that influence the results of this study bcfore accepting its conclusion. This report adds to the existing information demonstrating significant difficulties with the use of pulsc CYC, and it remains our strong opinion that when CYC is indicated in patients with WG, it should be administcred as daily oral therapy.

Carol A. Langford, MD, MHS Michacl C. Sneller, MD National Institute of Allergy and Infectious Diseases National Institutes of Health Bethesda, MD

1. Guillcvin L, Cordier J-F, Lhote F, Cohen P, Jarrousse B, Royer I, et al. A prospective, multicenter, randomizcd trial comparing steroids and pulse cyclophosphamide versus steroids and Ordl cyclophosphamide in the treatment of generalized Wegener’s gran- ulornatosis. Arthritis Rheum 1997;40:2187-98.

2. Hoffman GS, Lcavitt RY, Fleisher TA, Minor JR, Fauci AS. Treatment of Wegencr’s granulornatosis with intermittent high- dose intravenous cyclophosphamide. Am 3 Med 1990;89:403-10.

3. Reinhold KE, Kekow J, Schnabel A, Schrnitt WH, Heller M, Beige1 A, et al. Influence of disease manifestation and antineutrophil cytoplasmic antibody titer on the rcsponse to pulse cyclo- phosphamide therapy in patients with Wegener’s granulornatosis. Arthritis Rheum 1994:37:919-24.

4. Makuch R, Simon R. Sample size requirements for evaluating a conservative therapy. Cancer Treat Rep 1978;62:1037-40.

5. Makuch RW, Johnson MF. Some issues in thc design and interpre- tation of “negative” clinical trials. Arch Intern Med 1986;146:986-9.

6. Blackwelder WC. ”Proving the null hypothesis” in clinical trials. Controlled Clin Trials 1982;3:345-53.

7. Hoffman GS, Kerr GS, Leavitt RY, Hallahan CW, Lebovics RS, Travis WD, et al. Wegener granulomatosis: an analysis of 158 patients. Ann Intern Med 1992;116:488-98.

Reply To the Editor:

The comments made and clarifications requested by different colleagues regarding our group’s prospective trial of WG treatment demonqtrate the interest of the medical com- munity in the therapeutic approach to this severe relapsing disease. Our trial was the first attempt to establish a coherent treatment strategy that integrates the initial and maintenance treatments of WG. Tt was not intended to solve all of the problems related to WG therapy, but focused on some major aspects of thc management of the disease. The need to devise a therapeutic approach other than oral CYC plus prednisone became evident from the fact that, with this combination, 50% of the patients had relapses and the occurrence of long-term complications, e.g., malignancies, was alarmingly high (1). The clarifications requested by Drs. Stone and Hellman can be made easily:

1. A randomization list was preestablished, and the treatment arm was assigned at study enrollment when the treating physician contacted the coordinating center (Avicenne Hospital, Bobigny). Inclusion criteria were vcrified by the study coordinator at the time of randomization. This proce- dure was preferred to local randomization because the disease is rare, and it seemed improbable that a sufficiently high number of patients could be enrolled in a single center to obtain equally balanced groups.

2. Randomization was done immediately at the time of diagnosis and before initiation of treatment. Although the initial therapeutic phase was the same for both groups, the treatment group was assigned before thc administration of CYC. For some patients, especially those with life-threatening manifestations, steroid pulses were given before randomiza-