Cyclophosphamide-Induced Cystitis and Bladder Cancer in Patients with Wegener Granulomatosis

Methods

Patients

From 1967 to 1993, 145 patients with Wegener granulomatosis were treated with cyclophosphamide at the Warren G. Magnuson Clinical Center of the National Institutes of Health (NIH). The clinical features of the underlying disease in all but 3 of these patients have been reported previously [3]; the clinical and demographic characteristics of these patients are summarized in Table 1. Individual patients were followed for 0.5 to 27 years (median, 8.5 years), for a total of 1333 patient-years.

Treatment Protocol

We used the standard cyclophosphamide treatment regimen, which has been described previously [1, 3]. Therapy consisted of 1) oral cyclophosphamide, 2 mg/kg of body weight per day and 2) prednisone, 1 mg/kg of body weight per day. If patients improved substantially after the first month of treatment, the prednisone dose was gradually tapered to an alternate-day regimen, and prednisone therapy was eventually discontinued. Cyclophosphamide therapy was continued for at least 1 year after patients achieved complete remission. The cyclophosphamide dose was then tapered by 25-mg decrements of therapy every 2 to 3 months until discontinuation of therapy or until disease recurrence required an increase in dose. The cyclophosphamide dose was adjusted as needed to keep the absolute neutrophil count higher than 1.0 × 10⁶/L. If substantial toxicity required the permanent discontinuation of cyclophosphamide therapy and if signs of active vasculitis were present, azathioprine, chlorambucil, or (after 1990) low-dose weekly methotrexate therapy was allowed.

Eight patients with fulminant disease initially received intravenous cyclophosphamide at daily doses of 3 to 5 mg/kg. When their disease stabilized, these patients were switched to the standard oral cyclophosphamide regimen. Fourteen patients received monthly pulses of intravenous cyclophosphamide, 1 g/m² body surface area; 13 of these 14 patients also received one or more courses of the standard oral cyclophosphamide regimen. Thus, 144 of the 145 patients received oral cyclophosphamide therapy for some period of time.

Patients were evaluated at the NIH every 1 to 3 months. Those who achieved remission of disease and maintained it for 1 year were subsequently seen every 6 months. For each patient, urinalysis was done at every visit, and a cytologic examination of urine was done every 6 to 12 months.

All patients received cyclophosphamide as part of clinical research protocols approved by the National Institute of Allergy and Infectious Diseases (NI-AID) Institutional Review Board, the NIAID Clinical Director, and the director of the NIH Clinical Center. All patients gave written informed consent.

Urine Cytology

Cytologic examination of urine was done at each evaluation. Sediments obtained from voided urine specimens were fixed in Saccomanno solution (Lerner Laboratories, Pittsburgh, Pennsylvania), immobilized on membrane filters (Millipore, Chicago, Illinois) or by cytospin, placed in 95% ethanol, and stained with Papanicolaou stain. Urine samples obtained as much as 6 months before each cystoscopic examination were reviewed retrospectively and were correlated with subsequent bladder biopsy specimens. Cellular cytologic features were placed in the following categories according to the following criteria [8-10]:

1. Negative: no important epithelial abnormalities.

2. Atypia: some nuclear abnormalities in epithelial cells, but the changes could not be definitely placed in categories 3, 4, or 5.

3. Therapeutic or viral: cytologic changes consistent with polyomavirus infection or cyclophosphamide toxicity. In the absence of diagnostic inclusions, the two types of changes are indistinguishable, and we therefore grouped them together. The nuclear enlargement and hyperchromasia associated with polyomavirus or chemotherapeutic effect should be distinguished from high-grade dysplasia or carcinoma.

4. Dysplasia or possible low-grade transitional-cell carcinoma: a few cells in a voided urine sample, either singly or in clusters, that have slightly enlarged, irregular nuclei with increased granularity in chromatin distribution and small or absent nucleoli.

5. Transitional-cell carcinoma: high-grade lesions meeting unequivocal criteria of malignancy.

Definition of Terms

Nonglomerular hematuria was defined as microscopic or gross hematuria not associated with the presence of erythrocyte casts or declining renal function. Glomerulonephritis causing hematuria associated with erythrocyte casts (glomerular hematuria) occurred at least once in 116 of the 145 patients (Table 1). If hematuria persisted after the treatment of glomerulonephritis and the disappearance of erythrocyte casts, or if hematuria not associated with the presence of erythrocyte casts ever developed, patients were considered to have nonglomerular hematuria and had cystoscopy (see below).

Cyclophosphamide-induced cystitis was defined as nonglomerular hematuria associated with characteristic cystoscopic bladder changes. These changes included patchy areas of neovascularity and telangiectasia manifested as an increased number of tortuous, thin-walled veins and small areas of hemorrhage in or under the bladder epithelium. The mucosa between the hypervascular areas may appear normal or pale with decreased vascularity.

Cystoscopy

Cystoscopy was done to evaluate nonglomerular hematuria (microscopic or gross) in patients receiving cyclophosphamide. Only the results of cystoscopies done at the NIH Clinical Center by members of the Urologic Oncology Section of the National Cancer Institute are included in this report. Most of these cystoscopies were done by two of the authors; all cystoscopy results were reviewed by these two authors. Random biopsies were done only if the results of cytologic examination of urine suggested malignancy. Patients having cystoscopy had intravenous pyelography or retrograde pyelography, or both, at least once to evaluate the upper urinary tract.

Statistical Analysis

The frequencies of clinical findings were compared by using the Fisher exact test for association with bladder cancer and nonglomerular hematuria; adjustments were made for multiple comparisons of microscopic and gross hematuria with bladder cancer using the modified Bonferroni method [11]. The effects of fixed covariates (sex, history of smoking, age at onset of Wegener granulomatosis disease, and duration of disease before first cyclophosphamide treatment) and time-varying covariates (microscopic hematuria, gross hematuria, total cyclophosphamide dose, and duration of cyclophosphamide therapy) on the development of bladder cancer were examined using Cox proportional-hazards regression analysis [12, 13]. Medians and other percentiles for variables dependent on follow-up time were estimated by using the Kaplan-Meier method [14]. The cumulative distributions determined by the Kaplan-Meier method were compared with the log-rank test. Risk estimates for the development of bladder cancer in patients with Wegener granulomatosis who were treated with cyclophosphamide were determined by comparing observed rates with the expected rates for the United States population, which were obtained from the Surveillance, Epidemiology, and End Results (SEER) Cancer Statistics Review, 1973-1991 [15].

Results

Hematuria

Seventy-three of 145 patients treated with cyclophosphamide (50%) developed nonglomerular hematuria; the median time to development for all patients was 37 months of receipt of cyclophosphamide (95% CI, 32 to 55 months), and the median dose before development was 124 g (CI, 82 to 149 g) Table 2, Figure 1. Forty-one patients (56%) presented with microscopic hematuria; 32 (44%) presented with gross hematuria. Twenty-eight patients had more than one recurrent episode of nonglomerular hematuria (microscopic or gross), sometimes years after cyclophosphamide therapy had been discontinued. Eighteen of the 73 patients developed hematuria after cyclophosphamide therapy was discontinued. Manifestations of active vasculitis necessitated that cyclophosphamide therapy be continued in 26 of the 55 patients who developed nonglomerular hematuria while receiving this therapy.

Three patients (2.1%) had gross hematuria with clinically significant blood loss. Two men and one woman were transfused with 35 to 80 units of blood and required intravesical treatment with formalin or silver nitrate to stop the bleeding. Bleeding was attributed to cyclophosphamide-induced cystitis in two patients and was related to tumor (of colonic origin) invasion of the bladder in one patient. The total doses and durations of cyclophosphamide therapy for these three patients were 55 g administered over 12 months, 128 g administered over 43 months, and 163.7 g administered over 36.5 months.

Sixty of the 73 patients with nonglomerular hematuria (82%) had a total of 139 cystoscopies done at the NIH (mean, 2.3 cystoscopies; range, 1 to 13 cystoscopies). Cystoscopy was done with similar frequency in patients with and those without glomerulonephritis. Eleven of 29 patients (38%) who had no episodes of glomerulonephritis during the study period had a total of 20 cystoscopies; 49 of 116 patients (42%) who had at least one episode of glomerulonephritis had a total of 119 cystoscopies. Forty-two of the 60 patients (70%) having cystoscopy each had macroscopic changes consistent with cyclophosphamide-induced bladder injury on at least one occasion. Three patients were given a diagnosis of bladder cancer on the basis of their initial cystoscopies, which also showed changes consistent with cyclophosphamide-induced bladder injury. In the remaining 15 patients, the bladder mucosa appeared normal or showed nonspecific abnormalities. The cystoscopic manifestations of cystitis often changed over time in individual patients. These abnormalities resolved in some patients; in others, bladder lesions remained stable or progressed. No patient had evidence of Wegener granulomatosis involving the bladder.

Figure 1 shows the Kaplan-Meier estimates of the cumulative percentage of patients developing nonglomerular hematuria with cyclophosphamide therapy. Although men developed hematuria more often than women, a sex difference over time was not significant after adjustment for the total dose and duration of cyclophosphamide therapy. Although the median ages of men (39 years) and women (43 years) at the time of initial cyclophosphamide treatment were not statistically different, men tended to develop hematuria earlier than women (median for men, 56 years [CI, 48 to 62 years]; median for women, 65 years [CI, 55 to 73 years]). The effect of smoking on the development of hematuria is shown in (Figure 2). The proportions of smokers and nonsmokers who developed hematuria (47% and 57%, respectively) were not statistically different, but smokers developed hematuria after fewer months of cyclophosphamide treatment than did nonsmokers. The median time to the development of hematuria was 34 months for smokers compared with 56 months for nonsmokers (P = 0.03). Exact data are not available, but, in our experience, most patients with microscopic nonglomerular hematuria have no symptoms related to the urinary tract.

Transitional-Cell Carcinoma of the Bladder

Seven patients (5%) treated with oral cyclophosphamide developed transitional-cell carcinoma of the urinary bladder at a median age of 53 years (range, 49 to 83 years). The time receiving cyclophosphamide and the cumulative dose at which 5% of the sample developed bladder cancer were 46 months and 113 g, respectively. In 6 of the 7 patients who developed bladder cancer, the total cumulative cyclophosphamide dose exceeded 100 g, and the cumulative duration of cyclophosphamide treatment exceeded 2.7 years (Table 3). Before bladder cancer was diagnosed, all 7 patients had had one or more episodes of microscopic or gross nonglomerular hematuria Table 2 and Table 3. In contrast, none of the 72 patients who did not have a documented episode of nonglomerular hematuria developed bladder cancer.

At the time of diagnosis, three patients had noninvasive papillary lesions, one had invasion into the lamina propria, one had invasion into the deep muscle, one had carcinoma in situ, and one had a noninvasive papillary lesion and carcinoma in situ. Five patients had transurethral resection of their bladder tumors. Five patients had bladder instillation therapy with thiotepa, mitomycin, or Mycobacterium bovis bacille Calmette–Guérin vaccine. Two patients ultimately required cystectomy because of invasive disease; a third patient with invasive bladder cancer did not have a cystectomy because of severe underlying medical problems (Table 3).

The latency period, defined as the time from the initiation of cyclophosphamide therapy to the diagnosis of bladder cancer, ranged from 7 months to 15.3 years (Table 3). The lag time, defined as the time from the date of the last cyclophosphamide dose to the date of the diagnosis of bladder cancer, ranged from 0 to 10 years. The patient who developed bladder cancer after only 7 months of therapy was receiving cyclophosphamide for active Wegener granulomatosis at the time of diagnosis. The patient who developed bladder cancer 15 years after his first dose of cyclophosphamide had been in remission and had not received cyclophosphamide for 10 years. Figure 3 shows the Kaplan-Meier estimate of the cumulative risk for developing bladder cancer as a function of time from the first cyclophosphamide dose.

Six of the seven patients with bladder cancer had a history of smoking. None had a history of exposure to aromatic amines, including aniline dyes. No patient had bladder symptoms or cystoscopic findings compatible with decreased bladder capacity. Other clinical conditions that may enhance cyclophosphamide-induced bladder toxicity, such as obstructive uropathy and neurogenic bladder, were not seen in our cohort. No neoplasms of the ureter or renal pelvis were seen in our patient sample.

The rate of occurrence of bladder cancer among the 145 patients receiving cyclophosphamide who had Wegener granulomatosis was compared with that in the SEER Cancer Statistics Review (1973-1991) for adult men and women in the general U.S. population [15]. A 31-fold increase in the incidence of bladder cancer (CI, 13-fold to 65-fold) was seen in our patients on the basis of the SEER rate adjusted for all ages; those younger than 65 years of age had a 51-fold increase (CI, 14-fold to 132-fold; P < 0.001).

Predictors of the Development of Bladder Cancer

We analyzed the fixed covariates (sex, history of smoking, age at onset of Wegener granulomatosis, and duration of disease before first cyclophosphamide treatment) and the time-varying covariates (microscopic hematuria, gross hematuria, total cyclophosphamide dose, and duration of cyclophosphamide therapy) using the Cox proportional-hazards regression model to determine whether any of these factors were significant predictors of the development of bladder cancer. Of these variables, only microscopic nonglomerular hematuria was found to be statistically significantly associated with the development of bladder cancer (P < 0.01). Bladder cancer developed in 6 of 51 patients (12%) in whom microscopic nonglomerular hematuria was documented at some point during follow-up and in 1 of 94 patients without microscopic nonglomerular hematuria (P = 0.02).

A history of smoking tobacco was not found to be an independent risk factor for the development of bladder cancer. However, more patients with a history of smoking developed bladder cancer at a lower total dose of cyclophosphamide (P = 0.03) and after fewer months of cyclophosphamide therapy (P ≤ 0.01).

Urine Cytology

Five of the seven patients with bladder cancer had abnormal results on cytologic examination of urine before cancer was diagnosed by cystoscopy (Table 3). Cellular atypia was seen in two of these patients, and dysplasia suspicious for transitional-cell carcinoma was seen in three patients. In two of the seven patients, cytologic results did not suggest malignancy. Both of these patients had low-grade malignant lesions (Table 3). Patient 1, in whom bladder cancer was diagnosed after only 7 months of cyclophosphamide therapy, had had normal results on cytologic examination of urine 2 days before his bladder cancer was diagnosed by cystoscopy. The cytologic changes in urine seen in patient 3 were thought to be due to therapeutic or viral effects (Table 3).

The results of cytologic examination of urine did not consistently correlate with the degree of bladder injury seen at cystoscopy in patients with cyclophosphamide-induced cystitis. All 42 patients with evidence of cyclophosphamide-induced bladder changes had at least one cytologic examination of urine done within 2 months of one or more cystoscopic examinations. No epithelial-cell abnormalities were seen in 26 patients; atypia was seen in 5 patients; and therapeutic changes were seen in 15 patients. In the 15 patients with nonglomerular hematuria and no cystoscopic evidence of bladder injury, cytologic examination of urine showed no epithelial cell abnormalities in 10 patients, therapeutic effect in 1 patient, and atypia in 1 patient. In 3 patients, cytologic examination of urine was not done within 2 months of cystoscopy. For the detection of bladder cancer in the 60 patients with nonglomerular hematuria who had cystoscopy, the finding of dysplasia on cytologic examination of urine had a sensitivity of 43% and a specificity of 100%; the finding of atypia had a sensitivity of 29% and a specificity of 89%.

A total of 34 bladder biopsy specimens were obtained from 25 patients. Biopsy specimens from 9 patients were interpreted as showing acute and chronic hemorrhagic cystitis or chronic inflammation. Epithelial atypia was seen in 7 patients, epithelial dysplasia was seen in 1 patient, and no abnormalities were seen in 1 patient. Transitional-cell carcinoma of the bladder was reported in one or more biopsy specimens from 7 patients, of whom 1 had no cystoscopically identifiable lesion.

Discussion

Cyclophosphamide is highly effective for the treatment of Wegener granulomatosis and other serious autoimmune diseases, but its use can be associated with substantial toxicity of the urinary tract. The drug is metabolized in the liver to various chlormethine metabolites and acrolein [16]. The chlormethine metabolites are responsible for the therapeutic effects; the toxic effects on the urinary tract are thought to be mediated by the acrolein metabolite [17]. Although the entire urinary collecting system is at risk for acrolein-mediated toxicity, the bladder is most susceptible because of its prolonged exposure to the drug. Hemorrhagic cystitis and the induction of bladder cancer are well-recognized complications of oral cyclophosphamide therapy in patients with Wegener granulomatosis, systemic lupus erythematosus, and rheumatoid arthritis [3, 5-7]. However, the incidence of cyclophosphamide-induced cystitis and its relation to the development of bladder cancer have not been defined.

In our study, nonglomerular hematuria developed in 50% of patients treated with oral cyclophosphamide. It occurred earlier in patients who smoked tobacco and was chronic and recurrent in 38% of patients, even after the cessation of cyclophosphamide therapy. Seventy percent of patients with nonglomerular hematuria who had cystoscopy had macroscopic changes of cyclophosphamide-induced cystitis. Thus, nonglomerular hematuria usually indicates the presence of cyclophosphamide-induced bladder injury.

Bladder cancer developed in 5% of our patients treated with cyclophosphamide; compared with the general population, this represents a 31-fold increase in the incidence of bladder cancer and a 51-fold increase for persons younger than 65 years of age. By Kaplan-Meier analysis Figure 3, the estimated incidences of bladder cancer after first exposure to cyclophosphamide are 2% at 5 years, 5% at 10 years, and 16% at 15 years. The rate of bladder cancer in our study is higher than that previously reported for patients with Wegener granulomatosis who are treated with oral cyclophosphamide [1, 3, 5, 18], probably because follow-up was longer in our study. It should be emphasized that this is a minimum estimate of the occurrence of bladder cancer in patients receiving cyclophosphamide. Given the long latency period between the initiation of cyclophosphamide therapy and the diagnosis of bladder cancer Table 3, the cumulative incidence of bladder cancer in this patient cohort may continue to increase with time.

We examined several clinical variables for their association with the development of bladder cancer. Neither the total dose of cyclophosphamide nor the cumulative time receiving cyclophosphamide therapy were found to be significant risk factors for the development of bladder cancer in our patients. However, bladder cancer developed in 6 of the 51 patients (12%) whose cumulative cyclophosphamide dose exceeded 100 g. Continued follow-up of these patients may show the cumulative dose of cyclophosphamide to be an independent risk factor. Although 6 of the 7 patients had a history of smoking tobacco Table 3, this history was not found to be a statistically significant risk factor for the development of bladder cancer. However, smokers developed both nonglomerular hematuria and bladder cancer after fewer months of cyclophosphamide treatment than did nonsmokers, suggesting that smoking may enhance the urotoxicity of cyclophosphamide and its metabolites.

Of the variables examined, only microscopic nonglomerular hematuria was found to be a significant risk factor for the development of bladder cancer. Bladder cancer developed in 12% of patients with and in only 1% of patients without microscopic nonglomerular hematuria (P = 0.02). Although six of the seven patients with bladder cancer had at least one documented instance of nonglomerular microscopic hematuria, these episodes were not always frequent and recurrent. In one patient, bladder cancer was diagnosed 5 years after the last documented episode of microscopic hematuria (patient 2; Table 3). In combination with the cystoscopy results described above, these data strongly suggest that nonglomerular hematuria is a sensitive marker for cyclophosphamide-induced bladder injury and that it identifies a subgroup of patients at high risk for developing bladder cancer.

We found cytologic examination of urine to be of limited usefulness in monitoring cyclophosphamide treated patients for the development of bladder cancer. In three of the seven patients with bladder cancer, such examination showed changes suggestive of transitional-cell carcinoma. Atypia was seen in two patients and therapeutic or viral changes were seen in one patient; cytologic examination of urine in one patient showed no epithelial cell abnormalities (Table 3). These results, along with the findings of other investigators [5], indicate that cytologic examination of voided urine is a relatively insensitive method for detecting low-grade bladder cancer in patients treated with cyclophosphamide. However, the sensitivity of cytologic examination of urine increases for higher-grade malignant lesions.

The most sensitive method for detecting cyclophosphamide-induced cystitis or early malignant lesions is a routine urinalysis with careful examination for nonglomerular microscopic hematuria. Because nonglomerular hematuria may be the first manifestation of bladder cancer or, more likely, of cyclophosphamide-induced cystitis, we strongly recommend that all patients treated with cyclophosphamide have urinalysis every 3 to 6 months, even after cyclophosphamide therapy is discontinued. Patients who develop nonglomerular hematuria (microscopic or gross) should be evaluated by cystoscopy. If cystitis is documented, cyclophosphamide therapy may be continued if medically necessary to treat potentially life-threatening manifestations of Wegener granulomatosis. However, these patients are probably at greater risk for additional cyclophosphamide-induced bladder toxicity, and other immunosuppressive agents, such as methotrexate, should be considered [19].

It is still unclear how best to monitor the subgroup of patients with cyclophosphamide-induced cystitis for the development of bladder cancer. Cytologic examination of urine is useful for detecting higher-grade bladder cancer and should be done every 6 to 12 months. The finding of atypia or dysplasia on such examination should lead to a cystoscopic evaluation. However, cytologic examination of urine is relatively insensitive for detecting lower-grade malignant lesions, and this limits its usefulness in monitoring these patients. Thus, routine cystoscopy every 1 or 2 years should be considered for all patients with cyclophosphamide-induced hematuria. This approach would be similar to that taken in patients with ulcerative colitis, in whom the risk for developing carcinoma of the colon is estimated to be 7% to 14% after 20 to 25 years of disease [20-22]. For these patients, annual or biennial colonoscopic surveillance with random biopsies is advocated [23, 24].

In summary, 5% of our patients treated with cyclophosphamide developed transitional-cell carcinoma of the bladder. We found that nonglomerular microscopic hematuria was a frequent manifestation of cyclophosphamide-induced cystitis and that it identified a subgroup of patients at high risk for developing bladder cancer. The best way to monitor this subgroup of patients is unclear, but routine cystoscopy every 1 or 2 years should be considered. Because the risk for bladder cancer persists for many years after cyclophosphamide therapy has been discontinued, lifelong surveillance for bladder cancer may be necessary.

Acknowledgments: The authors thank the clinical fellows and physician assistants at the National Institute of Allergy and Infectious Diseases and the Clinical Center nursing staff for the care of our patients; and Drs. David Alling, Steven Banks, and Mary Foulkes for their assistance with the statistical analysis.

Dr. Hijazi: National Cancer Institute, Building 10, Room 2A 10, 9000 Rockville Pike, Bethesda, MD 20892.

Drs. Walther and Linehan: National Cancer Institute, Building 10, Room 2B 47, 9000 Rockville Pike, Bethesda, MD 20892.

Dr. Lubensky: National Cancer Institute, Building 10, Room 2N 212, 9000 Rockville Pike, Bethesda, MD 20892.

Dr. Kerr: Veterans Affairs Medical Center, 50 Irving Street NW, Washington, DC 20422.

Dr. Hoffman: Cleveland Clinic Foundation, 9500 Euclid Avenue, Cleveland, OH 44195.