Competing interests None.
Ethics approval This study was conducted with the approval of the local ethic committees.
Provenance and peer review Not commissioned; externally peer reviewed.
Statistics from Altmetric.com
Granulomatosis with polyangiitis (Wegener's) (GPA) (also known as Wegener's granulomatosis)1 and microscopic polyangiitis (MPA) are organ- and life-threatening diseases often grouped together under the general term anti-neutrophil cytoplasm antibody (ANCA)-associated small-vessel vasculitis (AAV).2 3 With the introduction of treatment regimens combining high-dose glucocorticoids with cyclophosphamide or various other cytotoxic agents, disease remission for the majority of patients is achieved and the life expectancy of patients surviving the first year is approaching that of the general population.4 Along with the better prognoses for AAV patients, potential long-term side effects of these therapies have progressively become a primary focus of interest. Notably, as has been reported for other inflammatory rheumatic5 6 and non-rheumatic7 8 diseases and organ transplant recipients,9 the de-novo development of cancers associated with immunosuppressive therapy is a major concern for AAV.
Previously published observations suggested that the overall incidence of cancer in treated AAV patients is 1.6–2.4 times higher than that of the general population.10,–,13 Moreover, cancer type-specific analyses demonstrated an increased risk of bladder cancer10,–,13 and leukaemia,10 12 with those findings spotlighting the known urothelial and haematological toxicities of cyclophosphamide.14 Higher lymphoma11,–,13 and non-melanoma skin cancer (NMSC)10 12 13 incidences were also inferred. However, the reported risk levels varied and those studies covered observation periods dating back to the 1970s and 1980s that may no longer reflect the cancer risk of currently applied regimens. In addition, the majority of data pertained to GPA10,–,12 and, to the best of our knowledge, no study has examined the comparative cancer risks of GPA and MPA.
In this study, we analysed follow-up data of patients with newly diagnosed AAV included in four recent multicentre clinical trials to assess these patients' subsequent cancer risk.
This analysis was based on patients with newly diagnosed GPA or MPA from 15 countries who, between March 1995 and September 2002, had been enrolled in four international multicentre, randomised clinical trials organised by the European Vasculitis Study Group (EUVAS). These trials included patients with ‘early systemic’,15 ‘generalised’,16 ‘generalised, renal’,17 or ‘severe renal’18 AAV. Notably, three of them evaluated the efficacy and safety of cyclophosphamide-sparing regimens and demonstrated, within their respective patient populations, that cyclophosphamide use could be confined to the remission–induction period,16 administered as periodic intravenous pulses,17 or replaced by methotrexate.15
Only participants in the early systemic AAV trial randomly assigned to receive methotrexate induction treatment did not receive cyclophosphamide induction therapy.15 Otherwise, patients achieving remission with cyclophosphamide induction were switched to azathioprine maintenance therapy 3–12 months after trial treatment onset. Trials were completed in 1215 18 or 1817 18 months. In addition, all patients received high-dose glucocorticoids; plasma exchange was allowed, if considered appropriate or according to randomisation.18 After completion or withdrawal from the trials, further treatment decisions were left to the discretion of the investigators at the primary sites. Patients with a previous cancer were ineligible for trial participation.15,–,18
Over the period 2004–7, a cross-sectional mailing survey of the trial physicians at the primary study sites was undertaken to gather information on the outcomes of the participants in the four clinical trials. Several reminder letters were sent to non-responders. The survey used a standardised questionnaire, composed of 22 questions, to collect follow-up data on survival, cause of death, renal outcome, relapses, cardiovascular and other adverse events, including cancers that had occurred during the initial study or follow-up. Information was also collected on the type and duration of treatments administered after the trial period; the amount of each drug given was not recorded, meaning cumulative doses could not be calculated. Overall survival data generated by the follow-up study were published previously.4
For each cancer notification, the questionnaire obtained information on the type of cancer and its site and date of diagnosis. To identify cancers occurring in study subjects for whom no follow-up data were obtained, we also reviewed the trial adverse event report forms for possible additional cancers. For subjects with multiple cancer notifications, both first and subsequent cancers were considered.
Calculation of cumulative cancer incidence rates
Cumulative cancer rates were computed using the Kaplan–Meier method. The observation time started at trial inclusion and ended, at the earliest, at the date of the first cancer diagnosis or that of last contact or death.
Calculation of standardised incidence ratios
For each patient, person-years of follow-up were calculated using the start and end dates described above, and for each of the considered cancer endpoints. The person-years at risk were calculated by sex and 5-year age groups, and separately for five geographical areas (see below). Standardised incidence ratios (SIR) were calculated for cancers at all sites, all sites excluding NMSC (non-NMSC), and each of the reported cancers at any site. Subgroup analyses concerned the following variables: sex, age at study entry (dichotomous according to median value), GPA versus MPA, trial, chronic kidney disease stage19 at study entry (≤2 vs >2), geographical area (see below) and disease activity assessed using the Birmingham vasculitis disease activity score (dichotomous according to median value).20 A further division was made by the time elapsed since the start of follow-up.
Because the study patients came from 14 different European countries (and five patients from Mexico) and in light of the known international variations of cancer incidence,21 the expected numbers of cancers were calculated by pooling the patients into five geographical areas and linking each area to existing cancer registries (figure 1). In an attempt to obtain the most accurate values, we chose national cancer registries from five large countries or areas, ie, the UK,22 the Nordic countries,23 Germany,24 France25 and Italy,26 and applied their rates to their respective countries but also their neighbouring countries and/or to countries located at the same latitude. Because NMSC are not recorded in the German and French national registries, and no general French population data on this cancer type are available, we also selected two regional German cancer registries from northern27 and southwestern regions,28 and assimilated data on NMSC incidences from those two registries into the German and French datasets, respectively. For all databases, we used year 2000 data except for the Italian registry for which 2000–3 data were used.
The expected numbers of cases of cancers and specific cancer types were calculated by multiplying the number of person-years for each sex, age and area group by the corresponding specific cancer incidence rates in the respective areas.29 SIR and their 95% CI were calculated assuming that the observed number of cases followed a Poisson distribution.30 31 In addition, subgroup findings were also analysed using log-linear Poisson regression models, with the logarithm of the expected numbers of cancers as an offset term, to generate RR for cancer incidence in each stratum compared with a reference stratum.
Continuous variables are expressed as means±SD and categorical variables are expressed as numbers (%). Comparisons of the baseline characteristics of patients diagnosed with cancer versus those who had no cancer used Student's t tests or, for small samples, Kruskal–Wallis tests and χ2 and, when appropriate, Fisher's exact tests. Survival rates were compared using the log-rank test. For all statistical analyses, a two-tailed p≤0.05 was considered significant and 95% CI were calculated.
Study sample characteristics
The four trials included a total of 535 patients. Follow-up data after trial completion were obtained by means of the mailing survey addressed to trial investigators for 467 (87.3%) patients; for the remaining 68 (12.7%), we used follow-up data accrued during the trial phase. During the overall 2650 person‑years' observation period (mean follow-up 4.95±3.22 years), 133 patients died.
Table 1 shows the main demographic, clinical and laboratory characteristics at diagnosis, treatments received and outcomes of the study population. Because information on therapies received by 42 patients was incomplete, the corresponding results were presented only for the 493 patients with detailed treatment data. Among the latter, 458 (92.9%) received cyclophosphamide and 331 (67.1%) received azathioprine at some time during their treatment.
Numbers and characteristics of observed cancers
Sixty-two cancers were notified in 53 patients. After review of the data provided, eight notifications were excluded because they were not considered definite cancers (two dysplasias, one Bowen's disease and five myelodysplasias), and one report was dismissed because it referred to a previous cervical carcinoma. The 53 reports of malignancies retained corresponded to 18 diagnoses of NMSC, ie, 13 basal and five squamous cell carcinomas, in 15 patients (including three patients with two distinct NMSC diagnoses) and 35 non-NMSC cancers (in 34 patients). Three patients were diagnosed with an NMSC and a non-NMSC and one patient was diagnosed with two non-NMSC. In total, 46 patients had at least one cancer diagnosed and 34 patients had at least one non-NMSC.
Among the four identified bladder cancers (three in GPA patients and one in an MPA patient), two were non-invasive (pTa according to the tumour–node–metastasis classification).32 All four patients had received cyclophosphamide for 6–36 months, and the trial inclusion to bladder cancer diagnosis intervals ranged between 2.1 and 6.6 years. One acute myeloid leukaemia and one chronic lymphocytic leukaemia were recorded in patients who had received cyclophosphamide for 7 or 3 months, respectively. All 15 patients with NMSC diagnoses had received cyclophosphamide, 13 (86.7%) of whom had also taken azathioprine.
The cumulative 5 and 8-year rates for cancers at all sites were 8.4% (95% CI 6.0 to 11.8) and 13.1% (95% CI 9.6 to 17.7), respectively, with respective non-NMSC rates of 7.6% (95% CI 5.3 to 10.9) and 9.9% (95% CI 7.1 to 13.9) (figure 2).
Among the 46 cancer patients at any site, 21 (45.7%) died and, among the 34 non-NMSC patients, 19 (55.9%) died; 16 (12.0%) of the total 133 deaths were attributed to cancer. Comparisons of the main characteristics of the AAV patients with and without cancer showed that the former were significantly older, had a significantly higher death rate and significantly longer follow-up (table 1).
Table 2 shows the calculated SIR for cancers at all sites, non-NMSC and for each of the identified 15 cancer sites. Because cancer type-specific incidence data for larynx and thyroid cancers were not available in the Italian cancer registry, and those for myeloma and liver cancer were not available in the German registry, we applied the corresponding incidence rates provided by the French registry to the person-years of these cancer sites and areas.
The SIR for all cancers indicated a 58% increased risk (SIR 1.58, p=0.003). The SIR for all non-NMSC sites was 1.30 (p=0.16). NMSC was the only site-specific cancer that was significantly increased (SIR 2.78, p=0.001).
Table 3 (and supplementary table S1, available online only) show subgroup analyses for incidences of all cancers and all non‑NMSC stratified by predefined variables. The results of those analyses indicated higher SIR for cancers at all sites in GPA than MPA patients (SIR: 1.92 vs 1.20) but their 95% CI overlapped; the corresponding RR was 1.60 (95% CI 0.90 to 2.86). Subgroup analyses of three follow-up periods (ie, ≤3, 3–5 and >5 years) did not indicate a clear cancer incidence trend over time.
Based on this analysis of follow-up information on 535 AAV patients initially treated within recent EUVAS clinical trials, we found a significantly higher SIR for cancers at all sites (1.58) and a non-significant SIR for non-NMSC (1.30) than general population expectations. NMSC was the only cancer site for which a statistically significant, 2.8-fold incidence increase was found. Despite our cancer SIR computed for overall cancer risk being at the lower range compared with previous reports that examined AAV patients10 12 13 (table 4), our data support evidence that cancer is an important cause of AAV-associated morbidity.
Over the past two decades, the known substantial toxicity of cyclophosphamide instigated many efforts to develop cyclophosphamide-sparing regimens for AAV treatment. Three of the four EUVAS trials, from which our study population was recruited, represented important advances in this respect and contributed to new standards of AAV care. Although the lower cyclophosphamide exposure achieved by these treatment regimens remains difficult to quantify, the lower SIR for cancers in any site, particularly for bladder cancer and leukaemia in our study population as compared with findings obtained earlier (table 4), could indicate that the more contained use of this drug might have started to show a benefit. Because only half of the patients enrolled in these studies were randomly assigned to receive the cyclophosphamide-sparing regimens, our study might still overestimate the cyclophosphamide-related cancer risk of current treatment protocols.
The accelerated NMSC rate observed concurs with previous studies on GPA10 12 and AAV,13 even though our risk estimate is lower than those reported earlier (table 4). Sun exposure is the main environmental cause of NMSC, and these observations support that AAV patients who receive or had received immunosuppressants should undergo regular skin cancer screening and be advised to protect themselves against ultraviolet radiation exposure. Some studies provided in-vitro and in-vivo evidence that azathioprine might hasten the induction of these tumours by sensitising the skin cell genome to ultraviolet A radiation.33 Herein, most of the NMSC diagnoses occurred in azathioprine-treated patients, but no inference can be derived concerning possible azathioprine responsibility for NMSC because the majority of analysed patients had been exposed to it at some time during their observation.
Our and previous data9 11 12 indicate that cancers in many sites contributed to the observed cancer excess (table 4). Although some of these findings may be fortuitous, they accord with the extensive range of cancers observed in other settings of iatrogenic or acquired immunodeficiency. Reports on organ transplant recipients and HIV-infected populations conclusively pointed out the heightened incidence of cancers attributable to infectious agents but also for malignancies for which no clear infectious or common societal factors were present.9 Those observations tend to underline the importance of a functional immune system in antimicrobial defence and the control of emerging cancer clones. That infectious cancers might occur more frequently in the AAV setting could be hypothesised based on our and previously observed elevated risks for NMSC and, albeit not statistically significant, for cancers of the larynx, vulva,13 oral cavity and pharynx,12 to which human papilloma virus is a possible (NMSC, larynx) or definite (oral cavity/pharynx, vulva) aetiological factor.34
An unresolved issue is whether AAV predisposes to lymphoma development. Authors of several studies reported a four to 11-fold increased risk of the incidence of lymphomas overall11 12 or non-Hodgkin's lymphoma13 in GPA, but this association was not replicated by another10 and this study (table 4). A heightened lymphoma risk was demonstrated for several other autoimmune diseases, eg, rheumatoid arthritis and systemic lupus erythematosus.36 37 Although the underlying biological mechanism is not fully understood, consensus is building that chronic immune stimulation plays a central role in autoimmune disease-related lymphomagenesis.38 It could be argued that, in AAV, the inflammatory response is not sufficiently intense for a prolonged period to induce lymphoproliferative diseases.
Stratification of cancer incidence according to the vasculitis diagnosis suggested higher cancer incidences among GPA patients than those with MPA, and this finding might have contributed to our observed lower overall cancer incidence, as opposed to that focused on GPA alone10,–,12 (table 4). Keeping in mind that the between-vasculitis comparison did not reach statistical significance, the higher SIR for GPA could be attributed to the more intensive immunosuppressive therapy given to these patients. Because cancer incidence and non-cancer mortality are competing events, the higher cancer risk for GPA could also be explained by the known superior survival of patients with GPA compared with MPA.4
A limitation of our study is that the analysed patients were initially enrolled in clinical trials, which could have led to either underestimating the cancer risk, because inclusion was incompatible with a history of cancer,15,–,18 or overestimating it because of increased or earlier cancer detection. It is also possible that it was too early to look for malignancies developing later during disease follow-up, particularly bladder cancer, which has reported latency periods of up to 15–19 years after initial cyclophosphamide treatment for AAV.10 12 35 A common caveat inherent to the design of our study pertains to the possibility of inaccurate or inadequate background data, which may hold true especially for NMSC that are possibly underreported,39 and the inconsistent recording of non-invasive bladder tumours in cancer registries. Finally, our dataset was not suitable for more in-depth examination of the relationship between the cumulative doses of the individual medications and cancer risk.
To conclude, our data strengthen the link between AAV therapy and cancer development. Although suggesting that the less extensive use of immunosuppressants in current treatment protocols might have resulted in a lower cancer risk, our study's results further highlight the persistent challenge of finding the optimal balance between AAV therapy risks and benefits. Longer follow-up data are warranted to appraise the risk of developing cancers later during the course of AAV.
The authors would like to acknowledge the contributions of the other EUVAS investigators involved in the ‘long-term follow-up study’: I Bajema, A Berden, S le Cessie (Leiden, The Netherlands); K de Groot (Offenbach, Germany); A Ekstrand (Helsinki, Finland); EC Hagen (Amersfoort, The Netherlands); K Herlyn (Bad Bramstedt, Germany); D Jayne (Cambridge, UK); R Luqmani (Oxford, UK); C Mukhtyar (Norwich, UK); I Neumann (Vienna, Austria); N Rasmussen (Copenhagen, Denmark); A Salmela (Helsinki, Finland); Daina Selga (Lund, Sweden); C Stegeman (Groningen, The Netherlands); and M Walsh (Hamilton, Canada). The authors are also very grateful to all the EUVAS clinical investigators at the primary study sites for their assistance in the data collection.
Competing interests None.
Ethics approval This study was conducted with the approval of the local ethic committees.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.