Abstract
Objective. Patients with ANCA-associated vasculitides (AAV) exhibit higher rates of malignancy than the general population. We assessed whether the cancer risk is increased in a well-characterized population-based cohort of AAV in southern Sweden, followed for a median time of 8 years.
Methods. With case record review, the outcomes and malignancy development in a cohort of 195 patients with AAV [granulomatosis with polyangiitis (GPA), microscopic polyangiitis, and eosinophilic GPA] diagnosed between 1997 and 2010 were assessed. The patients were followed until death or December 31, 2015. The age- and sex-standardized incidence ratios (SIR) were estimated using the Swedish population data as a reference.
Results. During the observation period of about 1500 person-years, we found 60 cancers in 52 of the 195 patients. SIR (95% CI) was 2.8 (2.1–3.6) for cancers at all sites, 1.8 (1.3–2.5) for all cancers excluding squamous cell carcinoma (SCC), 12.9 (8.4–18.8) for SCC, 4.3 (1.4–10.0) for bladder cancer, and 7.0 (1.4–20.5) for pancreatic cancer. Cumulative doses of cyclophosphamide (CYC) < 10 g were not associated with higher incidence of cancers other than SCC (SIR 1.63, 95% CI 0.8–2.9).
Conclusion. In contrast to previous publications assessing malignancy risk in patients with AAV, we show in this population-based cohort of patients a persistent increased risk for overall malignancy, bladder cancer, and pancreatic cancer as well as a markedly increased risk for SCC. There was no increase in incidence of cancers other than SCC for those treated with < 10 g CYC.
Systemic small vessel vasculitides, comprising granulomatosis with polyangiitis (GPA), microscopic polyangiitis (MPA), and eosinophilic granulomatosis with polyangiitis (EGPA), are in most cases associated with antibodies for proteinase 3 and myeloperoxidase in neutrophil granules1. GPA, MPA, and EGPA are grouped together under the name antineutrophil cytoplasmic antibody (ANCA)-associated vasculitis (AAV)2. With the introduction of immunosuppressive agents such as glucocorticoids and cyclophosphamide (CYC), survival has improved substantially but is still low compared with the general population in the majority of patients3. The standard treatment regimens today consist of high-dose corticosteroids combined with CYC in the acute phase followed by lower doses of corticosteroids in combination with azathioprine (AZA) or methotrexate (MTX)4. Rituximab (RTX) is a drug that has been approved as an alternative to CYC in the acute phase as well as an option for treating relapses5,6,7. AAV often relapses, requiring repeated treatments with immunosuppression8.
As survival of patients diagnosed with AAV has dramatically improved, the side effects of the immunosuppressive treatment have become a focus of interest, and the development of malignancies is a major concern9. Previously published observations suggested that the overall cancer incidence in patients treated for AAV is 1.6–2.4 times higher than in the general population, and the relationship between malignancy and cumulative dose of CYC has been highlighted in some of these studies10,11,12,13,14,15. Interestingly, in a more recent study, patients treated with RTX did not show an increased malignancy risk compared with the general population16.
Within the European Vasculitis Society (EUVAS), we previously assessed the longterm malignancy risk in patients with AAV who had participated in one of 4 randomized clinical trials12,17,18,19,20. Even though the EUVAS studies included patients with variable clinical manifestations and disease severity, patients with EGPA and those who had a known previous malignancy were excluded.
We investigated longterm malignancy risk in a population-based cohort of 195 patients with AAV without any referral or exclusion bias, including all the diagnoses within the AAV spectrum, and these patients were followed for a median time of 8 years.
MATERIALS AND METHODS
Study design and participants
The occurrence of malignancy was assessed in a cohort of all incident cases with AAV in a defined geographical area of about 0.7 million inhabitants in southern Sweden. Potential cases of AAV, diagnosed between 1997 and 2010, were identified through healthcare registries using a previously described validated search algorithm21 and were reviewed using case records. Patients were considered to have AAV if they had symptoms and signs compatible with small vessel vasculitis supported by histopathology, radiology, and/or serological findings, and if they had been classified as having GPA, MPA, or EGPA according to the European Medicine Agency 2007 algorithm22. By review of each case record and pathology report for the 195 patients in the cohort, data of any malignancy occurring after the diagnosis of AAV were collected until December 31, 2015. If a patient developed several squamous cell carcinomas (SCC), only the first was reported. Likewise, if a patient developed several other cancers excluding SCC, only the first was reported. However, if a patient developed both SCC and any other cancer, the first of both malignancies was reported. In the Swedish Cancer Registry, basal cell carcinoma has been reported only since 2004; therefore, the study did not include this diagnosis as no data on this disease are available for the general population throughout the period of investigation.
The study was performed in accordance with the Declaration of Helsinki and was approved by the Ethical Review Board in Lund (2010-517).
Baseline evaluation and treatment
Results from laboratory investigations collected at the diagnosis of AAV included plasma creatinine and ANCA serotype analyzed with ELISA and capture ELISA. According to then-current guidelines used in southern Sweden, the majority of patients received CYC and glucocorticoids as induction treatment. CYC was given orally, intravenously (IV), or in many cases, both orally and IV at different times. Patients with creatinine ≥ 500 µmol/l or those with rapidly progressive glomerulonephritis or pulmonary hemorrhage received plasma exchange and/or IV methylprednisolone in addition to CYC as induction treatment. All patients who were treated with IV CYC also received mesna to protect the urinary bladder. For those without systemic involvement (i.e., exclusively upper respiratory involvement), induction therapy was MTX together with corticosteroids. After stable remission, the treatment continued with AZA in the majority of patients. Thirty-one patients were treated with RTX, of whom 4 did not receive any CYC, 11 received 1–10 g, and 16 received > 10 g. For all included patients, medical charts were reviewed, and the cumulative doses of CYC given throughout the entire followup period were calculated. In addition, information on immunosuppressive drugs other than CYC was gathered.
Statistical analysis
Continuous variables are presented as medians and interquartile ranges (IQR). Categorical variables are expressed as counts and frequencies.
Estimation of cumulative cancer incidence and survival
Cumulative incidences of cancer, overall survival, and cancer-free survival were calculated using the Kaplan–Meier method. The calculations regarding cancer were performed both for all cancers and for SCC. Person-years (PY) of followup were calculated, starting from the date of diagnosis of AAV and ending at the earliest time of cancer of the selected kind, death, or the end of followup on December 31, 2015. The PY were calculated until the development of each specific cancer, regardless of a diagnosis of other cancers. To enable comparisons with our previous study12 as well as many others in the calculation of cumulative incidence, the occurrence of cancer was considered an event, and patients were followed until death or the end of the study period. For calculation of cancer-free survival, the occurrence of cancer or death was considered an event, and patients were followed through the end of the study period23.
Calculation of standardized incidence ratios (SIR)
For each patient, PY of followup were calculated using the start and end dates described in the previous section; these calculations were made separately for each of the identified cancer endpoints. The PY risk was calculated by sex and by age in 5-year groups. SIR were calculated to compare the observed malignancies in the cohort with the expected numbers in the general population matched for sex, 5-year age groups, and 1-year calendar time period (i.e., for each year the patients grow older) and when appropriate, the corresponding 5-year age groups in the general population were changed. A published SAS macro24 was modified and expressed in the MATLAB language for these calculations. The expected numbers of cases of cancers of each kind were calculated by multiplying the number of PY for each sex and age group by the corresponding cancer incidence rates as provided by the Swedish Cancer Registry25. Each healthcare provider in Sweden is obliged to report to the Swedish Cancer Registry every cancer case diagnosed at clinical, morphological, and other laboratory examinations as well as cases diagnosed at autopsy. SIR were calculated for cancers at all sites, for SCC and for each of the reported specific cancers in the cohort. SIR were stratified by disease phenotype, ANCA serotype, sex, and age group at entry (below vs above 60 yrs), chronic kidney disease group (CKD) at entry (CKD 1–2 vs CKD 3–5), previous malignancy (i.e., before diagnosis of AAV), and renal transplantation. We also performed a series of Poisson regressions to generate relative risks for cancer incidence in each subgroup.
RESULTS
Patients
One hundred ninety-five patients (98 female) with a median age of 69 years (IQR 55–77) at diagnosis of AAV fulfilled the inclusion criteria and were included in the study cohort. Among these patients, 94 (48%) were classified as having GPA, 90 (46%) as MPA, and 11 (6%) as EGPA. The followup time ranged between 0.3 and 18 years, with a median followup time of 8 years (IQR 4.0–11.9). Ninety-eight (50%) of the patients died during followup. Further demographic data are summarized in Table 1. Of the patients included in this cohort, 7 had also participated in the EUVAS longterm followup and malignancy studies12; of these 7 patients, 1 developed malignant melanoma and SCC and 1 developed lung cancer (both of whom participated in the MEPEX study)20.
Baseline and followup demographics in 195 patients with incident ANCA-associated vasculitis.
Numbers and characteristics of observed cancers
Fifty-two of 195 patients developed 60 malignancies during the followup time of 1300 PY. There were 26 SCC and 34 other types of cancers. The median time from the diagnosis of AAV to the first cancer diagnosis at any site was 5.13 years; to first SCC, 4.58 years; and to first malignancy of other types, 5.50 years. Overall survival rate at 1, 2, 5, and 10 years was 87%, 82%, 70%, and 55%, respectively, and the cancer-free survival rate for cancer at any site at 1, 2, 5, and 10 years was 84% (95% CI 78–89), 77% (71–83), 61% (54–68), and 44% (37–52), respectively. The cancer-free survival at any site excluding SCC was 86% (95% CI 81–91), 80% (75–86), 66% (60–73), and 49% (42–57), respectively (Figure 1). The cumulative 5- and 10-year incidence for cancers at all sites was 17% (95% CI 11–23) and 30% (22–38), respectively (Figure 2). The respective incidence for cancers of types other than SCC was 10% (95% CI 5–15) and 20% (12–26; Figure 2).
Overall survival (Dead), cancer-free survival for cancer at any site (D), and cancer-free survival for cancer at any site excluding squamous cell carcinoma (C).
Cumulative incidence of cancer at any site (B), and cumulative incidence of cancer at any site excluding squamous cell carcinoma (A).
Among the 52 patients with cancer at any site, 29 died and among the 34 patients with non-SCC, 18 died; 15 of the total 98 deaths were attributed to cancer. Sixteen patients had had malignancies prior to the diagnosis of AAV; 8 of these were diagnosed with cancer during the followup period of this study, and another 8 had no new cancer detected (p = 0.026). Seven patients underwent renal transplantation, and 4 of these were diagnosed with cancer.
SIR of malignancies
The SIR for all site cancers was 2.77 (95% CI 2.07–3.64, p < 0.001) and for all sites excluding SCC was 1.81 (95% CI 1.26–2.54, p = 0.001). Among the cancer site-specific analyses, incidences of SCC, bladder, and pancreatic cancers were significantly increased when compared with the general population. The SIR for SCC was 12.85 (95% CI 8.40–18.83, p < 0.001), the SIR for bladder cancer was 4.30 (95% CI 1.40–10.04, p = 0.003), and the SIR for pancreatic cancer was 7.00 (95% CI 1.44–20.45, p = 0.002; Table 2). During the first 5 years of followup, a total of 26 cancers were observed. The corresponding figures for SCC, for cancer at all sites excluding SCC, and bladder cancer were 14, 16, and 3, respectively. The cumulative followup time for the first 5 years was 735 PY, and SIR was 2.60 (95% CI 1.70–3.80) for all sites, 14.9 (8.14–25.0) for SCC, 1.66 (0.95–2.69) for other types, and 5.36 (1.11–15.7) for bladder cancer.
Calculated age, sex, and calendar year SIR for observed cancers in 195 patients with ANCA-associated vasculitis.
Table 3 shows subgroup analyses for SIR grouped by the predefined variables. The analysis indicates higher SIR for SCC in patients with EGPA, for those with negative ANCA, for patients younger than 60 years at diagnosis of AAV, and for those with renal transplants. The results for overall cancer excluding SCC indicate higher SIR for patients with previous malignancies and those with renal transplant, but a lower SIR for patients with EGPA and patients with negative ANCA.
Standardized incidence ratios (SIR) for cancer at all sites, all sites excluding SCC, and SCC according to selected subgroup variables.
Cumulative doses of CYC < 10 g were not associated with higher incidence of cancers other than SCC (SIR 1.63, 95% CI 0.8–2.9). Cumulative doses of CYC > 36 g exhibited SIR of 3.4 (95% CI 1.5–6.4) for cancers other than SCC. The risk for SCC was markedly increased in patients not receiving CYC (SIR 53.6 (95% CI 21.6–110.5). Among the 5 patients who developed bladder cancer, 3 were treated with cumulative CYC doses > 10 g (151, 31, and 17 g, respectively), and 2 patients were treated with doses < 10 g (7.5 and 8 g, respectively). Table 4 shows risk ratios calculated using Poisson regression, showing, for example, that patients with previous malignancies had a risk ratio of 3.9 (1.5–8.9, p < 0.001) compared with patients without previous malignancies.
Relative risks (RR), calculated by Poisson regression, according to selected subgroup variables.
DISCUSSION
In this population-based cohort of patients, which includes all 3 diagnoses within the AAV spectrum, we show that risk for overall malignancy, bladder cancer, pancreatic cancer, and SCC is still increased when compared with the general population.
A 2.77-fold increase in the risk for overall malignancy was seen in the AAV cohort compared with that in the general population. The highest risk was seen in SCC and was more than 12 times higher than that in the general population; the next highest risks were for pancreatic cancer, with a 7-fold increase, and bladder cancer, with a 4.3-fold increased risk compared to the general population. There were no other significantly increased risks for other reported malignancies.
Previous studies assessing malignancy risk in AAV have reported SIR for overall malignancy of different ranges. Earlier studies26 have shown a 33-fold increased risk for bladder cancer, which was believed to be due to the use of higher doses of CYC for longer periods of time, but with current treatment protocols, these high risks are no longer seen. In more recent studies, the SIR for overall cancers range from nonsignificant27 to 3.7611,12,13, the SIR for bladder cancer ranges from nonsignificant to a 3.6-fold increase, the SIR for the nonmelanoma skin cancer (including SCC and basal cell carcinoma) ranges between a 2.8 and 4.7 increase, and the SIR for leukemia ranges from nonsignificant to a 5.9-fold increase.
Two studies, one the EUVAS followup study of multi-center clinical trials12 with a followup time of almost 5 years, and the other a cohort of 138 patients identified through a pathology database13 and followed for a median of 10 years, both stress that the malignancy risk in patients with AAV treated according to current protocols is increased but that this increase is solely driven by the increased risk for SCC. However, in our study, the risk for SCC as well as the risk for both bladder and pancreatic cancer is significantly increased. The increased risk for bladder cancer is consistent with other previous studies from 2002 and 2006, but pancreatic cancer has not been shown to be increased in other studies. However, the patients who developed bladder cancer were treated with high cumulative doses of CYC, and the interpretation of the finding of increased risks for pancreatic cancer must be made with caution, because the cohort is small. The treatment with cumulative doses of CYC < 10 g, corresponding to IV CYC for 3–6 months, was not associated with a higher risk for malignancies excluding SCC, indicating that the dosing of CYC with current treatment protocols is probably safe and not associated with an increased malignancy risk. However, cumulative doses of CYC of > 10 g were associated with higher incidence of cancers beyond SCC and an even higher risk was seen with doses > 36 g. The cutoff of 36 g was chosen for comparison with a previous study from Faurschou, et al, corresponding to treatment with 100 mg CYC/day for > 1 year11.
The discrepancies, however small, between our findings and those in previous studies might have several explanations. The assessment of malignancies in Sweden is very thorough, and each healthcare provider is obliged to submit a report to the Swedish Cancer Registry for every cancer case diagnosed at clinical, morphological, and other laboratory examinations as well as in cases diagnosed at autopsy. The assessment of malignancies both in our cohort as well as in the general population is reliable, and the probability that any cancer case is missed is very low. In our current study we included patients with previous malignancies, and in this group the risk for additional malignancies is known to be higher28. In addition, the excess risk of SCC might be explained to some extent by the high risk in patients with EGPA who have not been included in previous studies, with the exception of the most recent one by van Daalen, et al16. In our study 7 patients received a renal transplant, and in these patients the SIR was higher, especially for SCC, also adding to the higher risk in our previous study compared to others assessing malignancy risk in AAV. The risk for SCC among those patients receiving transplants has in several studies been shown to exceed that in the AAV population29. The most probable explanation for the high risk for SCC that we see in patients not receiving CYC is that these patients received other oral immunosuppressive drugs such as AZA, known to increase the risk for nonmelanoma skin cancer including SCC30.
The main limitation of our study is the relatively small number of patients. A larger number of patients would have strengthened the results, but the significantly increased malignancy risk seen already in a smaller cohort shows that this still is an important factor to keep in mind in the followup of patients with AAV. Unfortunately, it is not possible to make a firm conclusion on the pathogenesis of the malignancies in our patients. In our study, the main focus was on the relationship between malignancies and immunosuppression because no data are available on other potential risk factors for cancer in AAV.
Our study has important strengths. We used a population-based cohort of patients with AAV in which the diagnoses of AAV are scrutinized by reviewing case records, and all patients are classified according to a defined algorithm using the European Medicines Agency classification. Further, all diagnoses of malignancies in the patients have been collected through case records, pathology, histology, and cytology databases without any missing data. In addition, the number of PY of followup is about 1500 for a number of cancer diagnoses, and we were able to show important differences when comparing with the general population. Finally, the Swedish Cancer Registry, which includes data from the entire Swedish population, enables comparison with the general population.
In this population-based cohort of patients, which includes all 3 diagnoses within the AAV spectrum, we show that the malignancy risk is still increased when compared with the general population, but in patients receiving CYC doses of < 10 g, this is solely driven by the increased risk for SCC. In contrast to the most recent publications assessing malignancy risk in patients with AAV, we show a persistently increased risk for bladder cancer and pancreatic cancer in patients receiving CYC > 10 g, indicating that the risk for the development of solid tumors in patients treated for AAV should not be totally ignored.
Footnotes
Supported by The Swedish Rheumatism Association (Reumatikerförbundet), The Swedish Society of Medicine, The King Gustaf V’s 80-year foundation, and Region Skåne (REGSKANE-280361).
- Accepted for publication August 19, 2019.
REFERENCES
In this issue
