Objective The risk of cancer after exposure to the β-interferons (IFNβs) for multiple sclerosis (MS) has not been established. We assessed whether IFNβ treatment for MS is associated with cancer risk or the risk of specific cancers in a population-based observational study.
Methods The British Columbia MS database was linked to the provincial Cancer Registry, Vital Statistics death files and Health Registration files. Using a nested case-control design, MS cancer cases were matched with up to 20 randomly selected MS controls at the date of cancer diagnosis by sex, age (±5 years) and study entry year using incidence density sampling. Associations between treatment exposure and overall or specific (breast, colorectal, lung and prostate) cancers were estimated by conditional logistic regression, adjusted for MS disease duration and age. Tumour size at cancer diagnosis was compared between treated and untreated patients using the stratified Wilcoxon test to explore potential lead time bias.
Results The cohort included 5146 relapsing-onset MS patients and 48 705 person-years of follow-up, during which 227 cancers were diagnosed. Exposure to IFNβ was not significantly different for cases and controls (OR 1.28; 95% CI 0.87 to 1.88). There was a non-significant trend towards an increased risk of IFNβ exposure in the breast cancer cases (OR 1.77; 95% CI 0.92 to 3.42), but no evidence of a dose–response effect. Tumour size was similar between IFNβ treated and untreated cases.
Conclusions There was no evidence of an increased cancer risk with exposure to IFNβ over a 12-year observation period. However, the trend towards an association between IFNβ and breast cancer should be investigated further.
Statistics from Altmetric.com
Multiple sclerosis (MS) is a devastating chronic disease of the brain and spinal cord that is challenging to treat. All disease-modifying drugs (DMDs) licensed for MS to date have focused on modulating the immune response. As the immune system serves as the primary defence against cancer, and MS patients can be subject to long-term chronic exposure to these therapies, it is feasible that treatment with any of the DMDs could alter cancer risk in treated patients.1 However, rare adverse events that might be linked to treatment, such as cancer, are unlikely to be detected until the postmarketing setting, primarily because clinical trials include relatively small samples of select patients with relatively short follow-up. While spontaneous reporting systems can play a useful role for suspected drug-associated events, the reliability of this information is severely limited by under-reporting.2 ,3 Therefore, identification of rarer, long-term potential adverse effects can only be well assessed using systematically collected population-based data and long-term follow-up.
The β-interferons (IFNβs) and glatiramer acetate (GA) represent the first DMDs approved for MS (in the mid-1990s). Pivotal clinical trial data indicated that these therapies reduce relapse events by approximately a third and can significantly impact the incidence of brain lesions.4–7 IFNβ is the most widely used treatment for relapsing-onset MS in many jurisdictions, but very few studies have systematically assessed the risk of cancer in MS with IFNβ treatment. Our main objective was to determine whether the risk of overall cancer was altered among MS patients exposed to IFNβ. The secondary objectives were to investigate the relationship between IFNβ treatment and the risk of major site-specific cancers (breast, colorectal, lung and prostate cancers), and to investigate the possible role of lead time bias (earlier cancer diagnosis) among IFNβ treated MS patients. Last, we explored the association between GA treatment for MS and overall cancer risk.
Study design and data source
This was a nested case-control study using prospectively collected linked data from population-based clinical and administrative databases in British Columbia (BC), Canada. Details about the source cohort within which this study was nested have been previously published.8 Briefly, the cohort included all MS patients who were initially registered at one of the four MS clinics in the province of BC between 1980 and 2004. Clinical and demographic data were collated within the BCMS database,9 including date of MS symptom onset, clinical course (eg, relapsing-onset) and start and stop dates of all DMDs and immunosuppressants for the treatment of MS. The BCMS data were linked via unique personal health number to health administrative data: BC Vital Statistics which captures death dates for all residents of the province; the BC Registration and Premium Billing files,10 which document registration dates in the provincial healthcare plan and hence confirms each individual's residency in BC; and the BC Cancer Registry which provides information on malignancies diagnosed in BC by cancer type, date and site. The BC Cancer Registry contains information on all cancer diagnoses occurring in the province since 1969. The available follow-up information from the four databases combined extended from January 1986 to the end of December 2007, when the databases were linked and locked, and all identifiers were removed.
Study cohort and follow-up
All patients in the source cohort with a neurologist confirmed diagnosis of MS (Poser or McDonald criteria)11 ,12 and a relapsing-onset disease course were included. For the main analysis, patients were followed from the most recent of: 1 January 1996, MS symptom onset, 18th birthday or first confirmed residency in BC. Patients with exposure to any DMD (including immunosuppressants) prior to the start of follow-up were excluded. Follow-up ended at the earlier of: first relevant cancer diagnosis, initiation of a non-IFNβ DMD, emigration from BC, death or 31 December 2007. The inclusion criteria were broadly adapted from the eligibility criteria for financial coverage of a first line DMD in BC, which include assessment by a neurologist at an approved MS clinic, age of at least 18 years and a relapsing-onset disease course. The study start date of 1 January 1996 represents the beginning of the first full year following approval of IFNβ in Canada. Availability prior to approval was primarily through clinical trials. Residency in BC was required to ensure the capture of cancer events by the BC Cancer Registry.
Cancer outcomes and nested case-control analysis
The primary outcome was the diagnosis of any invasive cancer, as captured by the BC Cancer Registry. Breast, lung, colorectal and prostate cancers were examined as secondary outcomes. All patients with an incident malignant cancer event during follow-up were considered cases. For the few patients who had more than one primary cancer, only the first relevant cancer event was considered. Using incidence density sampling, up to 20 controls, matched for sex, age (±5 years) and calendar year at start of follow-up, were randomly selected for each case. Thus, cases and controls were matched for follow-up duration, all potential controls were free of a cancer event at the time of the cancer diagnosis in the case, controls were eligible to be selected as a control for more than one case and eligible controls could become cases later during follow-up.
The main exposure of interest was at least 90 days cumulative treatment with IFNβ prior to the cancer diagnosis or prior to the analogous date for the control. Patients with less than 90 days exposure were considered ‘unexposed.’ All IFNβ products were combined and considered as one therapeutic group. The potential for a dose–response effect was also explored for all cancer and for breast cancer (the numbers of lung, colorectal and prostate cases were insufficient to investigate dose–response effects). As there is no clear consensus on how to estimate and compare the biological activity (potency) and hence cumulative dose of the different IFNβ products,13 two approaches were taken. First, using the million international units (MIU) equivalence, an estimated cumulative dose was generated by summing the exposed time by the MIUs,13 resulting in a weekly equivalence of: 6 MIU for IFNβ-1a (intramuscular); 18 MIU for low dose (22 µg/thrice weekly) IFNβ-1a (subcutaneous); 28 MIU for IFNβ-1b (250 µg subcutaneous alternate days); and 36 MIU for high dose (44 µg/thrice weekly) IFNβ-1a (subcutaneous). The resulting cumulative MIU exposure was categorised to represent unexposed (<100), low (≥100 to <1500) or high (≥1500) exposure (ie, for this variable a minimum dose of 100 MIU, rather than minimum exposure duration of 90 days, was used as the criterion for ‘exposed’). Second, the cumulative time exposed to IFNβ was considered as: unexposed (<90 days), ≥90 days to <2 years or ≥2 years. Finally, to account for a potential latency between exposure and outcome, the elapsed time since IFNβ initiation was explored (unexposed, <5 years or ≥5 years). Separate models were built for each defined exposure because of the anticipated high correlation between these measures.
To account for any variation in the number of controls per case between risk sets, the descriptive summaries of case and control characteristics and exposure histories were weighted inversely by the number of controls in each matched set. ORs were estimated using conditional logistic regression; all models were adjusted for MS duration and a more precise measure of age (age per year) at start of follow-up. The overall cancer risk analysis was also estimated for men and women separately.
Tumour size at cancer diagnosis
To assess cancer stage (and potential lead time bias) between treated and untreated MS cancer cases, the primary tumour size (the ‘T’ component of the TNM staging classification14) was compared, first for the four most common cancers combined: breast, prostate (adenocarcinoma), colorectal and lung (non-small cell) and, second, for breast cancer only. Treated and untreated cancer cases were matched on tumour site, sex, age and year at cancer diagnosis and tumour size was compared using the stratified Wilcoxon test (van Elteren test), where strata weights were inversely proportional to the stratum sizes.
We assessed the sensitivity of our estimates in the event that: a cancer diagnosis was missed due to data linkage and residency issues; survival bias was introduced by following patients from disease onset; or DMD exposure was misclassified as ‘unexposed.’ We repeated our case-control selection and analyses in a ‘restricted cohort’ which included only patients with fully linked data (97% of the cohort). Follow-up started from the most recent of: first MS clinic visit (to address potential survivor bias), 1 January 1996, 18th birthday or first confirmed residency in BC. In addition, follow-up ended at the earlier of first cancellation of provincial health insurance (to exclude time spent out of province by the 5% of patients who were transient with multiple periods of provincial healthcare coverage), the most recent MS clinic visit (to eliminate potential exposure misclassification between January 2005 and 31 December 2007 when a fifth MS clinic was available in BC which was independent of the BCMS database), the first cancer event, death or first exposure to a non-IFNβ DMD. Findings were compared with the results from the main analysis.
The relationship between exposure to GA and all cancers was explored by selection of a new case-control sample from the cohort as described above, but follow-up continued after initiation of either GA or IFNβ. Exposure to GA was defined as ‘unexposed’ (<90 days) versus ‘exposed’ (≥90 days), and the models was adjusted for IFNβ exposure. Follow-up was terminated at the start of any other DMD (mitoxantrone, natalizumab or alternative drugs administered as part of a clinical trial), which was rare in this cohort. Exposure to GA was examined only as a secondary outcome because of its relatively low use in BC.15
Analyses were performed using Stata V.11 (Stata-Corp; 2009) and R:A Language and Environment for Statistical Computing V.2.15.01 (R Foundation for Statistical Computing; 2010). The Clinical Research Ethics Board of the University of British Columbia provided ethical approval.
The linked source cohort included 5146 relapsing-onset MS patients (75% women) who were naive to any MS DMD at the start of follow-up and resident in BC after 1 January 1996 with 48 705 patient-years of follow-up (mean 9.5 years; SD 3.4). The mean age was 32 years (SD 9.6) at MS symptom onset and 42 years (SD 11.4) at the start of follow-up. Among the patients with no cancer event during follow-up, 3512 were followed to study end, 666 were followed to the start of a non-IFNβ (70% of whom started GA), 431 patients died and 309 were followed to emigration from BC. Not unexpectedly, patients who were younger at study start (mean age 39 years) or younger at MS onset (mean age 29 years) were more likely to have emigrated from BC before a cancer event was detected. Those who emigrated were distributed equally by sex (6% of men and 6% of women).
Overall, 227 incident cancer cases were identified. The mean age at cancer diagnosis was 57.8 years (SD 10.8) and 77% of cases were women. Cases and controls were similar for the matching variables, that is, sex, follow-up time to event and age, and also for the mean disease duration at cancer diagnosis (see table 1).
A similar proportion of all cancer cases (17%) and their matched controls (15%) were exposed to IFNβ treatment. Conditional logistic regression analyses revealed no significant difference by treatment exposure (‘ever vs never’) (OR 1.28; 95% CI 0.87 to 1.88) or by cumulative dose (whether measured as MIUs or cumulative time), or by elapsed time since initiation of IFNβ (see table 1). When split by sex, the odds of IFNβ exposure were similar between cases and their controls for women (OR 1.28; 95% CI 0.83 to 1.97) and for men (OR 1.31; 95% CI 0.57 to 2.99).
There were 63 breast (see table 2), 20 lung, 18 colorectal and 12 prostate (see table 3) cancers. There was no evidence of an altered risk of IFNβ exposure among lung, colorectal or prostate cancer cases and their matched controls. However, there was a non-significant trend towards a higher risk of breast cancer in IFNβ treated patients, with 25% of breast cancer cases exposed to IFNβ versus 17% of matched controls (OR 1.77; 95% CI 0.92 to 3.42). Nonetheless, there was no indication of a dose–response effect; neither increasing cumulative dose or cumulative time nor elapsed time since initiation of IFNβ was associated with increased breast cancer risk (see table 2).
Tumour size at cancer diagnosis
Of the 113 breast, lung, colorectal and prostate cancers diagnosed in the cohort during follow-up, 57% had documented tumour size data. After matching for sex, age at cancer diagnosis, year of cancer diagnosis and tumour site, there were no significant differences in the tumour size of combined cancers (n=64; Z=1.317; p=0.21) or of breast cancers (n=50; Z=0.898; p=0.42) between the IFNβ treated and untreated cases.
Following restriction of the source cohort for sensitivity analyses, there were 164 cancer cases, including 47 breast cancer cases. Findings were similar to those seen in the main analyses; there were no significant differences in the odds of exposure to IFNβs in the cancer cases (OR 1.08; 95% CI 0.68 to 1.72) or in the breast cancer cases (OR 1.76; 95% CI 0.83 to 3.74) compared with their controls.
By including observation time after initiation of GA, the available follow-up time increased to 50 971 patient-years (mean 9.9 years; SD 3.1). Only 2.6% of the 233 cases had a history of exposure to GA and their treatment history did not differ from the controls (OR 1.06; 95% CI 0.46 to 2.49).
In this case-control study nested in a large prospective cohort of relapsing-onset MS patients, we found no substantial differences in IFNβ exposure between patients with a diagnosis of malignant cancer and their matched controls. Neither was there evidence of an altered risk when lung, colorectal or prostate cancers were examined separately. While there was a trend towards an association between IFNβ exposure and the risk of breast cancer, this did not reach statistical significance and no dose–response relationships with either breast cancer or overall cancer were observed.
Very few longitudinal studies of MS cohorts have assessed the risk of cancer in patients treated with DMDs. Our findings are somewhat in contrast to those from a smaller study from Israel of 1338 MS patients (15 of whom were cancer cases following DMD treatment) which showed borderline associations between non-breast cancer risk and IFNβ treatment, as well as between breast cancer risk and GA treatment, neither of which reached statistical significance.16 This is the only published study that we are aware of, other than ours, to address this question using linked population-based cancer registry data to capture cancers. However, the authors were unable to explore the existence of a dose–response effect which can add biological plausibility to findings. We considered cumulative IFNβ dose, as well as the elapsed time since initiating IFNβ, and found no evidence of an association with overall cancer risk.
Our main findings concur with those from the study from France in which MS patients followed at 12 MS centres with or without cancer were identified through the European Database for MS; the findings revealed no increased risk of cancer with exposure to either the IFNβs or GA.17 Although this study reported a greater cancer risk with exposure to multiple DMDs or immunosuppressants, potential unequal follow-up time, and hence differential opportunity for treatment exposure, combined with cancer information obtained from a MS database rather than a cancer-specific population-based registry, may limit the interpretation of these observations.17 ,18
Two industry sponsored studies have reported no malignancy risk with IFNβ-1a (intramuscular) or IFNβ-1a (subcutaneous) treatment. However, these studies faced a number of challenges, including the identification of drug-related cancer cases from spontaneous reports of adverse events in the Global Drug safety Database19 ,20 of which only an estimated 6% are ever reported.3 Furthermore, data were supplemented with information from clinical trials20 or an insurance claims database19 and in both situations the observation periods were likely too short to rule out cancer risk (2–3 years).
There are a few published case reports that have raised concerns about a possible link between IFNβ (or GA) and breast cancer,21 and other sites of malignancy.22–27 It is reasonable to hypothesise that the IFNβs could be linked to an altered cancer risk, given that they are known to be immunomodulatory agents28 ,29 and that the immune system provides the primary defence against cancer. Although interferons (predominantly α-interferons) have been used to treat cancer in other settings, therapeutic use of the IFNβs in MS is quite different with exposure more likely to be chronic and often initiated at a relatively young age.
The reason why breast cancer may have emerged as possibly associated with IFNβ (albeit not significantly) is unclear, and the absence of a dose–response effect might argue against a true association. However, breast cancer occurs at a younger age on average compared with the other cancers assessed and is also the most frequent cancer in this predominantly female patient population. It is therefore feasible that a small effect on general cancer risk could be first apparent in breast cancers.
We were able to explore the possibility of a lead time bias which could result if patients who are exposed to IFNβ (or any DMD) are subject to enhanced monitoring, and hence an earlier detection of cancer. In the subcohort of cancer cases where this information was available, we found no evidence to suggest that treated patients were monitored more carefully for cancer, as the tumour size distribution at cancer diagnosis was comparable for IFNβ exposed and non-exposed cancer cases, even after consideration of the confounding influences of tumour site, sex and year of diagnosis. In addition, our findings do not provide support for the ‘autoimmune aggression’ hypothesis, that is, that IFNβ enhances tumour growth,21 ,22 ,24 ,25 ,30 in which case we might expect to see cancers diagnosed at a later stage (or larger size) in treated patients. However, we cannot rule out that the potential effects of more vigilant monitoring of treated patients and an enhancement of tumour growth of IFNβ cancel each other out. We are not aware of any other study exploring the possibility of lead time bias in treated MS patients.
Our study includes a substantial source cohort of more than 5000 relapsing-onset MS patients, all of whom were diagnosed by neurologists specialising in MS, along with population-based cancer registry, vital statistics and administrative health data which ensured reliable prospective coding of incident malignant cancers and minimal loss-to-follow-up and avoids the potential biases that can be associated with non-population-based sources of data. Furthermore, the BC MS database incorporated systematic documentation of all exposures to DMDs which provided accurate coding of individual treatment histories. By use of the case-control design and incidence density sampling method, cases were matched to controls for age, sex, date of study entry and follow-up time thus ensuring that controls had equal opportunity time for exposure to treatment during the same era. Among the few studies of MS cohorts that have addressed exposure to immunomodulatory agents and cancer risk,16 ,17 our study includes the largest number of cancer cases diagnosed after MS symptom onset.
There are, however, limitations to our study. Although our findings provide reassurance that after an average of 9.5 years of follow-up there was no evidence of an increased cancer risk associated with IFNβ exposure, it remains possible that an effect could be found with an even longer observation period or in a larger cohort. Further, although we had access to detailed start and stop dates for the IFNβs, we assumed that while ‘on drug’, patients were adherent. We did not have access to all potential confounders, which might include lifestyle-related factors (eg, exercise, diet, smoking), previous cancer diagnosis (ie, pre-1969 or preimmigration to BC) or a family history of cancer. Although none of these factors are known as contraindications to IFNβ treatment, they still may have influenced a patient's or physician's treatment decision or modified an effect of the IFNβs on cancer risk. Our definition of exposure to IFNβ (90 days minimum) was somewhat arbitrary, guided by clinical trial evidence of the minimum exposure time before a measurable effect on MRI outcomes is achieved.31 However, our dose–response analysis did not reveal even a trend towards a greater risk with higher cumulative exposure. Last, the power of this study to assess the relationship between the IFNβs and specific cancer sites was limited and few cancer cases were exposed to non-IFNβ DMDs in our cohort, allowing only a basic descriptive analysis of the potential effects of GA (although no suggestion of an association with cancer risk was observed).
In conclusion, our findings provide reassurance that in the real world clinical setting, the overall cancer risk does not appear to be increased by exposure to the IFNβs, in either men or women with MS, after up to 12 years of follow-up. While our observation surrounding breast cancer warrants further investigation, it is worth noting that the relatively small effect size observed suggests that any potential increased risk is unlikely to have a noticeable effect on absolute risk. Nonetheless, further study of these rare outcomes, particularly breast cancer, in larger or combined observational MS cohorts after a longer follow-up period is warranted, as is continued vigilance.
We gratefully acknowledge Dr Chris Bajdik (formerly at the BC Cancer Agency and School of Population and Public Health, University of British Columbia) for his contributions to the initial study development and successful funding application, and for his invaluable help with access to the BC Cancer Registry data. We are also grateful to Tom Duggan (Division of Neurology, University of British Columbia) for help with data handling and coding, Dr Dean Eurich (School of Public Health, University of Alberta) for advice with the initial study design and Dr John Spinelli (BC Cancer Agency and School of Population and Public Health, University of British Columbia) for helpful comments on the draft manuscript prior to submission. We thank the BC Vital Statistics Agency, the BC Ministry of Health and the BC Cancer Agency for approval and support with accessing provincial BC administrative health, BC Vital Statistics and BC Cancer Registry data; and Population Data BC for facilitating approval and use of the data. We gratefully acknowledge the BCMS Clinic patients for sharing clinical information and the BCMS Clinic neurologists who contributed to the study through patient examination and data collection (current members listed here): A Traboulsee, MD, FRCPC (UBC Hospital MS Clinic Director and Head of the UBC MS Programs); A-L Sayao, MD, FRCPC; V Devonshire, MD, FRCPC; S Hashimoto, MD, FRCPC; J Hooge, MD, FRCPC; L Kastrukoff, MD, FRCPC; J Oger, MD, FRCPC; D Adams, MD, FRCPC; D Craig, MD, FRCPC; S Meckling, MD, FRCPC; L Daly, MD, FRCPC; O Hrebicek, MD, FRCPC; D Parton, MD, FRCPC; and K Pope, MD, FRCPC. The views expressed in this paper do not necessarily reflect the views of each individual acknowledged.
Contributors EK was responsible for study coordination, designing the study, acquiring the data, ensuring the data integrity, managing the data, statistical analysis, interpretation of the results and drafting the manuscript. CE contributed to the study design and provided intellectual comments and editing of the manuscript. FZ contributed to statistical analysis and editing of the manuscript. JO and SH helped obtain funding for the study and provided intellectual comments and editing of the manuscript. HT also obtained funding, contributed substantially to the study design, acquisition of data and interpretation of the results, and provided intellectual comments and editing of the manuscript. All authors read and approved the final version of the manuscript.
Funding Funding support for this study was provided by the Canadian Institutes of Health Research (MOP-82738; PI:HT). EK was funded by the Michael Smith Foundation for Health Research and the MS Society of Canada (Postdoctoral Fellowships) and HT is funded by the MS Society of Canada (Don Paty Career Development Award); is a Michael Smith Foundation for Health Research Scholar; and the Canada Research Chair for Neuroepidemiology and Multiple Sclerosis. The BCMS database has been funded from multiple sources including an unrestricted grant from Dr Donald Paty and the ‘MS/MRI Research Group,’ the MS Society of Canada (PI:HT) and CIHR (PI:HT).
Competing interests This study was funded by the Canadian Institutes of Health Research. The study sponsor had no role in the study design, acquisition of data, statistical analysis, interpretation of results, drafting or editing of the manuscript, or decision to submit. EK was funded by the Michael Smith Foundation for Health Research and the MS Society of Canada (Postdoctoral Fellowships). She has received reimbursement of travel, accommodation or registration costs to present at and attend conferences from the endMS Research and Training Network and the International Society for Pharmacoepidemiology. CE was funded by the Michael Smith Foundation for Health Research (Postdoctoral Fellowship) and has received travel and accommodation reimbursement to present at and attend conferences from the European Committee for Treatment and Research in Multiple Sclerosis. FZ reports no disclosures. JO has received speaker honoraria, consulting fees, travel, research or educational grants from Bayer, Biogen-Idec, Novartis, Teva, Genentech and Serono. SH has received Advisory Board payments (Biogen Idec; Novartis; Teva); assistance for travel and education speaking engagements (Biogen Idec), and for outreach clinics (Bayer, Biogen Idec). HT is funded by the MS Society of Canada (Don Paty Career Development Award), and is a Michael Smith Foundation for Health Research Scholar and the Canada Research Chair for Neuroepidemiology and Multiple Sclerosis. She has received speaker honoraria and/or travel expenses to attend conferences from: the Consortium of MS Centres (2013), the MS Society of Canada (2013), the National MS Society (2012), ECTRIMS (2011, 2012, 2013), the University of British Columbia MS Research Program, the UK MS Trust (2011), the Chesapeake Health Education Program, US Veterans Affairs (2012; honorarium declined), Bayer Pharmaceuticals (speaker, 2010; honorarium declined), Teva Pharmaceuticals (speaker 2011), Novartis Canada (2012) and Biogen Idec (2014; honorarium declined). Unless otherwise stated, HT speaker honoraria are donated to an MS charity or to an unrestricted grant for use by her research group.
Ethics approval The Clinical Research Ethics Board of the University of British Columbia approved the study, which included patient consent.
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.