Cancer risk among patients with multiple sclerosis and their parents

Cancer risk among patients with multiple sclerosis (MS) is important, as they may be subject to immunomodulatory therapies potentially altering cancer risk. A large Danish study¹ and some smaller studies^2–4 reported a lower overall cancer risk in MS, although there are contradictory results.⁵ MS has been associated with a reduced risk for digestive, respiratory, prostate, and ovary cancers¹ and non-Hodgkin lymphoma⁶ but increased risks for urinary tract and nasopharyngeal cancers.¹ Breast cancer risk has been reported as higher,^1,4 lower,³ or unchanged.^2,5

Besides assessing cancer risk in a nationally representative group of patients with MS diagnosed over several decades with long follow-up and general population-based controls, this article is concerned with establishing if any reduced cancer risk is due to changes following disease onset. This could include lifestyle changes following diagnosis, or immunologic changes due to disease activity: these influences may not be mutually exclusive and younger age at MS onset could confer greater protection. Some genetic MS susceptibility factors may also protect against cancer. Reduced cancer risk among parents of patients with MS would be consistent with this genetic explanation. As inflammation increases cancer risk,⁷ we hypothesized that patients with MS are at higher risk of brain tumors due to chronic neurologic inflammation.

METHODS

Patients who received a diagnosis of MS in Sweden between 1969 and 2005 were identified using the national Inpatient Register,⁸ which records the discharge diagnoses of all inpatients in Sweden, and the national Swedish Multiple Sclerosis Register, which requires written informed consent for inclusion.⁹ To assess diagnostic accuracy, we randomly sampled 112 patients in the MS Register, and after excluding suspected MS as identified in the register, the diagnosis was confirmed in 105/109; 96% (95% confidence interval [CI]: 91–99%).

Subjects with MS were individually matched with 12 individuals (fewer in a small minority) without the disease among the general Swedish population by Statistics Sweden. The matching criteria for controls were year of birth, sex, vital status, region of residence (the 24 counties of Sweden), and age at the time of diagnosis in the matched case. The Multi-Generation Register¹⁰ was used to identify both biological parents of subjects. The Swedish migration and death registers have provided information on date of migration or death among index subjects and their parents. We used the national Swedish Cancer Register to identify the date and diagnosis of cancer among cases, controls, and their parents. This register was established in 1958 and records all newly diagnosed malignant tumors in Sweden, and reportedly over 98% complete.¹¹ All health care providers in Sweden are obliged to report newly detected cancer diagnoses. The Cancer Register used ICD-7 codes and a WHO pathology code where applicable.

A six-category socioeconomic index was based on occupation identified from the census nearest in time to study entry (manual workers, nonmanual workers, professionals, self employed, farmers, and others). All data were linked using the unique personal identity number issued to all Swedish residents.

From among 20,543 cases and 204,163 controls, we excluded 267 cases and 212 controls due to missing information. We had information on biological fathers for 11,284 cases with MS and 123,158 controls, and biological mothers for 12,006 cases and 129,409 controls. There were no differences in the demographic characteristics of subjects with and without MS when those with linked data on mothers and fathers were compared with those who did not have linked data. Index subjects (with MS and the comparison group without MS) with linked data had a higher mean age at exit (69 years compared with 50 years), and longer follow-up time (39 years compared with 29 years). This is mainly because those without linked data for parents were born earlier and censored at an earlier timepoint. They could not be linked to their parents because they were born before 1932 or their parents died before 1947, the time that the unique personal identity number was first issued.¹⁰ It is important to note that the parents of cases and controls who entered the study had the same age distribution at final follow-up, average age at diagnosis of cancer, and average follow-up time (table 1). Therefore, there was nothing to indicate differential bias between cases and controls.

This study was approved by the Karolinska Institutet regional ethics committee.

RESULTS

The study population is described in table 1 and includes 20,276 patients with MS (65.2% women) and 203,951 subjects without the disease. Some 29.3% and 2.0% of patients with MS were censored due to death and emigration, and the corresponding percentages for subjects without the disease were 15.9% and 4.1%. The analysis of parents used information on 11,284 fathers and 12,006 mothers of patients with MS, with 123,158 fathers and 129,409 mothers of index subjects without MS.

DISCUSSION

This large general population-based register study of patients with MS and their parents confirmed a reduction in cancer risk of approximately 10% among those with MS and this effect was more pronounced among women. A raised risk was observed for some specific cancers including brain tumors and cancer of the urinary organs. No notable pattern of cancer risk or protection was observed among parents of those with MS.

The cancer risk reduction among patients with MS is consistent with previous studies,^1–5 demonstrating decreased risks for cancers of the digestive system, respiratory system, ovary, and prostate.¹ We also identified notably reduced risk for stomach and pancreatic cancers. As there were fewer men, this limited precision for estimating risk of less common cancers among men.

An evolutionary hypothesis is that genetic factors protect from cancer but only result in MS after specific triggering events, leading to their persistence in the gene pool. However, the results from parents of those with MS suggest this explanation is unlikely. There was a modest risk increase for cancers of bone, kidney, and lymphoma among fathers, and endocrine cancers among mothers. While these could be chance results (due to a few events), genetic or environmental factors relevant to MS risk in offspring might also explain these findings. A Danish study¹² reported familial clustering of young adult onset Hodgkin lymphoma in first-degree relatives of patients with MS (RR = 1.93), consistent with the paternal lymphoma risk identified here. We cannot rule out a genetic basis for cancer protection due to differences in gene expression between parents and offspring.

The lower cancer risk among patients with MS could be associated with lifestyle changes, treatment, disease-related activity, or a combination of these factors. Behaviors could include smoking; we have no information on smoking, but it is notable that several studies have found that smoking prevalence is modestly increased in MS.^13–15 This suggests the reduction in lung and other cancer risks may exist despite some increase in tobacco smoke exposure. Bladder cancer is also associated with smoking, so its increased risk is in contrast with that for lung cancer. We speculate that the association with urinary bladder cancer is consistent with chronic irritation due to micturition problems and urinary infections experienced by patients with MS. Patients with MS have on average a lower body mass index than the general population¹⁶ and body mass index is a risk factor for several cancer types,¹⁷ so lower body weight may explain some cancer risk reduction. Other behavioral changes or exposures following diagnosis may also be relevant; however, few known behavioral factors have been linked with pancreatic or prostate cancer. We speculate that cancer protection might partly result from the increase in systemic autoimmune responses as against myelin antigens observed among patients with MS.¹⁸ If autoimmune cells such as these react specifically against tumor autoantigens, this could represent an effective tumor defense mechanism. The existence of such mechanisms is well established in paraneoplastic autoimmune phenomena like the Eaton-Lambert syndrome, where anti-channel antibodies cross-react with lung tumors, thus controlling them. Unlike MS, rheumatoid arthritis—another autoimmune disease—is associated with a modest increase in overall cancer risk and for lymphoma and lung cancer but also some decreased risks.¹⁹

The observed increase in brain tumor risk may be due to neurologic inflammation as chronic inflammation is a recognized cancer risk.⁷ Other studies did not identify an increased risk, perhaps due to shorter follow-up time¹ or fewer events.⁴ Surveillance bias, due to frequent neuro-radiologic examinations, may be responsible for this finding as there was no clear excess of malignant brain tumors among patients with MS. We found moderate increased risks for cancers of urinary organs and nonmelanoma skin cancer among women and a more than twofold increased risk of small intestine cancer among men. It is notable that there is no increase in thyroid cancer risk, as there is comorbidity between MS and immune-mediated inflammation of this organ. While immunomodulatory therapy used in MS could potentially increase cancer risk,^2,3,20,21 this is an unlikely explanation for cancer in these specific sites, as immunosuppressive drugs such as azathioprine and cyclophosphamide are used extremely rarely for MS in Sweden. Since approximately 1993, there has been an increase in the use of interferon beta drugs. However, as there was no notable change in cancer risk associated with MS after this timepoint, it does not indicate greater risk due to the introduction of these therapies.

It could be argued that the observed lower overall cancer risk is because patients with MS may have shorter life expectancy than the general population,^22–24 with an average of approximately 5 to 10 years of life lost.²⁵ This is unlikely to be a major factor here as the Cox models estimated age-specific risks and hazard ratios. Masking of cancer symptoms and signs by MS is another unlikely explanation for our results, as those with a chronic disease such as MS are more likely to undergo frequent investigations. Overestimation of cancer risk among patients with MS is more likely, due to potential surveillance bias. Our main analyses were of first cancer diagnosed in an individual, but the results were similar when the analysis was extended to all cancers. As the national Swedish Cancer Register is more than 98% complete,¹¹ this reduces the risk of information bias. As this register began in 1958, earlier diagnoses would be missed: stratification by year of birth (before of after 1958) demonstrated this was not a problem. A potential limitation is the lack of information on behavior, such as drinking, smoking, and exercise. To tackle this issue we were able to adjust for a measure of socioeconomic circumstances (which is strongly associated with many behaviors), as well as for region of residence and period.

The reduced cancer risk associated with age at diagnosis is equivocal as no association was observed for age at symptomatic MS onset. This may be a chance finding or because information on symptomatic onset is less precise and available only for a subset. Another possibility is that age at diagnosis is a proxy marker of disease phenotype or severity, not reflected by onset age.

The diagnostic accuracy of MS could be a concern.²⁶ There is evidence of relatively high diagnostic accuracy for MS among patients included in the Multiple Sclerosis Register. Those with suspected MS are excluded and among a sample, 92% (2,933) had positive test results for CSF oligoclonal bands, consistent with the diagnosis with over 95% specificity after the usual exclusion criteria,²⁷ with similar indication of diagnostic specificity from MRI findings. However, these data were only available for 64% of the patients with MS; they are less frequently available for patients diagnosed in earlier periods, particularly before the mid-1970s.

While there is greater evidence of high diagnostic specificity among patients who entered or remained in the Register during the later period, the temporal pattern of association between MS and cancer risk does not indicate that substantial variation in MS specificity accounts for the results. Our review of 112 randomly chosen patient records further indicates a high degree of diagnostic accuracy of 96% for the MS Register, suggesting that the other signs of diagnostic accuracy for this register are not a function of selection bias. A diagnosis of MS is likely to be less specific for the Inpatient Register and include patients with a suspected rather than confirmed diagnosis, as well as being influenced potentially by other sources of error. Other diseases such as Crohn disease have revealed diagnostic accuracy of around 85%.²⁸ MS diagnoses are likely to be more reliable in the MS Register, consistent with the greater reduction in cancer risk in this subset of patients.

AUTHOR CONTRIBUTIONS

The statistical analyses were conducted by Shahram Bahmanyar and Scott M. Montgomery.