Article Text

PDF

Associations between specific autoimmune diseases and subsequent dementia: retrospective record-linkage cohort study, UK
  1. Clare J Wotton,
  2. Michael J Goldacre
  1. Unit of Health-Care Epidemiology, Nuffield Department of Population Health, University of Oxford, Oxford, UK
  1. Correspondence to Professor Michael J Goldacre, Unit of Health-Care Epidemiology, Nuffield Department of Population Health, University of Oxford, Old Road Campus, Oxford OX3 7LF, UK; michael.goldacre{at}dph.ox.ac.uk

Abstract

Objective To determine whether hospital admission for autoimmune disease is associated with an elevated risk of future admission for dementia.

Methods Retrospective, record-linkage cohort study using national hospital care and mortality administrative data, 1999–2012. Cohorts of people admitted to hospital with a range of autoimmune diseases were constructed, along with a control cohort, and followed forward in time to see if they developed dementia. 1 833 827 people were admitted to hospital with an autoimmune disease; the number of people in cohorts for each autoimmune disease ranged from 1019 people in the Goodpasture's syndrome cohort, to 316 043 people in the rheumatoid arthritis cohort.

Results The rate ratio for dementia after admission for an autoimmune disease, compared with the control cohort, was 1.20 (95% CI 1.19 to 1.21). Where dementia type was specified, the rate ratio was 1.06 (1.04 to 1.08) for Alzheimer's disease and 1.28 (1.26 to 1.31) for vascular dementia. Of 25 autoimmune diseases studied, 18 showed significant positive associations with dementia at p<0.05 (with 14 significant at p<0.001) including Addison's disease (1.48, 1.34 to 1.64), multiple sclerosis (1.97, 1.88 to 2.07), psoriasis (1.29, 1.25 to 1.34) and systemic lupus erythematosus (1.46, 1.32 to 1.61).

Conclusions The associations with vascular dementia may be one component of a broader association between autoimmune diseases and vascular damage. Though findings were significant, effect sizes were small. Clinicians should be aware of the possible coexistence of autoimmune disease and dementia in individuals. Further studies are needed to confirm or refute our findings and to explore possible mechanisms mediating any elevation of risk.

Statistics from Altmetric.com

Introduction

There are suggestions that Alzheimer's disease (AD) may have an autoimmune component, and that autoimmune and inflammatory mechanisms may play a role in the development of dementia.1 ,2 AD has been shown to have a similar female-to-male ratio as that in some autoimmune diseases such as multiple sclerosis (MS) and rheumatoid arthritis,3 and the risk of AD has been shown to be reduced in people who regularly take non-steroidal anti-inflammatory drugs (NSAIDs)4–6—drugs used to treat diseases such as rheumatoid arthritis—although findings from a recent systematic review were mixed.7 A relationship between levels of antithyroid antibodies and AD has been reported,8 and studies have found that in people with early-onset dementia (those diagnosed when aged under 65), MS accounts for the dementia in 3–4% of cases.9 ,10 In a recent study, we report an association between admission to hospital with type 1 diabetes (ie, diabetes with an autoimmune component) and subsequent admission for dementia.11

We have used national Hospital Episode Statistics (HES) for England to construct retrospective cohorts of people admitted to hospital with a range of autoimmune diseases, in order to follow them forward in time for a hospital admission record of dementia. The objectives of the study were to determine whether hospital admission with an autoimmune disease is associated with future admission with dementia more often than expected by chance.

Methods

Population and data

We undertook a retrospective cohort analysis using a data set of English national HES in which successive episodes of care for each person were linked to each other, and linked to a data set of all deaths in England, during the period 1 April 1998 to 31 March 2012.12 The HES data include clinical, demographic and administrative information on all hospital admissions, including day cases (people admitted who do not stay overnight), in all National Health Service (NHS) hospitals in England. The great majority of medical admissions in England are in NHS rather than private hospitals. The HES data were supplied by the English national Health and Social Care Information Centre; and the linked data set (filename da_cips_14yr_v01b_alluhcert) was built by the Oxford record-linkage group using, as identifiers, the encrypted values of the NHS number and of their HES number (both unique to each individual). Mortality data derive from death registration records, and were supplied by the Office for National Statistics, with identifiers encrypted in the same way. The methods used were similar to those described in previous studies.11 ,13 ,14 The same methods were used for each autoimmune disease followed by an admission for dementia, and are described here using MS as the example. Analyses by dementia subtype, age at first hospital admission for dementia and by sex were carried out in the same way.

Exposure cohort definition (for each autoimmune disease, described using MS as the example)

A cohort of people with MS was constructed by identifying each person's first recorded episode of day case care or inpatient admission for MS coded in any diagnostic field on the hospital discharge abstract, that is, coded as either the primary reason for admission or as a secondary diagnosis. The International Classification of Disease (ICD) code, 10th revision, used for MS was G35; codes used for each of the other autoimmune diseases are given in table footnotes.

Control cohort definition

A control cohort was constructed by identifying the first admission for each individual with various other medical and surgical conditions and injuries as the primary discharge diagnoses (listed in table 2 footnotes; primary diagnosis selected to avoid including people whose primary diagnosis was a more serious condition). This is based on a ‘control group’ of conditions that has been used in other studies of associations between diseases.11 ,14 Standard epidemiological practice was followed, in using hospital controls, of selecting a very diverse range of clinical conditions rather than relying on a narrow range (in case any of the latter are themselves atypical in their risk of dementia).

Outcome definition: dementia

We searched the data set for any subsequent NHS hospital care for, or death from, dementia in the MS and the control cohorts. The ICD codes used for dementia in the 9th and 10th revisions were 290 and 331.0 in ICD9, and F00-F01, F03 and G30 in ICD10 (ICD9 was used for deaths in 1999 and 2000, and ICD10 from 2001; ICD10 was used for hospital discharge data throughout the period covered by this study). We considered dementia cases as those coded in any diagnostic field on the discharge abstract. Codes for dementia secondary to other conditions were not included in our analyses of dementia, to exclude possible associations with those other conditions. We subdivided people with an admission for dementia into those who were specifically recorded as having an admission for AD (331.0 in ICD9 and F00 and G30 in ICD10) and those recorded as having an admission for vascular dementia (290.4 in ICD9 and F01 in ICD10). People were included in the MS cohort, or in the control cohort, if they did not have an admission for dementia either before or at the same time as the admission for MS or control condition. Additionally, we analysed the data separately for men and women and according to time interval between the first known record of autoimmune disease admission and the first known dementia admission to obtain information on temporality.

Statistical analysis

In comparing the MS cohort with the control cohort, we stratified and then standardised by age (in 5-year age groups), sex, calendar year of first recorded admission, region of residence and quintile of patients' Index of Deprivation score (a standard measure of socioeconomic status used in England). Statistical methods are given in detail elsewhere.13 All calculations were performed, comparing the cohorts, within the individual strata (eg, within 5-year age groups, single year of hospital admission, etc). The direct method of standardisation was used with the combined MS and control cohort as the standard population; and the stratum-specific rates were then applied to the number of people in each stratum within, first, the MS cohort and then, separately, the control cohort. This calculation gave the expected number of cases of dementia in each stratum in, respectively, the MS and control cohorts. The observed and expected numbers within each stratum were then summed to give totals in the broader age groups shown in the tables and then at all ages. Results were expressed as stratum-standardised rate ratios (RRs) comparing the MS cohort with the control cohort, using the formula Oms/Ems divided by Ocon/Econ, where O and E are the observed and expected numbers of dementia in the MS and control cohorts, respectively. The calculation of expected values, the RRs, their CIs and p values used standard, published statistical methods.15

Ethical approval for our multipurpose programme of work linking and analysing routine medical data sets was obtained from the Central and South Bristol Multi-Centre Research Ethics Committee (04/Q2006/176).

Results

In total, 1 833 827 people were admitted to hospital with an autoimmune disease. The number of people in the cohorts for each specific autoimmune disease ranged from 1019 people in the Goodpasture's syndrome cohort, to 316 043 people in the rheumatoid arthritis cohort. Around seven million people were included in the control cohort; numbers varied slightly depending on each specific autoimmune disease being studied (see exclusion criteria when dementia was recorded on the exposure or control record prior to, or at the same time as, the autoimmune disease or control condition, in the Methods section).

Table 1 shows the mean age at entry, the mean period of follow-up and the percentage of men in the cohort for each of the autoimmune diseases.

Table 1

Mean age at entry into cohort, period of follow-up and percentage males for each autoimmune disease

Dementia risk in people with an autoimmune disease

Overall, people admitted to hospital with an autoimmune disease were 20% more likely to have a subsequent admission for dementia than those without an admission for an autoimmune disease (RR 1.20; 95% CI 1.19 to 1.21; table 2). Of the 25 autoimmune diseases studied, 18 showed significant positive associations with dementia at p<0.05 (with 14 significant at p<0.001) including conditions as diverse as Addison's disease (RR 1.48, 95% CI 1.34 to 1.64), MS (1.97, 1.88 to 2.07), polyarteritis nodosa (1.43, 1.12 to 1.82), psoriasis (1.29, 1.25 to 1.34), systemic lupus erythematosus (1.46, 1.32 to 1.61) and thyrotoxicosis (1.31, 1.27 to 1.34). Though significant, effect sizes were small. To ensure that the diagnosis of dementia was not the result of an admission soon after the autoimmune disease admission, as a result of its incidental diagnosis in the latter (ie, to guard against surveillance bias), we additionally examined the data after the exclusion of cases of dementia that were first recorded within a year of the first specific autoimmune disease admission. There were no notable differences in the RRs for dementia whether or not the first year cases were included. Accordingly, the main results shown are for all cases.

Table 2

Occurrence of dementia in people admitted to hospital after an admission for an autoimmune disease

The type of dementia was often not recorded: of a total of 81 502 people with an autoimmune disease and dementia, 20 032 had a record of AD and 22 536 of vascular dementia. The RR for AD in people admitted with an autoimmune disease was 1.06 (1.04 to 1.08) and that for vascular dementia was 1.28 (1.26 to 1.31). RRs for vascular dementia after individual autoimmune disease were, on the whole, slightly higher than those for AD. The risk of vascular dementia, but not AD, was elevated in association with hospital admission for idiopathic thrombocytopenia purpura, pemphigus, polyarteritis nodosa, scleroderma, Sjogren's syndrome and systemic lupus erythematosus. The risk of AD was elevated in association with hospital admission for Addison's disease, myxoedema, pemphigoid, pernicious anaemia, psoriasis and thyrotoxicosis (table 2). Although there was an elevated risk of vascular dementia after an admission for rheumatoid arthritis, the RR for AD was actually significantly low in people following an admission for rheumatoid arthritis (0.89, 0.86 to 0.93).

Men and women

Autoimmune diseases are generally more common in women than men (tables 1 and 3). We analysed data for men and women separately, as well as together. For all autoimmune diseases combined, the RR was slightly but significantly higher in men as evidenced by the non-overlapping CIs (1.32; 1.30 to 1.35, p<0.001) than women (1.17; 1.16 to 1.18, p<0.001). The excess risk of dementia was significantly higher in men with MS than in women (RR 2.52; 2.32 to 2.74 in men vs 1.79; 1.69 to 1.89 in women). For most of the individual autoimmune diseases, RRs for dementia were broadly similar for men and women.

Table 3

Occurrence of dementia in men and women admitted to hospital after an admission for an autoimmune disease

Time intervals

It is difficult to be sure of the temporal sequence of autoimmune disease and dementia in this type of study. To try to address this, we studied the time intervals between the first recorded admission for the autoimmune disease and the first recorded admission for dementia (table 4). Most of the associations remained significant 5 or more years after admission for the autoimmune disease, although associations were generally stronger with shorter time intervals.

Table 4

Time intervals between autoimmune disease admission and subsequent admission for any type of dementia, for selected associations

Discussion

Main findings and comparisons with other studies

There have been suggestions that autoimmune and inflammatory mechanisms may play a role in the development of dementia,2 and there is evidence that people with AD may have higher levels of antithyroid autoantibodies when compared with people without dementia.8 However, there is little epidemiological data on whether autoimmune disease and dementia are associated. We recently reported that hospitalisation with type 1 diabetes was associated with subsequent hospitalisation with dementia,11 but further evidence of a link between autoimmune disease and dementia is lacking. Here, we report that people admitted to hospital with autoimmune disease were more likely to be admitted to hospital with dementia, at a later date, than a control group. Our findings for thyroid autoimmune disease and rheumatoid arthritis are consistent with previous data4 ,8 and there is little else published with which to compare our findings.

The risk of vascular dementia after an admission with an autoimmune disease was slightly higher than the risk of AD; and some individual autoimmune diseases were significantly associated with an elevated risk of future vascular dementia, but not AD. The associations with vascular dementia might reflect associations between autoimmune disease and risk factors for cardiovascular and cerebrovascular diseases more generally. We have undertaken similar analyses to those reported here for coronary heart disease (CHD) and ischaemic stroke (which we will report in detail elsewhere). In brief, the RRs for CHD and ischaemic stroke in people with autoimmune disease (all those in table 1 combined) were, respectively, 1.53 (1.52 to 1.54) and 1.46 (1.44 to 1.47).

A previous admission with rheumatoid arthritis seemed to protect against AD, but elevated the risk of vascular dementia. Although evidence that NSAIDs reduce the risk of AD is not conclusive,4–7 our findings might add circumstantial evidence to the data supporting the hypothesis that NSAIDs protect against AD. People with rheumatoid arthritis generally take NSAIDs to manage their condition, so if rheumatoid arthritis is at least partially a proxy for NSAID use, the real association may be between NSAID use and a reduced risk of AD. Although one could argue that the association may actually be between rheumatoid arthritis and AD, one study did find that the association between NSAID use and AD did not appear to be confounded by arthritis.6 We do not have data on potential confounders (other than basic demographics like age, socioeconomic status and region of residence), so cannot explore the association further. A recent case–control study found that NSAID use was positively associated with vascular dementia (OR 1.33, 1.29 to 1.38).16 Thus, it seems possible that NSAID use may reduce the risk of AD but increase the risk of vascular dementia. If there is a protective effect of NSAIDs on some autoimmune diseases, and if its mechanism of action is anti-inflammatory, then steroids may also play a protective role against AD. NSAIDs and steroids are prescribed for several of the autoimmune diseases we studied, but we did not find any other significantly negative findings. However, some of the RRs were <1, albeit non-significant, so further negative associations cannot be ruled out.

Strengths and limitations

The main strengths of the study are that it is very large in size, and allowed us to study the risk of dementia, both overall, and for AD and vascular dementia separately, following a wide range of autoimmune diseases, in a single, geographically defined population.

Limitations include the fact that the study is restricted to people who were admitted to hospital or who received day case care, and the lack of data on confounders. Effect sizes found in our study are small. Factors which may have led to underestimation of rates include under-recording of autoimmune diseases and of dementia, when present, with dilution of effects. Factors that may have led to overestimation of risks include unmeasured confounding such that, for example, people with an autoimmune disease and also with subsequent dementia might be more likely to be admitted than others if they are socioeconomically disadvantaged. In mitigation, we standardised for deprivation; but the limits of the deprivation data may be such that the standardisation was too simple. Also, the factors that determined whether or not a person with dementia was admitted to hospital should be the same in the control cohort as in each autoimmune disease cohort.

The autoimmune disease cohorts are based on prevalent cases, the first recorded hospital admission or episode of day case care, rather than being cohorts with follow-up from the date of diagnosis. The diagnoses of the conditions studied are confined to those recorded on the hospital discharge summary for each person, and, due to privacy regulations, we are unable to examine the records to ascertain their accuracy. Confining data collection to hospital admissions and episodes of day case care will mean that we miss cases of autoimmune disease and dementia that did not require hospital day case care or admission. Dementia subgroups, in particular, are not well coded on hospital records, indeed are often not coded at all. It is possible that hospital coding of dementia subtypes is subject to some bias and that this may affect our findings. The separate analyses of AD and vascular dementia, therefore, should be viewed with some caution; but we suggest that the results with dementia as a whole have fewer limitations, although there is still likely to be some underdiagnosis. The likelihood is that cases that really are AD are underestimated as AD, coded as such in HES; and that, accordingly, unspecified dementia in HES actually contains a higher percentage of AD than that of vascular dementia. Because of the nature of the data, some misdiagnosis and miscoding are inevitable. However, it is worth noting that any lack of diagnostic specificity of dementia type will affect both cases and controls alike and should not distort the RRs themselves.

We studied a large number of associations between autoimmune diseases and subsequent dementia. The effect of making multiple comparisons needs to be considered. For this reason, we give exact p values in the tables, as well as CIs, so that the reader can see the degree of significance of each association. It is possible that some of the associations that are significant at a level of p<0.05 or even p<0.01 may result from making multiple comparisons and the play of chance. On the other hand, even with the number of comparisons that we have made, findings with a significance level <0.001 are unlikely to be attributable to chance alone. Further, 18 of the 25 disease–dementia associations were significant: by chance alone, only one or two would be significant at p=0.05.

The list of autoimmune diseases chosen is not exhaustive but it includes diseases for which there is acceptance of a definite autoimmune component and which are common, chronic and longstanding; and which have been included by us in other studies of the epidemiology of autoimmune diseases.17 ,18 There is scope for the study of other diseases, not covered by us, which are considered to have an autoimmune component.

Possible mechanisms

Our study design cannot be used to investigate mechanisms of association. Diseases may occur in combination more often than expected by chance because one (the ‘exposure’ disease) predisposes to another (the ‘outcome’ disease); or because they share environmental and/or genetic mechanisms in common; because a specific treatment used in the exposure disease alters the risk of the outcome disease; or because the exposure disease causes changes in risk factors for pathology of which the outcome disease is just a part. Considering the latter, a speculative example is that autoimmune diseases (or their treatment) may increase the risk of circulatory disease, generally, of which vascular dementia is one component in some people.

Conclusions

For reasons given above, our findings should be considered as indicative rather than definitive. However, the negative association between rheumatoid arthritis and AD reflects the literature on NSAID use and AD risk, and gives some face validity to our findings. People admitted to hospital with an autoimmune disease, likely to be those at the severe end of the disease spectrum, do appear to have an elevated risk of dementia. This finding is consistent with autoimmune disease predisposing to vascular risk and vascular dementia. It is also, separately, consistent with the theory that AD may have an autoimmune component. If our findings are confirmed in other studies, clinicians and epidemiologists will wish to know that some people with some autoimmune diseases have an elevated risk of dementia. The effect size, at least in our study, is small. The findings may be relevant to furthering understanding of the pathogenesis of AD and of vascular dementia. Studies should be considered to confirm or refute our findings; and, if confirmed, to explore possible mechanisms mediating the associations.

What is already known on this subject

  • It has been suggested that autoimmune mechanisms may play a role in the development of dementia. We recently reported an association between type 1 diabetes and dementia, but there is no published evidence on other associations between autoimmune diseases and dementia.

What this study adds

  • We found an association between hospital admission for an autoimmune disease and admission for dementia. Many of the individual autoimmune diseases studied were associated with subsequent dementia; however effect sizes were small. Clinicians should be aware of the potential coexistence of these diseases in individuals.

Acknowledgments

Over many years, the linked data sets were built by Leicester Gill and Matt Davidson, Unit of Health-Care Epidemiology, University of Oxford.

References

View Abstract

Footnotes

  • Contributors CJW proposed the study, analysed the data and wrote the first draft. MJG is the guarantor of, and designed, the study. Both authors contributed to the interpretation of the data and revision of the manuscript for important intellectual content.

  • Funding The Unit of Health-Care Epidemiology was funded by the English National Institute for Health Research to build the linked data set.

  • Disclaimer The views expressed in this paper do not necessarily reflect those of the funding body.

  • Competing interests None declared.

  • Ethics approval Ethical approval for analysis of the record-linkage study data was obtained from the Central and South Bristol Multi-Centre Research Ethics Committee (04/Q2006/176).

  • Provenance and peer review Not commissioned; externally peer reviewed.

Request permissions

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.