Population and data

The Oxford Record Linkage Study (ORLS) [9] includes brief statistical abstracts of records of all hospital admissions (including day cases) in National Health Service (NHS) hospitals, and all deaths regardless of where they occurred, in defined populations within the former Oxford NHS Region. The original ORLS covered the years 1963 to 1998, and is referred to here as ORLS1. The hospital data were collected routinely in the NHS as the region's hospital statistics system and were similar to English national hospital episode statistics (HES). A second dataset, referred to as ORLS2, has been linked and built as the Oxford regional subset of the English national Hospital Episode Statistics (HES), and runs from 1999 to 2008. The data items available for linkage changed between 1998 and 1999 and the two datasets are not themselves linked together. We have also used the complete dataset of the English national HES (1999 to 2008). Death data in each of the three datasets derive from death certificates. The population covered by ORLS1 gradually expanded over time from an initial population of part of Oxfordshire (an approximately 300,000 resident population) to all four counties of the former Oxford NHS region (resident population 2.5 million); and the ORLS2 covers the same four counties. The population of England is 50 million. The datasets used in this study (versions m6v2 for ORLS and v08a for ORLS2 and England) have been constructed by staff in the Oxford Unit of Healthcare-Epidemiology.

People were included in the rheumatoid arthritis or reference cohort if they did not have an admission for VTE either before or at the same time as the admission for rheumatoid arthritis or the reference condition. We then searched the database for any subsequent NHS hospital care for, or death from, VTE in these cohorts. We considered that rates of VTE in the reference cohort would approximate those in the general population of the region while allowing for migration in and out of it (data on migration of individuals were not available).

Statistical methods

We calculated rates of VTE based on cohort analysis. For each broad age group of people with rheumatoid arthritis, and for VTE, we took "date of entry" into each cohort as the date of first admission for rheumatoid arthritis, or reference condition, and "date of exit" as the date of first record of VTE, death, or the end of the data file (31 December 1998 for ORLS1; 31 March 2008 for ORLS2 and English HES), whichever was the earliest. In the ORLS cohorts, in comparing the rheumatoid arthritis cohort with the reference cohort, we first calculated rates for VTE, stratified and then standardised by age (in five-year age groups), sex, calendar year of first recorded admission, and district of residence, to ensure that the results of group comparisons were equivalent in these respects. We used a similar approach to standardisation in the England dataset, stratifying by age (in five-year age groups), sex, calendar year of first recorded admission, region of residence, and quintile of patients' Index of Deprivation score (as a measure of socio-economic status). We used the indirect method of standardisation, using the combined rheumatoid arthritis and reference cohorts as the standard population. We applied the stratum-specific rates in the combined rheumatoid arthritis and reference cohorts to the number of people in each stratum in the rheumatoid arthritis cohort, separately, and then to those in the reference cohort. We calculated the ratio of the standardised rate of occurrence of VTE in the exposure cohorts relative to that in the reference cohort. The confidence interval for the rate ratio and χ² statistics for its significance were calculated as described elsewhere [13]. Calculated in this way, the rate ratio provides a measure of relative risk of VTE in the rheumatoid arthritis cohort, compared with the reference cohort. The fact that there is unmeasured migration in the populations covered by the study, and the use of an internal reference cohort for comparison with the VTE rates in the rheumatoid arthritis cohort (and others), preclude meaningful calculation of absolute risks.

We analysed the occurrence of an admission for VTE within 90 days of the admission for each immune-related disease, and at 91 days and more, to help establish whether any elevated risk of VTE was confined to the short-term after the episode of hospitalisation or was more prolonged.

In the analysis of diabetes mellitus, we used hospital admission for diabetes mellitus when aged under 30 as a proxy for type 1 diabetes, as the type of diabetes is not well recorded in routine hospital statistics. We analysed the data for males and females separately, as well as together, to ascertain whether or not there were differences between them in risk of VTE.

Males and females

On the whole, the rate ratios for males and females were similar although some of the numbers in ORLS1 and ORLS2 became rather small when subdivided by sex. In the England dataset, there was a significant elevation of VTE risk in males with coeliac disease (RR 1.47; 1.26 to1.70), but not in females (1.09; 0.97 to 1.23). Similarly, the VTE rate ratio for males with Hashimoto's thyroiditis was significantly elevated (2.98; 1.86 to 4.51), but the rate ratio in females was not (1.23; 0.97 to 1.54). There were no other significant differences between males and females.

Short (0 to 90 days) and long-term (91+ days) associations with VTE

Absolute risk

Although we could not calculate meaningful absolute risks for the ORLS1 and ORLS2 cohorts, because of migration into and out of the Oxford region, approximate absolute risks can be calculated in the English national data. For example, there were 81,950 people in the MS cohort of whom 1,509 had an admission for VTE (1.8%) during the 10-year period covered by the study. As an approximation, in people with MS there was one case of VTE per 264 person-years at risk. Similar calculations show that 2.7% of people with SLE had an admission for VTE in the period covered by the study, and that the event rate was one case of VTE per 169 person-years at risk. These calculations assume that migration was small and that its effect was unimportant.

Methodological issues: datasets, population, and multiple comparisons

Strengths of the datasets include their size, with large numbers of fairly uncommon diseases. The ORLS1 data provide long duration of follow-up; the English data provide a much larger and more recent population but with shorter follow-up. The risk of VTE was therefore studied for a large number of immune-mediated diseases, all within a single population and using the same methodology. Accordingly, levels of risk associated with different immune-mediated diseases can be directly compared within the same study populations.

Our study should be regarded as exploratory rather than definitive. We studied a wide range of immune-mediated diseases with differing aetiologies. We did so because, as an exploratory study, with very large linked datasets in which many diseases can be studied, we saw no reason to be restrictive in our selection of diseases.

The datasets have limitations. The cohorts are based on prevalent cases, the first recorded hospital admission or episode of day case care for each person with each condition, rather than being cohorts with follow-up from the date of first diagnosis. Data are not recorded on patients who move out of the area covered by data collection or who are treated in hospitals outside the area (mainly affecting ORLS). The datasets are limited to people who were admitted to hospital, or who received day case specialist care. This would not capture all people with each immune-mediated disease, although it should identify the great majority with subsequent diagnosed VTE. These factors are part of our reasoning for including a comparison cohort of patients admitted to hospital, or in receipt of day case care, from the same database and for 'matching', through stratified analysis, for area of treatment and for year of first recorded diagnosis as well as for age and sex. The two Oxford datasets are not linked to each other due to changes in the data items available for linkage between 1998 and 1999. Consequently, it is likely that some people have been recorded as having a 'first admission' for each immune-mediated disease in each of the time periods studied.

We lack clinical and laboratory data. We lack treatment data for the immune-mediated diseases; and elements of their treatment could themselves influence VTE risk. There is very limited information on potential confounding factors such as socioeconomic status, and none on smoking or ethnicity. As we comment above, our results should be regarded as speculative rather than definitive: they represent results from what can be done using very large-scale, routinely collected administrative data. They need further work, in different study designs, to confirm or refute the findings, although epidemiological studies involving direct patient contact may be quite formidable logistical undertakings on the scale required.

Our rate ratio for VTE in people admitted to hospital with rheumatoid arthritis in the English dataset was 1.75, which is comparable with a previous study that reported a relative risk of 1.99 [6]. Our rate ratio for VTE in people aged under 30 with diabetes was 2.58 in the English dataset, which compares with a reported figure of 1.73 in hospitalised patients with diabetes aged 20 to 29 in the USA [7]. Although the literature is sparse on VTE in people with immune-mediated diseases, our findings seem broadly comparable with findings of others.

We used age at admission under 30 as a proxy for type 1 diabetes mellitus, as the type of diabetes is not routinely recorded on hospital admission records. Although this will mostly consist of people with type 1 diabetes, there may be a few people with type 2 diabetes in the cohort as well.

We studied a large number of associations between diseases. The effect of making multiple comparisons needs to be considered. For this reason, we have given exact P-values, as well as confidence intervals, so that the reader can judge the degree of significance of each immune-mediated disease and subsequent VTE. It is possible that some of the associations that are significant at a level of P < 0.05 or P < 0.01 may result from making multiple comparisons and the play of chance. This may particularly be so where there is no prior hypothesis to support the finding. On the other hand, even in a study with the number of comparisons that we have made, findings where the significance level is <0.001 or less, are unlikely to be attributable to chance alone. There were differences in levels of risk between the diseases studied: for example, the rate ratios for SLE were significantly and substantially higher than those for myxoedema. Even if the fairly low levels of elevation of risk, such as those associated with myxoedema and thyrotoxicosis, are considered unimportant, the high levels of risks associated with SLE, polyarteritis nodosa and diabetes mellitus are striking. The large numbers of findings with highly significant results in the England cohort needs comment. They no doubt in part reflect the very large numbers of patients in the cohort, such that many differences are significant, even if fairly small (for example, coeliac disease, thyroiditis, myxoedema, thyrotoxicosis), as a result of high statistical power.