Autoimmune thyroiditis as a risk factor for stroke

DISCUSSION

Our study provides strong evidence for a slightly increased risk of stroke in patients with AIT, particularly in the first year after AIT diagnosis. Our analyses indicated that people with AIT were more likely to develop hyperlipidemia, CHD, and CHF, adding evidence to previous discussions about whether AIT is associated with these factors, and that some, but not all, of the increased risk of stroke among patients with AIT was mediated via these classic cardiovascular risk factors.¹⁴

Few previous studies have focused on potential associations between hypothyroidism or AIT and stroke. Our systematic review identified 7 studies, which showed effect sizes ranging from 0.8 to 1.6 (table e-2). All of these studies except for the Scottish study from 2006 were small and had effect estimates with wide CIs that overlapped with those from the present study. Other methodologic limitations of previous studies included overadjustment for variables that could be mediators of increased stroke risk, and residual confounding (for example, the Scottish study lacked information on smoking, alcohol use, and medication history).^{7,12,22,–,26} Also, studies included individuals with a history of stroke. This could explain their slightly higher effect estimates, because thyroid hormone status is often assessed at stroke units, resulting in better AIT ascertainment in individuals with a stroke history (who are at higher risk of a subsequent stroke). Three of the studies were conceptually different from this study because they used asymptomatic patients identified as hypothyroid by laboratory tests who were mostly untreated with thyroxine.^7,12,25

We found differences between individuals with and without AIT in the prevalence of traditional cardiovascular risk factors at baseline. Individuals with AIT were more likely to have AF and hypertension at baseline compared to those without AIT; however, they were not at increased risk of developing these conditions during follow-up. This finding is consistent with that from a recent population-based study from Germany, which reported that hypothyroidism was associated with prevalent but not incident hypertension.³³ Hypertension and AF may not be on the causal pathway between AIT and stroke, although the differences we observed at baseline could represent early changes induced by undiagnosed AIT. Differences in AF (and possibly hypertension) at baseline between individuals with AIT and those without could also be explained by differences in ascertainment of AIT, if individuals with AF were more likely to have thyroid-stimulating hormone levels measured.

Our main model of interest, adjusted for these baseline variables, is based on the assumption that pathologic mechanisms on the causal pathway to stroke did not start until after AIT diagnosis and initiation of thyroxine treatment. Thus, these factors were treated as potential confounders. However, it is likely that the date of AIT diagnosis/start of treatment did not accurately capture start of disease. Given that patients with AIT often present with nonspecific symptoms, diagnosis of AIT can be delayed. Thus, it is plausible that cardiovascular causal processes could have started before formal diagnosis of AIT. If so, model 3 (which adjusted for these variables) may have provided a conservative estimate of relative stroke risk, and model 2 might provide an estimate closer to the true effect. However, the effect estimates produced by these 2 models were similar; also, the sensitivity analysis excluding all individuals with cardiovascular disease–related conditions at baseline showed results similar to the main analysis.

Our finding of a higher effect of AIT on stroke risk in the first year after diagnosis is compatible with the hypothesis of increased cerebrovascular risk in patients with AIT due to prediagnosis hypothyroidism and the time taken to reduce this risk after thyroid hormone replacement. There was no evidence for an increasing effect with increased length of follow-up (a possible proxy for a long-term effect mediated by autoimmune processes). An alternative explanation is that the observed effect of AIT on stroke was caused by thyroxine treatment because thyroxine can have a procoagulant effect.³⁴ However, if the observed effect was due to thyroxine, we might expect to have seen the excess stroke risk restricted to the period soon after start of treatment, whereas the risk remained increased 6 to 12 months after diagnosis in our study. Moreover, a thyroxine-mediated effect might also be expected to generate short-term differences in incidence of AF between patients with AIT and individuals without AIT because thyroxine in high doses can induce AF; this was not the case. Nevertheless, we cannot exclude that some of the effect attributed to hypothyroidism could have been due to thyroxine, and we were unable to examine this formally because all patients with AIT received thyroxine.

This study has several strengths. We used a considerably larger study population than all previous studies, and thus had greater power to ascertain effect estimates. We had information on a variety of potential confounding factors that could be considered in the analysis. Unlike previous studies, we excluded individuals with a history of stroke, avoiding introduction of bias. Moreover, excluding people with other autoimmune diseases from our study population enabled us to observe the effect of AIT unmodified by other autoimmune disorders. This might not have been the case in other studies, because patients with AIT are at higher risk of acquiring other autoimmune disorders, some of which (e.g., rheumatoid arthritis, type 1 diabetes) are associated with considerably increased risk of cardio- and cerebrovascular disease.^16,35

Potential limitations of our study also need consideration. Almost all noniatrogenic hypothyroidism in the United Kingdom is due to AIT, and we excluded individuals who had other reasons for hypothyroidism. However, our Read code list included nonspecific hypothyroidism codes, which increased sensitivity of AIT diagnosis but could have included a few non-AIT cases. Because hypothyroidism can remain undiagnosed for an unspecified time, some patients with AIT could have been misclassified as unexposed. This is likely to have happened largely independently of subsequent stroke risk, and thus may have driven the RR toward 1.0. Those with preexisting AF could have been more likely to have their AIT diagnosed, which could have resulted in some overestimation of the effect size of AIT on the risk of stroke. However, our sensitivity analyses excluding individuals with AF at baseline indicate that this is unlikely to have been a major problem.

Differential misclassification of the outcome was unlikely in this study because stroke is a serious condition and should have been recorded equally in both groups independent of AIT status, or health-seeking behavior. However, we could not differentiate between cases of ischemic and hemorrhagic stroke, because 85% of stroke codes did not specify the stroke subtype, consistent with previous studies of stroke using electronic health data.^23,36 We hypothesized that AIT might be associated specifically with ischemic stroke because hypothyroidism is linked with classic risk factors for ischemia.^{3,15,–,18,20,21,37} Moreover, carotid artery intima-media thickness and AF, 2 large risk factors for ischemic stroke, have been linked to AIT.^22,23,27 Ischemic stroke causes approximately 80% of stroke cases in adults, and effect sizes for TIA (an ischemic process) were comparable to those shown for stroke in our study, further supporting that AIT is associated with ischemic stroke. Nevertheless, AIT could also affect the risk of hemorrhagic stroke, because hypothyroidism has been shown to be associated with the development of von Willebrand syndrome and bleeding events.³⁸

Individuals with AIT had higher consultation rates than those without AIT. It is therefore possible that they were more likely to be diagnosed and treated for conditions such as hyperlipidemia or hypertension. Monitoring and prompt treatment of these conditions could have decreased stroke risk associated with these conditions and might partly explain the relatively small increase in stroke risk associated with AIT. Nevertheless, this study provides a realistic picture of the residual effect of AIT after standard treatment and surveillance in the United Kingdom. Data concerning the potential confounders of ethnicity and socioeconomic status were not available. However, socioeconomic status was partly accounted for by allowing for clustering by practice.

Although THIN has been shown to be representative for the UK population, the results of this study might not be generalizable to all patients with AIT. We included only patients with AIT who received thyroxine treatment. This excluded patients who had never had hypothyroidism. If hypothyroidism is on the causal pathway between AIT and stroke, the observed effect relates to treated rather than all patients with AIT. Excluding individuals with AIT who had a second autoimmune disorder allowed us to pinpoint the specific effect of AIT, but may have excluded some patients with AIT who had stronger autoimmune processes and thus might not have captured all of the long-term risk of stroke mediated by autoimmune pathology.

Our study has demonstrated a 10% to 14% increased risk of stroke among patients with hypothyroid AIT after extensive consideration of confounding and potential biases. Given the relatively high prevalence of AIT and the morbidity and mortality associated with stroke, even small increases in stroke risk might be of high clinical relevance. The results of our study highlight the potential for regular screening for cardiovascular risk factors and preventive treatment (e.g., with statins) in patients with AIT.