No association between HIV disease and its treatment and thyroid function

Introduction

HIV disease has long been associated with endocrine dysfunction. In the pre-highly active antiretroviral therapy (HAART) era, this was primarily associated with opportunistic infections such as cytomegalovirus and tuberculosis. In the era of HAART, a more complex situation has developed, with many patients experiencing insulin resistance, diabetes, sex hormone abnormalities and osteoporosis, with no unifying mechanism established for these conditions. Recently, thyroid dysfunction has been reported [1–7], with a reported increased prevalence of abnormal thyroid function tests, in particular subclinical hypothyroidism in both adults and children on HAART [1,2,4].

Recent population (non-HIV)-based studies have revealed ethnic differences in the prevalence of thyroid dysfunction [8–10]. In iodine-replete communities, autoimmune disease is the most common cause of both hypothyroidism and hyperthyroidism. In women the prevalence of thyrotoxicosis is 0.5–2% and that of hypothyroidism is 1–2%, both conditions being 10 times more common in women than in men [11]. In the USA, the National Health and Nutrition Examination Survey (NHANES) 111 study revealed a significantly lower prevalence of subclinical hypothyroidism amongst black men and women compared with white individuals [10]. A study from South Africa suggested that black subjects were 10 times less likely to develop thyrotoxicosis than white subjects [9]. These data are relevant when investigating the factors associated with thyroid dysfunction in our heterogeneous HIV-positive population.

The aims of this study were as follows.

Methods

Patients included in our study were attendees at the out-patient clinic for HIV-positive individuals at the Royal Free Hospital, London, UK. Details on data collection methods for the cohort are given elsewhere [12]. The clinic database currently contains data up to August 2004. We additionally obtained the results and dates of all thyroid function test (TFT) measurements carried out from August 2001 onwards from the clinical biochemistry database of the hospital. Since this date, any new patients and established clinic patients have been offered annual screening for tests including TFT, syphilis serology and hepatitis A, B and C virus antibodies. Additional data regarding the prescribing of thyroxine or carbimazole for treatment of overt hypothyroidism and overt hyperthyroidism were obtained from pharmacy records.

Inclusion criteria were patients who had at least one serum thyroid-stimulating hormone (TSH) and free thyroxine (FT4) measurement recorded from August 2001 onwards. TSH and FT4 were measured using Modular E170 Elecsys methods (Roche Diagnostics, Sussex, UK). Assay precision across the measurement range was TSH [CV% (coefficient of variation which is the standard deviation/mean expressed as a percentage)] 0.46 mU/L (CV=1.6%) 4.7 mU/L (CV=1.5%) 28.8 mU/L (CV=1.8%); free T4 (FT4), 8.8 pmol/L (CV=4.2%) 21.5 pmol/L (CV=4.1%) 82.1 pmol/L (CV=8.2%). Each patient could have more than one measurement taken. Follow-up began on the date of the first TFT measurement and ceased on the date of the last TFT measurement. All the results were included in the analysis, and patients were classified by biochemical definition, according to whether they had ever had an abnormal result. Longitudinal results were also obtained by comparing the first and second tests.

A patient was considered to be overtly hypothyroid if, on at least one occasion during the follow-up period, a serum TSH measurement were recorded of >4.2 mU/L and a FT4 measurement were recorded of <12 pmol/L. A patient was also considered to have overt hypothyroidism if pharmacy returns indicated that the patient had received thyroxine treatment for hypothyroidism in the current year. Similarly, a patient was considered to be overtly hyperthyroid if, on at least one occasion during follow up, they had a TSH measurement of <0.02 mU/L and a FT4 measurement of >22 pmol/L, recorded or if pharmacy returns indicated a prescription for carbimazole for hyperthyroidism.

Individuals without overt hypothyroidism or overt hyperthyroidism were further categorized according to the following criteria.

All other individuals were assumed to have normal TFTs.

The above criteria were used to obtain an estimate of the ‘period’ prevalence of thyroid dysfunction in our population. Factors associated with the presence of overt hypothyroidism were investigated using the Mann–Whitney U-test, the χ² test or Fisher's exact test as appropriate. Those factors that were significant at the 5% level in a univariate analysis were included in a multivariable logistic regression model to further investigate which factors were independently associated with overt hypothyroidism. As the number of patients was low, investigation of the factors associated with overt hyperthyroidism was not possible. Analyses were carried out using SAS version 8.2 (SAS Institute Inc., Cary, NC, USA).

Results

From August 2001 onwards, time-point data on thyroid function tests was routinely collected and 2151 HIV-positive individuals visited the out-patient clinic. Of these 2151 individuals, 1565 (73% of the clinic population) had at least one TFT taken and could be included in this study. The characteristics of those who did not have TFT measured were similar to the characteristics of those screened for TFT (data not shown). A total of 3584 samples were analysed. Individuals had a median of 2 (range 1–22) TFTs taken.

We were able to determine from pharmacy returns for the last year that 13 patients in this cohort had prescriptions for thyroxine. These were mainly for pre-existing diagnoses of primary hypothyroidism, but also included one patient with pituitary failure, and two who were hyperthyroid and had been treated with radioactive iodine, so needed replacement therapy. Of these 13 patients, eight had a normal TSH and FT4 result, and hence had pre-existing diagnoses of hypothyroidism that was being adequately replaced. The other five were new diagnoses requiring replacement. One further patient was identified as having hyperthyroidism from pharmacy returns for carbimazole and had normal TFTs. All these patients were included in the ‘overt’ disease category so that ‘period’ prevalence could be calculated.

Table 1 shows the characteristics of those with overt thyroid disease compared with the whole study population. A total of 1565 patients, of whom 1233 (79%) were male, were screened at least once. Overall, 1043 patients (66%) were white and 365 (23%) were black African, and in 969 (62%) the main risk for HIV infection was homosexual sex. Median age at baseline was 38 years [interquartile range (IQR) 33–44 years]. Two hundred and thirteen (14%) patients were in-patients in hospital during the study period and 900 (58%) were on HAART at the start of the study.

Thirty-nine patients (2.5%) were found to have overt hypothyroidism; eight (<1%) had overt hyperthyroidism (both these may have included diagnoses pre-dating HIV diagnosis); 61 (4%) had subclinical hypothyroidism; five (<1%) had subclinical hyperthyroidism, and 263 (17%) had a nonthyroidal illness pattern of TFT. Of the remainder, 1118 (75.5%) had normal TFT function results. On further analysis, the 39 overt hypothyroid patients were more likely to be women (P=0.009), older (P=0.06), have heterosexual risk (P=0.03), to be on HAART (P=0.06), to have an AIDS diagnosis (P=0.01) and to have a lower nadir CD4 cell count (median=99 cells/μL; P=0.07). However, when all these factors were included in the multivariable model (Table 2), gender and risk group appeared to explain the association between AIDS diagnosis and nadir CD4 cell count and overt hypothyroidism, and although the numbers are small, none of these independent variables was significantly associated with overt hypothyroidism.

When those on HAART with overt or subclinical hypothyroidism were compared with the remainder of the cohort, no significant difference was found (P=0.34). Eighteen (18%) of the 100 patients with overt or subclinical hypothyroidism were currently receiving stavudine (d4T), compared with 254 (17%) of the 1457 patients with neither overt hyperthyroidism nor hypothyroidism (P=0.89). Forty-five (45%) of the 100 patients with overt or subclinical hypothyroidism had ever received stavudine (d4T), compared with 601 (41%) of the 1457 patients with neither overt hyperthyroidism nor hypothyroidism (P=0.46).

Overall, 825 patients (53%) had more than one TSH and FT4 measurement during the study period. Of these, 23 patients (3%) already had a diagnosis of overt thyroid dysfunction. A further 683 patients (82%) had ‘normal’ TFTs at the time of their first visit, of whom 604 (88%) had a ‘normal’ result at their second visit. Further longitudinal results are shown in Table 3. Only eight patients (1%) developed clinically significant changes in TFTs over the 3-year period (seven with overt hypothyroidism and one with hyperthyroidism). In this group only one patient had started HAART between the two measurements.

Discussion

Prior to HAART, there was no evidence of a uniform pattern of association between thyroid dysfunction and HIV infection [9]. However, recent work suggests an increased prevalence of thyroid dysfunction since the introduction of HAART [1–7]. These data from Europe were obtained in various subpopulations with differing background prevalences of thyroid dysfunction. Multivariate analyses suggested that the cumulative dose of d4T was significantly associated with an increased risk of developing subclinical hypothyroidism [1,2] but no other risk factors, other than increasing age, were found. In this cohort, neither HAART regimen nor specifically d4T use was significantly associated with either overt hypothyroidism or subclinical hypothyroidism.

Beltran et al. [3] and Collazos et al. [5] found a similar prevalence of overt hypothyroidism to that found in this work (2.6% and 3.5%, respectively). The numbers screened in this work are greater than those in previous studies and we have examined longitudinal data. However, the data are retrospective and we do not have significant thyroid antibody data, which may be helpful in establishing if HAART or HIV infection is important. Previous work on 84 patients also found an increased prevalence of subclinical hypothyroidism, and this was in patients treated with HAART compared with drug-naïve patients [2]. Quirino et al. [6] investigated 687 patients, of whom 7% had subclinical hypothyroidism, but they found no significant relationship between the condition and drugs or CD4 cell count. The only other work with similar numbers to this study is by Sen et al. [7]. In that study, 2437 people were screened once and 1% were reported as hyperthyroid and hypothyroid. They reported a class effect, with protease inhibitors being significantly associated with hypothyroidism and NNRTIs associated with hyperthyroidism. Their data also suggested an increase in hypothyroidism during the HAART era.

There is no clear evidence that either sub clinical hypothyroidism or subclinical hyperthyroidism should be treated [14]. Recommendations in current guidelines are not to treat, and therefore the value of identifying subclinical disease that progresses slowly (if at all) to clinically significant disease is questionable. It must be acknowledged that thyroid dysfunction is often secondary to nonthyroidal illness and immediate treatment is not always appropriate. Changes in TSH and FT4 in systemic illness generally are poorly understood. False-positive and false-negative results are not uncommon, and so it is wise to be cautious before drawing conclusions regarding thyroid dysfunction and HIV infection until bigger studies are analysed [15]. Also, no clear mechanism is known for such proposed thyroid dysfunction. The relatively high prevalence of nonthyroidal illness may in part be attributable to the fact that TFTs are taken routinely from newly diagnosed patients who are more likely to be unwell and hospitalized. Fourteen per cent of the study population were hospitalized at some point during the study period. We also do not have data on drugs that patients may have been taking, which may cause a nonthyroidal illness pattern of results, such as steroids, heparin and nonsteroidal drugs. Differentiation of nonthyroidal illness from secondary hypothyroidism was performed by repeat sampling (Table 3), and if necessary further routine clinical and biochemical hormonal evaluation.

Epidemiological studies have revealed the heterogeneity of the population infected with HIV in the UK [16]. This cohort has a large proportion of black African patients. As increasing numbers of women are diagnosed HIV-positive, and as the HIV-infected population becomes older as a result of the improved prognosis, more thyroid disease will be seen. Also, with increasing numbers being diagnosed and treated for hepatitis C virus (HCV) infection, more patients will become hypothyroid as a consequence of this treatment, irrespective of any influence of HIV disease itself or HAART. In this cohort, only four patients had become hypothyroid as a result of treatment for HCV infection, but with more patients being treated it is likely that this number will rise. It seems that larger studies will be needed to establish whether there is a significant increase in the prevalence of thyroid dysfunction related to HIV infection. Such an increase is likely given the slow natural history of evolving thyroid disease in the non-HIV-infected population, where the prevalence of overt hypothyroidism is 1–2% and that of overt hyperthyroidism is 0.5–2% in women, both conditions being 10 times more common in women than in men [11]. Therefore, a clear answer to whether HIV disease or HAART impacts on thyroid function is likely to take some time and require large cohorts. However, our study implies that the prevalence of overt thyroid disease is likely to be similar to that observed in HIV-negative populations and that routine screening of this specific population is not necessary, especially given that patients with overt clinical disease do present clinically and that TFTs are a test clinicians have a low threshold for performing.