Subclinical hypothyroidism and the risk of cancer incidence and cancer mortality: a systematic review

A systematic search was performed on Ovid MEDLINE from the date of its inception until November 13th, 2017, combining words related to thyroid and cancer (MeSH and non- MeSH terms) to identify studies of thyroid dysfunction and incident cancer and those examining thyroid dysfunction and mortality in cancer patients (Appendix). A second search was performed using Pubmed from the date of its inception until March 12, 2020 (Appendix). No language limitations were used. In addition, we hand-searched the references of included studies to identify additional relevant studies that were not identified in our electronic search.

Inclusion criteria comprised of randomized clinical trials assessing treatment of SCH with levothyroxine treatment, cohort and case-control studies reporting SCH, where the thyroid dysfunction chronologically preceded the cancer incidence or mortality by at least a year to reduce the possibility of including studies with reverse causation, meaning that the cancer or its treatment induced the thyroid dysfunction and not the opposite [8]. Articles where the SCH was primary (not secondary to a medical treatment or procedure) were included. Articles reporting iatrogenic hypothyroidism, thyroid dysfunction during pregnancy or were medication induced were not included since the effect of these causes of SCH on cancer might have a different pathophysiology considering that SCH is commonly induced by an autoimmune mechanism [9, 10]. Exclusion criteria were studies using outdated methods to screen or diagnose thyroid dysfunction or cancer, suspected to have a high risk of bias (i.e. recall) according to the study methods, and studies in which the reference and comparison groups have cancer. Studies were screened by two investigators (J.G. and O.Y.) and discrepancies were resolved by a third investigator (K.B.F.).

Data were extracted using a pilot-tested form that included author, year of publication, study design, study period, sample size (overall and by group), type of population, age of the patients, source of the data or database used for the study, follow-up duration, objective of the study, if anti-thyroperoxidase antibodies (TPOAb) had been measured, effect measure, results (adjusted rates, survival times, and mortality rates), and conclusions.

Quality assessment was performed using the Cochrane tool to assess risk of bias in cohort studies [11] and the Newcastle-Ottawa quality assessment scale for case control studies [12]. Data extraction and quality analyses were completed independently by two authors (J.G. and O.Y.). Disagreements were resolved by discussion with a third independent reviewer (K.B.F.).

Meta-analysis was not possible due to the high heterogeneity in study designs, effect measures reported, and outcomes (cancer incidence, mortality, survival time). All the methods followed the PRISMA guidelines for systematic reviews of the literature [13, 14].

A total of 37,073 records were identified (Fig. 1); 36,954 were retrieved from Ovid MEDLINE, 117 were retrieved from Pubmed and 2 were retrieved from other sources [15, 16] (cited in other articles obtained via our search). 36,886 articles were found non-relevant after screening the titles and abstracts. One hundred and eighty-four articles were screened via further assessment of the methods described and subsequently 53 articles were thoroughly reviewed for eligibility. Forty-four articles were excluded for the following reasons: 23 studies identified concurrent thyroid dysfunction and cancer (reverse causation cannot be excluded) [17‐38], 1 study used an outdated method to screen for breast cancer (i.e. thermography) [39], 12 studies did not report SCH [40‐51], 3 studies have potential high risk of recall bias (no specific tool was applied to assess for risk of recall bias at this point; it was identified by the assessor after reading the methods of the article) [52‐54], 2 studies included iatrogenic hypothyroidism [55, 56], 2 studies had an inappropriate definition of hypothyroidism (both defined hypothyroidism based on use of thyroid hormone replacement without reporting levels of TSH or T4) [57, 58], and 1 study involved cases and controls with a history of cancer at baseline [59]. A total of 9 studies were included in our systematic review.

The 9 included studies comprised of 2 case-control studies [7, 60], 3 retrospective cohort studies [16, 61, 62], and 4 prospective cohort studies [8, 15, 63, 64] (Tables 1 and 2). There were no randomized controlled trials identified that studied the effects of treatment of SCH on the risk of cancer or cancer mortality. One included study assessed the effect of SCH on colorectal cancer [7], one study on breast cancer [8], one study on prostate cancer [60], one on thyroid cancer [61] and one on hepatocellular carcinoma [62]. One study analyzed overall cancer incidence [15], and three studies focused on overall cancer mortality [16, 63, 64]. Two of these studies assessed cancer mortality as a secondary outcome [16, 64].

Overall, the cohort studies had good quality (Table 3). The study by Fighera et al. [61] had high risk of bias because the exposed and non-exposed groups did not come from the same source population [61]. Also, it was not clear if SCH was identified prior to the cancer outcome [61]. Two studies did not fully adjust for confounder factors present in the exposed and non-exposed groups [16, 62]. Razvi et al. [16] assessed cancer mortality as a secondary outcome and the variables that were adjusted in the analyses were focused for cardiovascular outcomes. Finally, Pinter et al. [62] assessed the association between thyroid dysfunction and overall survival among patients with hepatocellular carcinoma. Unfortunately, the data collected for these patients lacked information on a number of potential confounders related to patient characteristics, including comorbidities.

Case control-studies had very good quality overall (Table 4). The identification of cases involved record linkage only to a primary care database in the study by Boursi et al. [7] without independent validation.

The risk of cancer incidence varied depending on the type of cancer studied. Hellevik et al. [15] found no association between hypothyroidism (i.e. inclusion of SCH and overt hypothyroidism) and overall cancer risk (patients with TSH > 3.5 mIU/L, risk of all cancers adjusted hazards ratio [HR]: 0.96; 95% CI: 0.82–1.12). In this study, individual cancer risks for the more common cancers were also assessed (i.e., lung, colon, prostate, and breast cancer). There were no observed increased risks of lung (adjusted HR: 0.87; 95% CI: 0.43–1.74), colon (adjusted HR: 0.95; 95% CI: 0.60–1.50), prostate (adjusted HR: 0.86; 95% CI: 0.55–1.35), and breast cancer (adjusted HR: 0.85; 95% CI: 0.57–1.25) among patients with a TSH level of > 3.5 mIU/L relative to patients with a TSH level within normal range (TSH: 0.5 to 1.4mIU/L). In contrast, Boursi et al. [7] reported a modest increased risk of colorectal cancer in patients with untreated hypothyroidism and SCH without thyroid hormone replacement compared to euthyroid patients (adjusted odds ratio [OR]: 1.16; 95% CI: 1.08–1.24). This study also assessed the effects of levothyroxine treatment for SCH compared to euthyroid patients and found a protective effect of thyroid hormone replacement against colorectal cancer (adjusted OR: 0.92; 95% CI: 0.86–0.98).

Two studies assessed the association between thyroid dysfunction and breast cancer risk. Kuijpens et al. [8] demonstrated that low levels of free T4 (fT4) are an independent risk factor for the development of breast cancer in peri- and post-menopausal women (risk of cancer in patients with low fT4 levels (≤10th percentile which is equivalent to ≤12.5 pmol/L, OR: 2.3; 95% CI: 1.2–4.6). As mentioned above, Hellevik et al. [15] found no increased risk of breast cancer in patients with hypothyroidism.

Only one study assessed the risk of prostate cancer associated with thyroid function. Mondul et al. [60] reported that hypothyroid men (i.e. inclusion of men with SCH and overt hypothyroidism) had a significantly reduced risk of overall prostate cancer compared to euthyroid men (adjusted OR: 0.71; 95% CI: 0.47–1.06 for TSH ≥ 2.2mIU/L). This result conflicted with that found by Hellevik et al. [15] whereby there was no association between the risk of prostate cancer and SCH or hypothyroidism. Finally, Fighera et al. [61] assessed the risk of thyroid cancer associated with thyroid function and reported an association between serum levels of TSH of > 1.64mIU/L and thyroid malignancy (OR: 2.57; 95% CI: 1.41–4.70).

Four articles reported cancer-related mortality outcomes [16, 62‐64]. Pinter et al. [62], assessed overall survival for 667 patients with hepatocellular carcinoma and found that patients with SCH had a slightly decreased survival time compared to euthyroid patients (median overall survival in patients with SCH: 6.1 months; 95% confidence interval [CI]: 0–13.5 versus 10.8 months; 95% CI: 6.5–15.2). Tseng et al. [63], performed a prospective cohort study of 115,746 patients followed in Taiwan with 10 years of follow-up and reported that SCH was independently associated with increased risk of cancer mortality (adjusted relative risk [RR]: 1.51; 95% CI: 1.06–2.15). Their study showed that the risk of cancer mortality among patients with SCH compared to euthyroid patients was more pronounced in bone, skin, and breast cancer. In a sub-analysis from the same study [63] restricted to patients with SCH defined as having TSH levels between 5 and 9.9 mIU/L, the association between SCH and increased cancer-related mortality remained significant (adjusted RR: 1.61; 95% CI: 1.12–2.31). Razvi et al. [16] conducted a retrospective cohort study using the General Practice Research Database (GPRD; since renamed the Clinical Practice Research Datalink), which is representative of patients followed in primary care in the United Kingdom. They determined that patients with SCH between the ages of 40 and 70 years treated with levothyroxine (n = 1634) had lower cancer-related mortality compared to untreated patients with SCH (n = 1459) (adjusted hazard ratio [HR]: 0.59; 95% CI: 0.21–0.88). Similarly, Waring et al. [64] conducted a prospective study using the Osteoporotic Fractures in Men (MrOS) study cohort which is comprised of 1587 men age ≥ 65 years with baseline thyroid function testing. This study assessed the association between thyroid function, including SCH and the risk of mortality over a mean follow-up of 8.3 years. The results of this study found no association between SCH and the risk of cancer-related mortality compared with euthyroid men (relative hazard: 0.88; 95% CI: 0.44–1.74) (Table 2).

Only one study addressed the association between TPOAb and the risk of cancer, specifically breast cancer risk. Kuijpens et al. [8] found that TPOAb were more prevalent in women with a previous or current diagnosis of breast cancer (OR: 3.0; 95% CI: 1.41–6.46). However, the presence of TPOAb was not associated with the development of incident in situ breast cancer during follow-up (OR: 1.1; 95% CI: 0.4–2.7) (Table 2).