We searched PubMed (1970 to May 2015) and EMBASE (1987 to May 2015) for eligible studies using common keywords related to alcohol intake and the risk of thyroid cancer. The keywords for literature search were as follows: “alcohol,” “thyroid cancer,” and “cross-sectional study,” “casecontrol study,” or “cohort study.”

We included observational studies such as case-control studies and prospective or retrospective cohort studies that investigated the association between alcohol intake and the risk of thyroid cancer, reporting outcome measures with adjusted odds ratios (ORs) or relative risk (RRs) and 95% confidence intervals (CIs). If data were duplicated or shared in more than one study, the first published study or most comprehensive study was included in the analysis. The language of publication was restricted to English.

Two of the authors (S.-H.H. and H.S.K.) independently evaluated the eligibility of all studies retrieved from the databases based on the pre-determined selection criteria. Disagreements between evaluators were resolved by discussion. We extracted the following data from studies included in the final analysis: study name (along with the name of the first author and year of publication), type of study design, country, study period (years), population (age, sex), definitions of alcohol use (highest vs. lowest category), OR/RR with 95% CI, and adjustments of variables.

We assessed the methodological quality of included studies based on the Newcastle-Ottawa Scale (NOS) for assessing the quality of case-control studies and cohort studies in meta-analyses [13]. The star system of the NOS ranged from 0 to 9 and consisted of three subscales: selection of studies, comparability, and exposure. In the current study, we considered a study awarded more than a mean score of each study type to be a high-quality study because criteria for quality of a study have not been established. Two pooled studies [11,12] with 14 case-control studies, five cohort studies and a cross-sectional study were excluded from this assessment of methodological quality because no required information for the methodological assessment of each study was provided.

We investigated the association between alcohol intake (highest use vs. lowest use) and risk of thyroid cancer using adjusted ORs or RRs with 95% CIs for the main analysis. We also performed subgroup analyses based on type of thyroid cancer, gender, region of study population (America, Asia, and Europe), amount of alcohol consumed, type of alcohol, methodological quality, and type of study design (case-control or cohort studies).

To calculate a pooled OR or RR with 95% CI, we used adjusted ORs or RRs and 95% CIs reported in individual articles whenever possible. We evaluated heterogeneity in results across studies using Higgins, which measures the percentage of total variation across studies [14]. I² was calculated as follows:

where, Q is Cochran’s heterogeneity statistic and df is the degrees of freedom. Negative values of I² are set at zero; I² lies between 0% (no observed heterogeneity) and 100% (maximal heterogeneity). An I² value greater than 50% was considered to indicate substantial heterogeneity.

When substantial heterogeneity was not observed, the pooled estimate calculated based on the Fixed-effect model was reported using Woolf’s method (inverse variance method). When substantial heterogeneity was observed, the pooled estimate calculated based on the random-effects model was reported using the DerSimonian and Laird method [15].

We also examined publication bias regarding the studies included in the final analysis using Begg’s funnel plot and Egger’s test. If publication bias exists, the Begg’s funnel plot is asymmetrical or the p-value is less than 0.05 by Egger’s test. Statistical analyses were conducted using the Stata SE ver. 12.1 software package (StataCorp., College Station, TX).

Fig. 1 shows a flow diagram of how relevant studies were identified. A total of 213 articles were identified by searching PubMed and EMBASE. We excluded 28 duplicated articles and an additional 140 articles that did not satisfy the selection criteria. The full texts of the remaining 45 articles were reviewed, and 29 additional articles were for the reasons shown in Fig. 1. The remaining 16 articles [11,12,16-29] were included in the final analysis.

We identified a total of 33 observational studies with two cross-sectional studies, 20 case-control studies, and 11 cohort studies from 16 articles, which included 7,725 thyroid cancer patients among 3,113,679 participants. Table 1 shows the general characteristics of the studies in the final analysis. The included studies were performed in the United States (n=15), Sweden (n=3), Japan (n=3), Korea (n=2), Norway (n=2), Canada (n=1), French Polynesia (n=1), Greece (n=1), Italy (n=1), New Caledonia (n=1), Switzerland (n=1), and the United Kingdom (n=1). The mean age was 48.5 years (range, 10 to 89 years), and 74% of the study participants were women. The enrolment period for participants across studies ranged between 1964 and 2013, and the follow-up periods in the cohort studies ranged from 7.2 to 23 years.

Fig. 2 shows the effect of alcohol intake on the risk of thyroid cancer in the meta-analysis of all studies. Alcohol intake (highest vs. lowest intake) was significantly associated with decreased risk of thyroid cancer in a fixed-effect model meta-analysis of all 33 observational studies (OR or RR, 0.74; 95% CI, 0.67 to 0.83; I²=38.6%). As shown in Fig. 3, in the subgroup meta-analysis by type of study, alcohol intake also decreased the risk of thyroid cancer in both case-control studies (OR, 0.77; 95% CI, 0.65 to 0.92; I²=29.5%; n=20) and cohort studies (RR, 0.70; 95% CI, 0.60 to 0.82; I²=0%; n=11), but not in cross-sectional studies (OR, 0.82; 95% CI, 0.62 to 1.09; I²=91.3%; n=2). There was no publication bias in the 33 studies (Begg’s funnel plot was symmetric; Egger’s test, p for bias=0.97) (Fig. 4).

Table 2 shows the methodological quality of studies included in the final analysis based on the NOS. The quality scores ranged from 4 to 9, with an average score of 6.3 for case-control studies and 6.5 for cohort studies. The high-quality studies (score of 7 or higher) included two case-control studies and three cohort studies.

Table 3 summarizes the results from the subgroup meta-analyses based on several factors. In the subgroup meta-analyses by type of thyroid cancer, sex, and amount of alcohol consumed, alcohol intake was significantly associated with a decreased risk of thyroid cancer. The subgroup meta-analyses by geographic region showed that alcohol intake was associated with a decreased risk of thyroid cancer in the American region (OR or RR, 0.70; 95% CI, 0.60 to 0.81; I²=0%; n=12), but not in the European (OR or RR, 0.60; 95% CI, 0.36 to 1.00; I²=13.1%; n=2) or Asian regions (OR or RR, 0.82; 95% CI, 0.48 to 1.42; I²=76.3%; n=4).

Moreover, subgroup meta-analyses by methodological quality of study indicated that alcohol intake decreased the risk of thyroid cancer in both high-quality studies (OR or RR, 0.64; 95% CI, 0.50 to 0.82; I²=0%; n=5) and low quality studies (OR or RR, 0.68; 95% CI, 0.55 to 0.85; I²=6.7%; n=7).

Tables 4 and 5 show the findings from the subgroup meta-analyses based on various factors within each study design (case-control study and cohort study). Overall, subgroup meta-analyses of cohort studies by various factors indicated that alcohol intake significantly decreased the risk of thyroid cancer, whereas those of case-control studies indicated that alcohol intake decreased the risk of thyroid cancer, although this decrease was not statistically significant.

In this meta-analysis, we found that there was an inverse association between alcohol intake and the risk of thyroid cancer. Moreover, overall, alcohol intake was associated with a decreased risk of thyroid cancer in the subgroup meta-analyses by type of thyroid cancer, sex, amount of drink, and methodological quality of study. In the subgroup meta-analyses by various factors within each study design, alcohol intake consistently showed a significantly decreased risk of thyroid cancer in both case-control and cohort studies, but a non-significantly decreased risk in cross-sectional studies.

There are several biological mechanisms that may explain the inverse association between alcohol intake and the risk of thyroid cancer observed in our meta-analysis. First, alcohol intake may have a protective effect on developing thyroid cancer by decreased levels of thyroid-stimulating hormone (TSH). TSH made by the pituitary gland is a well-established growth factor for thyroid nodules, and it is known that higher serum TSH levels are associated with higher rates of thyroid cancer [30]. An animal study demonstrated that chronic alcohol exposure reduces the responsiveness of the hypo-thalamic-pituitary-thyroid axis to central stimulation [31]. Moreover, many cross-sectional and longitudinal studies have reported that a blunting of the plasma TSH response in the thyrotropin releasing hormone stimulation test was consistently observed in patients of alcohol dependence compared to healthy people [32]. Another biological mechanism is that alcohol has a direct toxic effect on thyroid cells and reduces thyroid volume [33,34], which might lead to a decreased risk of thyroid cancer. However, a cross-sectional study found an increasing dose-response association between alcohol intake and the risk of thyroid enlargement [35].

To the best of our knowledge, this study is the most comprehensive meta-analysis of observational studies on the association between alcohol intake and the risk of thyroid cancer. Many epidemiological observational studies have reported inconsistent findings on this issue. In 2003, Mack et al. [11] conducted a pooled analysis of 14 case-control studies that indicated alcohol intake was associated with a reduced risk of thyroid. However, after adjustment for smoking, the reduced risk was not maintained. In 2012, a pooled analysis of five prospective cohort studies from the United States revealed that alcohol intake of more than seven drinks per week was inversely associated with the risk of thyroid cancer after adjustment for smoking [12]. Our meta-analysis involved 14 additional observational studies with cross-sectional, case-control, cohort studies as well as 20 case-control and cohort studies from those two pooled analyses. Four additional case-control studies have been conducted since the pooled analysis by Mack et al. [11] was published, most of which reported a protective effect of alcohol intake against the risk of cancer. Conversely, Xhaard et al. [28] showed an increased risk of thyroid cancer, although this increase was not significant. In the present study, a meta-analysis of 11 cohort studies (including six cohort studies published individually and five cohort studies from a pooled analysis) indicated that alcohol intake decreased the risk of thyroid cancer. Given that a cohort study generally shows a higher level of evidence than a case-control study, we believe that the results of the meta-analysis of cohort studies are more reliable than those from the meta-analysis of case-control studies.

When we performed subgroup meta-analyses by various factors, alcohol intake was consistently associated with a decreased risk of thyroid cancer, regardless of the type of thyroid cancer, sex, and amount of alcohol consumed. Even though alcohol intake had no association with thyroid cancer risk in the subgroup meta-analyses by study region (European and Asian regions), type of alcohol (wine, beer, or liquor), or methodological quality of study (low-quality studies), it is considered that those findings are attributable to the relatively small number of studies included in the analysis.

The subgroup meta-analyses by various factors within each study design showed that alcohol intake was consistently associated with a significantly decreased risk of thyroid cancer in both case-control and cohort studies, but a non-significantly decreased risk in cross-sectional studies. The non-significant findings in cross-sectional studies are also thought to be due to the small numbers of studies in the subgroup meta-analysis.

Choi et al. [27] conducted a study in Korea that showed an apparent increased risk of thyroid cancer in drinkers, unlike the other studies. We do not have any clear explanation for this finding. However, it should be noted that the study conducted by Choi et al. [27] had several limitations. Specifically, they determined whether respondents consumed alcohol simply based on asking “drinking or not,” not the amount of alcohol consumed. Moreover, they used a cross-sectional design, which is generally considered to provide a lower level of evidence than case-control studies or cohort studies. Conversely, a recent case-control study conducted in Korea reported that acute high-dose alcohol intake (151 g or more per event or on a single occasion) and long-term intake of alcohol for 31 or more years are associated with an increased risk of differentiated thyroid cancer [36]. Overall, further prospective studies are warranted to confirm the association between heavy drinking and the risk of thyroid cancer.

It should be noted that our study has several limitations. First, definitions for one drink of alcohol in each study included in the analysis were not identical. For example, Allen et al. used the total number of alcoholic drinks containing about 10 g of alcohol to categorize study participants into five groups [22], whereas Kitahara et al. [12] considered one drink to be equal to 14 g of alcohol. However, this discrepancy in the definitions for one drink would not lead to serious bias in our main findings on the association between alcohol intake and the risk of thyroid cancer. We were also not able to explain the exact reasons for the decreased risk of thyroid cancer associated with alcohol intake because we only included observational studies. Although we described possible mechanisms previously in the discussion section, further research is required to confirm these potential mechanisms. Moreover, we have limited information regarding alcohol intake during follow-up in cohort studies because information was collected only at one specific baseline point in time. Finally, due to a paucity of information in individual studies, we were unable to investigate the effects of alcohol intake on the risk of thyroid cancer according to types of alcohol beverages.