Process of study selection and description of qualified studies
A total of 2415 studies were identified on our online databases.After exclusion of duplicate references,129 articles were considered. After screening the abstract and title, 102 articles were excluded. After careful review of the full texts, 8 studies have been excluded because 5 of them did not provide relevant data, and 3 articles did not have full-text. 19 articles published between 1978 and 2019 met the inclusion criteria (Fig. 1).
A total of 367416 samples from 19 studies involving were enrolled in this meta-analysis[4, 8–10, 15–29]. 6 cohort studies and 13 case-control studies were included in the study. Twelve articles were studied in the European population, five in the North American population, and two in the Asian population. All articles are of high quality because of NOS score no less than 6. The chief characteristics of the enrolled materials are detailed in Table 1.
Table 1
Main characteristics of the included studies in our-analysis
Study | Year | Region | Sample | Follow-up | Median (Mean)Age | NOS | Study design |
(years) | (years) |
Hoffman | 1984 | Swedish | 1665 | 21.9 | 47.2 | 8 | Cohort |
Kuijpens | 2005 | Netherlands | 2775 | 9 | 50.5 | 8 | Cohort |
Sandhu | 2009 | Canada | 179462 | 10 | 74.9 ± 7 | 8 | Cohort |
Hellevik | 2009 | Norwegian | 29691 | 9 | ≥ 20 | 7 | Cohort |
Søgaard | 2016 | Danish | 61873 | 35 | 71 | 7 | Cohort |
Grani | 2012 | Italy | 380 | 5 | 59 | 7 | Case-control |
Ditsch | 2010 | Germany | 130 | NA | 58.6 ± 13.5 | 7 | Case-control |
Cristofanilli | 2005 | USA | 2224 | 3 | 51.6 ± 12.6 | 6 | Case-control |
Simon | 2002 | USA | 9257 | 4 | NA | 6 | Case-control |
Talamini | 1997 | Italy | 5157 | 3 | 55 | 7 | Case-control |
Smyth | 1996 | Ireland | 400 | 1 | 57.2 ± 1.4 | 7 | Case-control |
Shering | 1996 | Ireland | 350 | NA | NA | 7 | Case-control |
Moseson | 1993 | Canada | 1101 | 4 | 54 | 7 | Case-control |
Turken | 2003 | Prague | 250 | 4 | 63 | 6 | Case-control |
Brinton | 1984 | USA | 2612 | 4 | NA | 7 | Case-control |
Kalache | 1982 | UK | 2352 | 11 | NA | 6 | Case-control |
Adami | 1978 | Sweden | 358 | 1 | 64 | 7 | Case-control |
Weng | 2018 | USA | 103466 | NA | 53.3 | 8 | Case-control |
Kim | 2019 | Korea | 67416 | 4 | ≥ 40 | 8 | Cohort |
Table 2
Stratiedanalysis of the relationship between hypothyroidism and breast cancer risk.
Variable | No.of studies | OR(95%CI) | P | Heterogeneity |
| | | | I² | Ph |
Region | | | | | |
Europe | 12 | 0.93 (0.88–0.99) | < 0.001 | 0 | 0.877 |
North America | 5 | 0.86 (0.60–1.11) | < 0.001 | 93.8% | 0 |
Asia | 2 | 1.17 (0.98–1.35) | < 0.001 | 0 | 0.319 |
Study design | | | | | |
Case-control | 13 | 0.85 (0.62–1.09) | < 0.001 | 80.4% | 0 |
Cohort | 6 | 0.96 (0.91–1.01) | < 0.001 | 0 | 0.517 |
Follow-up date | | | | | |
> 4 | 7 | 0.96 (0.91-1.00) | < 0.001 | 0 | 0.435 |
≤ 4 | 9 | 0.80 (0.54–1.70) | 0.015 | 81.0% | 0 |
Relationship between hypothyroidism and breast cancer risk
There were 20 studies reported the relationship between hypothyroidism and breast cancer risk. With obvious heterogeneity (I 2 =78.2%, p = 0.000) among these studies, so a random effect model was used for assessment. The pooled analysis suggested that hypothyroidism can reduce the risk of breast cancer(OR 0.90, 95% CI 0.77–1.03; p < 0.001)( ( Fig. 2a).
Subgroup analysis of hypothyroidism and risk of breast cancer
To further explore the relationship between hypothyroidism and breast cancer risk, subgroup analysis was conducted from three aspects: study type, population distribution, and follow-up time. In the subgroups with a follow-up date of no more than four years, the case-control study, and the North American study, the results showed that the risk of breast cancer in patients with hypothyroidism decreased by 20%, 15%, and 14%, respectively( Fig. 3). But in Asian patients with hypothyroidism, the risk of breast cancer increases by 17% (OR 1.17, 95% CI 0.98–1.35; p < 0.001) ( Fig. 3a).
Relationship between thyroid hormone replacement therapy and breast cancer risk
A total of 10 studies reported the relationship between the use of thyroid hormone replacement therapy and the risk of breast cancer[4, 8, 9, 16, 18, 22, 24, 26, 27]. As obvious heterogeneity observed, the fixed-effect model was used(I 2 =86.3%, p = 0.000). The pooled analysis suggested that patients who received thyroid hormone replacement therapy had a lower risk of developing breast cancer(OR = 0.87, 95% CI 0.65–1.09; p < 0.001) ( Fig. 4a).
Publication bias
Figure 5A shows a funnel plot of the 20 articles included in this meta-analysis. The Begg test (Pr = 0.529) and the Egger test(P = 0.892) were used to detecting publication bias showed that there was no possibility of publication bias. As shown in Fig. 5B, there were no publication biases in the ten articles on the study of thyroid hormone replacement therapy. The Egger test was P = 0.672, and the Begg test (Pr = 0.858).
Sensitivity analysis
The results of sensitivity analysis are generally stable, and the primary source of heterogeneity is in the research of Cristofanilli et al[24]. (Fig. 6a). So we excluded the literature of Cristofanilli and analyzed the other studies. The results revealed that the hypothyroidism could reduce the risk of breast cancer (OR:0.96 95%CI:0.92 -1.00, P < 0.001), and there was no heterogeneity(I2 = 0, P = 0.577)( Fig. 2b).
As shown in Fig. 6b, we can see that the source of heterogeneity is the Cristofanilli and Sandhu articles[24, 26]. Therefore, we removed the two articles and re-executed the meta-analysis. With no obvious heterogeneity (I 2 =0, p = 0.922) among these studies, so a fixed-effect model was used for assessment. The results show that patients taking thyroid hormone replacement therapy have a lower risk of breast cancer(OR:0.96 95%CI:0.83–1.08, P < 0.001( Fig. 4b).