Oh, no--equations! Reanalysis of the data reveal even stronger associations between Hashimoto's thyroiditis and differentiated thyroid cancer.
E. Rijntjes,L. Schomburg
DOI: https://doi.org/10.1089/thy.2014.0301
IF: 6.506
Thyroid
Abstract:With great interest, we have been reading the paper by Paparodis et al. on the association between differentiated thyroid cancer (DTC) and Hashimoto’s thyroiditis (HT) (1). The paper readdresses the interaction of the common destructive autoimmune disease with malignant transformation of the thyroid gland almost exactly 60 years after its first mentioning (2) and at a time when the potential association is still of central interest to thyroidologists (3). The particular strengths of the study by Paparodis et al. constitute the wealth of data and a sufficiently large study cohort allowing for biostatistical calculations on the association between DTC and HT. A significantly increased relative risk (RR) for DTC in clinical hypothyroid HT (hypo-HT) patients receiving levothyroxine is reported, as compared to the subjects without HT (non-HT); RR = 1.04 (1.01–1.07, p = 0.0163). The authors further split the hypo-HT group into tertiles of levothyroxine replacement dose, with only the RR for DTC in the lowest tertile of levothyroxine replacement (RR = 1.04) reaching significance, and the other tertiles having a RR of 1.00 as compared to the non-HT. As important and clinically relevant news should do, these data rapidly found their way into the major German physician magazine Ärztezeitung Online (4). However, we felt that the reported significant RRs between 1.01 and 1.13 on first sight advocate a most meek effect size, suggesting the association between DTC and HT is present on paper but limited at the patient level. However, after reanalyzing the raw data reported in the unconventionally presented contingency table in the original manuscript, we feel that the study is of far bigger value and the effects are of a far stronger nature than reflected by the low RR values reported by the authors and in the media. The RR is a statistical measure typically used in cohort studies to describe the incidence of an event (here, DTC diagnosis) in the exposed subjects (in this study, hypo-HT) divided by the incidence of an event in the nonexposed subjects (in this study, non-HT): [a/(a + b)]/[c/(c + d)] = 1.32 (see Table 1). The odds ratio (OR) describes the odds of a subject developing DTC while having hypo-HT divided by the odds of a subject developing DTC while being non-HT: (a/b)/(c/d) = 1.53. When performing these calculations with the raw data reported in the study of Paparodis et al. (1), the association between hypo-HT and DTC becomes far stronger than reported in their publication. The RR for DTC in the lowest tertile of levothyroxine replacement amounts to an impressive 2.02, exceeding the reported value of 1.04 considerably. Patients in the two other tertiles remain at a RR of 1.01 for the middle and 0.95 for the higher tertile and, hence, close to the reported RR of 1.00. We assume that the RR values reported in the paper rather reflect the probability of a subject not having HT while having non-DTC divided by the probability of not having HT while having DTC: [d/(b + d)]/[c/(a + c)] = 1.04. By presenting the association in such an uncommon format, we are afraid that the data might not receive the attention and interpretation they deserve, as this uncommon ‘‘RR’’ calculation yields a value close to 1.0. After recalculating the data in the orthodox way, the RR values unfold to be far more impressive, thereby strongly supporting the claims and conclusions made by the authors. The HT pathology significantly and substantially increases the risk for DTC in euthyroid patients (RR = 1.62) and in patients with partially functional thyroid glands (RR = 2.02), whereas high thyroid peroxidase antibody titers apparently confer some protection against DTC in HT patients. Whether this protection is based on a more effective clearance of malignant thyrocytes by activated infiltrating immune cells in HT (immunosurveillance) remains to be determined in future studies.