Download PDF
Horm Metab Res 2014; 46(06): 440-444
DOI: 10.1055/s-0034-1371832
Endocrine Care
© Georg Thieme Verlag KG Stuttgart · New York

Underestimation of Thyroid Dysfunction Risk Due to Regression Dilution Bias in a Long-Term Follow-Up: Tehran Thyroid Study (TTS)

A. Amouzegar
1   Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, I. R. Iran
,
M. Beigy
1   Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, I. R. Iran
,
S. Gharibzadeh
1   Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, I. R. Iran
2   Department of Epidemiology and Biostatistics, School of Public Health, Tehran University of Medical Sciences, Tehran, I. R. Iran
,
F. Azizi
1   Endocrine Research Center, Research Institute for Endocrine Sciences, Shahid Beheshti University of Medical Sciences, Tehran, I. R. Iran
› Author Affiliations
Further Information

Publication History

received 15 October 2013

accepted 27 February 2014

Publication Date:
05 May 2014 (online)

Also available at
    Permissions and Reprints

Abstract

Thyroid dysfunction is linked with mortality and particular diseases. Intra-individual variability of measured thyroid function parameters may bias its association with outcomes, the so called “regression dilution” bias. Single measurements of thyroid function parameters result in underestimation of real associations between outcome rates with the “usual life-long levels” of the aforesaid parameters. The aim of this study was to examine the intra-individual variability of FT4 and TSH of study cohorts in the Tehran Thyroid Study (TTS) and to investigate the extent of the risk underestimation during the 4 phases (Ph) of TTS, with median follow-up of 4, 7, and 10 years between the Ph2-Ph1, Ph3-Ph1, and Ph4-Ph1 intervals; respectively. We estimated regression dilution ratios (RDRs) by the Rosner method of linear regression of repeated measures for FT4 and TSH. RDR1, RDR2, and RDR3 were obtained by regressing the repeated measures of the aforesaid parameters of the last 3 TTS follow-ups on the baseline measurements. Calculations showed 0.64 RDR1, 0.58 RDR2, and 0.52 RDR3 for TSH; and 0.62 RDR1, 0.57 RDR2, and 0.55 RDR3 for FT4. A single measurement-based risk estimation in the TTS was underestimated for FT4 about 61.2, 76.5, and 80.4%; and for TSH as 55.8, 73.1 and 93% after 4, 7, and 10 years of follow-up; respectively. In conclusion, using only single measurements of TSH and FT4 the association between thyroid function and outcome rates is considerably underestimated, especially after a long follow-up period.

Key words

regression dilution bias - correction methods - long-term variability - thyroid disorders - cohort studies - risk assessment

Supporting Information

 

  • References

  • 1 van de Ven AC, Netea-Maier RT, Medici M, Sweep FCGJ, Ross HA, Hofman A, de Graaf J, Kiemeney LA, Hermus AR, Peeters RP, Visser TJ, den Heijer M. Underestimation of Effect of Thyroid Function Parameters on Morbidity and Mortality due to Intra-Individual Variation. J Clin Endocrinol Metab 2011; 96: E2014-E2017
    CrossrefPubMedGoogle Scholar
  • 2 Andersen S, Bruun NH, Pedersen KM, Laurberg P. Biologic variation is important for interpretation of thyroid function tests. Thyroid 2003; 13: 1069-1078
    CrossrefPubMedGoogle Scholar
  • 3 Maes M, Mommen K, Hendrickx D, Peeters D, D’Hondt P, Ranjan R, De Meyer F, Scharpe S. Components of biological variation, including seasonality, in blood concentrations of TSH, TT3, FT4, PRL, cortisol and testosterone in healthy volunteers. Clin Endocrinol 1997; 46: 587-598
    CrossrefPubMedGoogle Scholar
  • 4 Collaboration TFS . Regression dilution methods for meta-analysis: assessing long-term variability in plasma fibrinogen among 27 247 adults in 15 prospective studies. Inter J Epidemiol 2006; 35: 1570-1578
    CrossrefPubMedGoogle Scholar
  • 5 Morris RW, Emberson JR. Commentary: Over-correction for regression dilution bias? Not for blood pressure vs. coronary heart disease. Inter J Epidemiol 2005; 34: 1368-1369
    CrossrefPubMedGoogle Scholar
  • 6 Clarke R, Shipley M, Lewington S, Youngman L, Collins R, Marmot M, Peto R. Underestimation of risk associations due to regression dilution in long-term follow-up of prospective studies. Am J Epidemiol 1999; 150: 341-353
    CrossrefPubMedGoogle Scholar
  • 7 Frost C, Thompson SG. Correcting for regression dilution bias: comparison of methods for a single predictor variable. J Royal Statist Soc: Series A 2000; 163: 173-189
    PubMedGoogle Scholar
  • 8 MacMahon S, Peto R, Cutler J, Collins R, Sorlie P, Neaton J, Abbott R, Godwin J, Dyer A, Stamler J. Blood pressure, stroke, and coronary heart disease. Part 1, Prolonged differences in blood pressure: prospective observational studies corrected for the regression dilution bias. Lancet 1990; 335: 765-774
    CrossrefPubMedGoogle Scholar
  • 9 Collaboration TFS . Correcting for multivariate measurement error by regression calibration in meta-analyses of epidemiological studies. Statist Med 2009; 28: 1067-1092
    CrossrefPubMedGoogle Scholar
  • 10 Azizi F, Amouzegar A, Delshad H, Tohidi M, Mehran L, Mehrabi Y. Natural course of thyroid disease profile in a population in nutrition transition: tehran thyroid study. Arch Iran Med 2013; 16: 418-423
    PubMedGoogle Scholar
  • 11 Rosner B, Willett WC, Spiegelman D. Correction of logistic regression relative risk estimates and confidence intervals for systematic within-person measurement error. Statist Med 1989; 8: 1051-1069 discussion 1071–1053
    PubMedGoogle Scholar
  • 12 Frost C, White IR. The effect of measurement error in risk factors that change over time in cohort studies: do simple methods overcorrect for ‘regression dilution’?. Int J Epidemiol 2005; 34: 1359-1368
    CrossrefPubMedGoogle Scholar