Underestimation of Risk Associations Due to Regression Dilution in Long-Term Follow-Up of Prospective Studies

Nuffield Department of Clinical Medicine, Radcliffe Infirmary, Oxford, England.
American Journal of Epidemiology (Impact Factor: 5.23). 09/1999; 150(4):341-53. DOI: 10.1093/oxfordjournals.aje.a010013
Source: PubMed


In prospective studies, disease rates during follow-up are typically analyzed with respect to the values of factors measured during an initial baseline survey. However, because of "regression dilution," this generally tends to underestimate the real associations of disease rates with the "usual" levels of such risk factors during some particular exposure period. The "regression dilution ratio" describes the ratio of the steepness of the uncorrected association to that of the real association. To assess the relevance of the usual value of a risk factor during particular exposure periods (e.g., first, second, and third decades) to disease risks, regression dilution ratios can be derived by relating baseline measurements of the risk factor to replicate measurements from a reasonably representative sample of study participants after an interval equivalent to about the midpoint of each exposure period (e.g., at 5, 15, and 25 years, respectively). This report illustrates the impact of this time interval on the magnitude of the regression dilution ratios for blood pressure and blood cholesterol. The analyses were based on biennial remeasurements over 30 years for participants in the Framingham Study (Framingham, Massachusetts) and a 26-year resurvey for a sample of men in the Whitehall Study (London, England). They show that uncorrected associations of disease risk with baseline measurements underestimate the strength of the real associations with usual levels of these risk factors during the first decade of exposure by about one-third, the second decade by about one-half, and the third decade by about two-thirds. Hence, to correct appropriately for regression dilution, replicate measurements of such risk factors may be required at varying intervals after baseline for at least a sample of participants.

Download full-text


Available from: Sarah Lewington,
23 Reads
  • Source
    • "While late-life leptin levels seem to be associated with dementia risk, there was no evidence for an association between midlife leptin levels and late-life dementia risk in Swedish women followed for 31 to 35 years [13]. This raises the possibility that reverse causation leads to an association between late-life leptin and dementia-related outcomes due to declining leptin levels and weight loss linked with neuropathological changes [14], or that midlife leptin levels are a poor indicator of overall leptin levels during follow-up or late-life leptin levels due to regression dilution bias [28]. Our results make the former less likely by demonstrating that the associations observed in late-life are not accounted for by midlife BMI or change in BMI. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: US studies suggest that leptin, a fat-derived hormone, may be protective against the development of dementia. Objective: To investigate the complex relationship between leptin levels and cognitive decline in elderly Italians. Methods: We studied circulating fasting leptin levels in 809 elderly adults free from dementia who participated in the prospective Italian population-based InCHIANTI study between 1998 and 2009 (mean follow-up of 8.0 years). Global cognitive decline was defined as a reduction of ≥5 points on the Mini-Mental State Examination (MMSE). Trail-Making Tests A and B were also incorporated, with cognitive decline defined as discontinued testing or the worst 10% of change from baseline. We also investigated whether any association could be explained by midlife weight and whether cognitive decline was associated with changing leptin levels. Results: The multivariate adjusted relative risk ([RR]; 95% confidence interval [CI]) of cognitive decline on the MMSE was 0.84 (95% CI 0.73-0.97) in relation to baseline sex-standardized log-leptin levels. High leptin levels showed a non-significant trend toward a reduced risk of decline on the Trail-Making Tests A (RR = 0.85, 95% CI 0.71-1.02) and B (RR = 0.90, 0.79-1.02). Adjusting for midlife weight or change in weight did not alter the pattern of results, and cognitive decline was not associated with changing leptin levels. Conclusions: High leptin levels were independently associated with a reduced risk of cognitive decline in elderly Italians.
    Journal of Alzheimer's disease: JAD 12/2014; 44(4). DOI:10.3233/JAD-141836 · 4.15 Impact Factor
  • Source
    • "In the analysis of exposure categorized in quintiles, regression dilution ratios (RDR) were calculated based on repeated health examinations in 133,820 subjects in the full Me-Can database in order to adjust RRs for random errors in the measurement of exposure variables at baseline [27,28], this process has been described in detail previously [13]. In brief, only measurements in the same cohort with the same fasting time before any incident cancer diagnosis were used. "
    [Show abstract] [Hide abstract]
    ABSTRACT: Obesity is associated with an increased risk of esophageal adenocarcinoma (EAC) and a decreased risk of esophageal squamous cell carcinoma (ESCC). However, little is known about the risk of EAC and ESCC related to other metabolic risk factors. We aimed to examine the risk of EAC and ESCC in relation to metabolic risk factors, separately and combined in a prospective cohort study. The Metabolic Syndrome and Cancer cohort includes prospective cohorts in Austria, Norway and Sweden, with blood pressure, lipids, glucose and BMI available from 578 700 individuals. Relative risk (RR) for EAC and ESCC was calculated using Cox's proportional hazards analysis for metabolic risk factors categorized into quintiles and transformed into z-scores. The standardized sum of all z-scores was used as a composite score for the metabolic syndrome (MetS). In total, 324 histologically verified cases of esophageal cancer were identified (114 EAC, 184 ESCC and 26 with other histology). BMI was associated with an increased risk of EAC (RR 7.34 (95% confidence interval, 2.88-18.7) top versus bottom quintile) and negatively associated with the risk of ESCC (RR 0.38 (0.23-0.62)). The mean value of systolic and diastolic blood pressure (mid blood pressure) was associated with the risk of ESCC (RR 1.77 (1.37-2.29)). The composite MetS score was associated with the risk of EAC (RR 1.56 (1.19-2.05) per one unit increase of z-score) but not ESCC. In accordance with previous studies, high BMI was associated with an increased risk of EAC and a decreased risk of ESCC. An association between high blood pressure and risk fo ESCC was observed but alcohol consumption is a potential confounding factor that we were not able to adjust for in the analysis. The MetS was associated with EAC but not ESCC. However this association was largely driven by the strong association between BMI and EAC. We hypothesize that this association is more likely to be explained by factors directly related to obesity than the metabolic state of the MetS, considering that no other metabolic factor than BMI was associated with EAC.
    BMC Cancer 02/2014; 14(1):103. DOI:10.1186/1471-2407-14-103 · 3.36 Impact Factor
  • Source
    • "Our study also has some limitations. First, a single measurement of plasma markers might have caused regression dilution bias (Clarke et al, 1999) and precluded us from evaluating associations between the long-term levels of these cytokines and risk of CRC. However, these markers have been shown to be generally stable over time (Pischon et al, 2003b; Platz et al, 2010). "
    [Show abstract] [Hide abstract]
    ABSTRACT: Background: Chronic inflammation may mediate risk of colorectal cancer (CRC); however, the association between circulating inflammatory markers and risk of CRC has been inconsistent. Methods: We prospectively evaluated the association of plasma C-reactive protein (CRP), interleukin-6 (IL-6), and the soluble tumour necrosis factor receptor 2 (sTNFR-2) with incident CRC among 274 cases and 532 matched controls nested in the Health Professionals Follow-up Study. Results: Multivariate relative risk (RR) of CRC comparing the extreme quartiles of plasma IL-6 was 1.54 (95% confidence interval (CI), 0.99–2.40; Ptrend=0.02). However, after excluding cases diagnosed within 2 years of blood draw, this association was not statistically significant (RR=1.26, 95% CI, 0.78–2.05; Ptrend=0.21). In analyses restricted to cases diagnosed at least 2 years after blood draw, the association of IL-6 with CRC appeared to differ by body mass index such that the significantly positive association was only present among lean individuals (Pinteraction=0.03). We did not observe any significant association between CRP or sTNFR-2 and CRC. Conclusion: Plasma inflammatory markers are not generally associated with risk of CRC among men. However, the possibility that plasma IL-6 is associated with increased risk of CRC among lean men requires further investigation.
    British Journal of Cancer 04/2013; 108(9). DOI:10.1038/bjc.2013.172 · 4.84 Impact Factor
Show more