Sir,
I was interested to read the paper by Sabour and colleagues published in the March 2011edition of the Journal of Research in Medical Sciences.[1] The purpose of the authors was to assess obesity predictors of people with spinal cord injury (SCI) according to age, time since injury, level, and completeness of injury. Obesity predictors were measured in 162 patients in a cross-sectional study.
For prediction studies, we need two different cohort datasets or at least split one cohort dataset to develop our prediction model and then validate it. So, using a cross-sectional design, we cannot say anything about prediction.
Waist circumference (WC) was measured at the level of the lowest rib and classified based on standard classification (men: >102 cm, women: >85 cm). There were 131 (80.9%) male patients and 31 (19.1%) female patients.
Standard classification for WC is not applicable for the Iranian population as there is no difference between men and women regarding mean WC.[2] So, applying such a classification will actually lead to overestimation of obesity in women as well as underestimation of that in men, which means we will be faced with differential misclassification of the outcome and biased results. Therefore, internal validity of such studies can be questionable.[3]
Self-reported height and weight were used in this study which means that validity (accuracy) and reliability (precision) of the data cannot be guaranteed. As these variables have been used to calculate other variables of the study such as body mass index (BMI) and so on, any interpretation based on these variables cannot be clinically correct and useful. Most of the time, misleading results (if we do not call it biased) are the main outcome of such researches.[4]
Moreover, for prediction purposes, using the Pearson correlation test is one of the common mistakes in reliability analysis as well as prediction researches.[3,4]
REFERENCES
- 1.Sabour H, Javidan AN, Vafa MR, Shidfar F, Nazari M, Saberi H, et al. Obesity predictors in people with chronic spinal cord injury: An analysis by injury related variables. J Res Med Sci. 2011;16:335–9. [PMC free article] [PubMed] [Google Scholar]
- 2.Hadaegh F, Zabetian A, Sarbakhsh P, Khalili D, James WP, Azizi F. Appropriate cutoff values of anthropometric variables to predict cardiovascular outcomes: 7.6 years follow-up in an Iranian population. Int J Obes (Lond) 2009;33:1437–45. doi: 10.1038/ijo.2009.180. [DOI] [PubMed] [Google Scholar]
- 3.Jeckel JF, Katz DL, Elmore JG, Wild DM. The study of causation in epidemiologic investigation and research. In: Jeckel JF, editor. Epidemiology, Biostatistics and Preventive Medicine. 3rd ed. Philadelphia, PA, United State: Saunders Elsevier; 2007. pp. 64–6. [Google Scholar]
- 4.Rothman KJ, Greenland S, Lash TL. Cohort studies. In: Rothman KJ, editor. Modern Epidemiology. 3rd ed. Baltimore, United States: Lippincott Williams and Wilkins; 2008. pp. 79–85. [Google Scholar]