Mozaffer Rahim Hingorjo, Sadiqa Syed, Masood Anwar Qureshi.
Overweight and obesity in students of a Dental College of Karachi: lifestyle influence and measurement by an appropriate anthropometric index.
J Pak Med Assoc Jan ;59(8):528-32.
Objective: To compare body mass index (BMI), waist circumference (WC), and body fat percentage (%BF), as index of overweight and obesity in young adults. We also intended to find an association between lifestyle behaviours and obesity. Methods: A cross-sectional study was conducted at Fatima Jinnah Dental College, Karachi, during 2007 to 2008, with 192 first year dental students, (18-21years) of high socioeconomic class. All were questioned regarding lifestyle behaviours. Overweight and obesity were estimated by measuring %BF, BMI, and WC. For %BF, skinfold thickness was measured using skinfold calipers. BMI > 23.0-24.9 kg/m2 was taken as overweight and > 25.0 kg/m2 as obese (Asians criteria proposed by Western Pacific Regional Office of World Health Organization). WC using Asian cutoff values for overweight and obesity were: males > 78cm and > 90cm; females > 72cm and > 80cm, respectively. Body fat percentage used to define overweight and obesity was: males 22.1-27.0 and >27.1; females 27.1-32.0 and >32.1, respectively. Pearson`s correlation was done between the BMI, WC and %BF with statistical significance taken at P <0.01. Results: BMI (Expressed as mean±SD) in males and females was 23.82±3.88 and 20.98±4.12 respectively. WC was 83.63±10.20cm in males and 70.22±9.36cm in females. %BF was 22.32±6.27 in males and 28.73±6.65 in females, with an overall 60.8% females and 44.4% males found to be overweight or obese. Obesity was underpredicted by BMI when compared to skinfold calipers method. The obese were seen to skip breakfast more often [odds ratio (OR):2.39], take frequent snacks (OR:1.58), watch television more (OR:1.58), and were physically less active than their non-obese counterparts. Conclusion: Body fat percentage using skinfold caliper is a reliable index of obesity. Lack of sleep and skipping of breakfast, are prominent promoters of obesity, in addition to other lifestyle behaviours (JPMA 59:528; 2009).
USER COMMENTS
AOA, I think the objectives and Statistical analysis are not matched.
The best analysis for first objective was to fine sensitivity and specificity. OR are also good to report but not for first objective.. Thanks a lot
Asif Hanif Biostatistician and Research Methodologist.
Posted by: biostat on Sep 2009
Hi writer, well done clinically, but some weak areas statistically. Your first objective could only be proved by multiple regression model or logistic model. ANOVA can also be used.
when you use OR you must write confidence interval for its statistical significance.
Pearson correlation can be used to determine the relationship between variables BMI. BF, WC.
My brother (ID BIOSTATS) suggested sensitvity and speificity which is only in the case of Diagnostic and screening procedures.
Posted by: ibrahim_ap on Sep 2009
The prevalence of obesity is rising in our society especially children and young adults of middle and upper social class. girls are more often affected with many social and psychological impacts.
Posted by: bvhospital on Nov 2009
Several issues, first, there is no gold-standard test in this study to determine body fat. While authors write that measuring triceps fold thickness tells about percentage body fat, it is not true. These are easy to use instruments, but have lot of measurement error. Thus, it is not possible to say that one instrument is better than the other unless we exactly measure body fat content. Authors may have tried to use regression calibration or similar other methods to correct for measurement error.
Second, the least one need to report for the first objective is sensitivity and specificity of the tests against 'standard test' (which in this case could be a regression calibration corrected instrument of measurement). Better would be to actually have ROC curve and give c-statistic. However appropriate reporting would have been to do report misclassification rates.
Third, depending upon whether the outcome is continuous (BMI, skin-fold thickness, WC) or dichotomous (obesity) multiple linear regression or logistic regression can be used. Regardless whichever test is used, 95%CI should be reported.
Posted by: rqayyum on Dec 2009
Several issues, first, there is no gold-standard test in this study to determine body fat. While authors write that measuring triceps fold thickness tells about percentage body fat, it is not true. These are easy to use instruments, but have lot of measurement error. Thus, it is not possible to say that one instrument is better than the other unless we exactly measure body fat content. Authors may have tried to use regression calibration or similar other methods to correct for measurement error.
Second, the least one need to report for the first objective is sensitivity and specificity of the tests against 'standard test' (which in this case could be a regression calibration corrected instrument of measurement). Better would be to actually have ROC curve and give c-statistic. However appropriate reporting would have been to do report misclassification rates.
Third, depending upon whether the outcome is continuous (BMI, skin-fold thickness, WC) or dichotomous (obesity) multiple linear regression or logistic regression can be used. Regardless whichever test is used, 95%CI should be reported.
Posted by: rqayyum on Dec 2009
Several issues, first, there is no gold-standard test in this study to determine body fat. While authors write that measuring triceps fold thickness tells about percentage body fat, it is not true. These are easy to use instruments, but have lot of measurement error. Thus, it is not possible to say that one instrument is better than the other unless we exactly measure body fat content. Authors may have tried to use regression calibration or similar other methods to correct for measurement error.
Posted by: rqayyum on Dec 2009
PakMediNet -Pakistan's largest Database of Pakistani Medical Journals - http://www.pakmedinet.com