BMI and an Anthropometry-Based Estimate of Fat Mass Percentage Are Both Valid Discriminators of Cardiometabolic Risk: A Comparison with DXA and Bioimpedance

Table 5

Prediction/discrimination of impaired fasting glucose and the metabolic syndrome with degree of obesity as defined by dual-energy X-ray absorptiometry (DXA) bioimpedance analysis (BIA), an anthropometry-based estimate of fat mass percentage (FM%-equation) and BMI.

										ROC analyses^n
Reference method/model^a	New method/model^b	𝑛^c		Reclassification index, %^f				IDI, %^k		Men		Women
		Cases^d	Non-cases^e	Net^g	𝑃^hnri	Cases^i	Non-cases^j	𝐼^lintegr.	𝑃^midi	Δ AUC^o	𝑃^p	Δ AUC	𝑃
Impaired fasting glucose (≥5.6 mmol/L = 100 mg/dL)
DXA	BIA InBody^q	164	249	−6%	0.181	−7%	1%	−0.5%	0.506	−0.03	0.102	−0.01	0.394
	BMI	191	276	−2%	0.727	−4%	2%	0.3%	0.723	−0.04	0.286	−0.02	0.394
	Estimate^r	191	276	−1%	0.771	−4%	2%	0.2%	0.809	−0.03	0.404	−0.01	0.597
BIA InBody	BMI	164	249	2%	0.752	1%	0%	0.2%	0.796	−0.01	0.888	−0.01	0.744
	Estimate	164	249	1%	0.769	1%	1%	−0.1%	0.882	0.00	0.890	0.00	0.981
BMI	Estimate	191	276	0%	1.000	0%	0%	−0.1%	0.733	0.01	0.514	0.01	0.386
Impaired fasting glucose (≥6.1 mmol/L = 110 mg/dL)
DXA	BIA InBody	70	343	−1%	0.901	−4%	3%	3.5%	0.009	−0.01	0.584	0.01	0.462
	BMI	80	387	6%	0.394	3%	4%	3.2%	0.009	0.00	0.900	0.00	0.918
	Estimate	80	387	3%	0.616	0%	3%	2.6%	0.023	0.02	0.438	0.01	0.796
BIA InBody	BMI	70	343	7%	0.341	6%	1%	−0.7%	0.609	0.02	0.648	−0.02	0.504
	Estimate	70	343	2%	0.754	1%	1%	−1.5%	0.205	0.04	0.253	−0.01	0.799
BMI	Estimate	80	387	−3%	0.251	−3%	-1%	−0.6%	0.176	0.02	0.315	0.01	0.248
Metabolic syndrome (AHA/NHBLI)^s
DXA	BIA InBody	144	268	−4%	0.400	−6%	2%	−0.7%	0.691	−0.03	0.120	0.01	0.625
	BMI	165	301	4%	0.461	0%	4%	2.5%	0.257	−0.02	0.610	0.02	0.309
	Estimate	165	301	3%	0.519	−1%	4%	1.7%	0.407	−0.02	0.466	0.02	0.329
BIA InBody	BMI	144	268	3%	0.595	2%	1%	0.9%	0.662	0.01	0.697	0.01	0.429
	Estimate	144	268	4%	0.480	2%	1%	0.8%	0.681	0.01	0.812	0.01	0.409
BMI	Estimate	165	301	−1%	0.577	−1%	0%	−0.7%	0.252	−0.01	0.622	0.00	0.958

aMethod of measurement, based on which participants are classified in categories of obesity.
^bDifferent method of estimating obesity, the predictive power of which is compared to reference model/reference method.
^cNumber of participants.
^dNumber of participants that are positive with regard to respective outcome.
^eNumber of participants that are negative with regard to respective outcome.
^fPercentage improvement (+) or deterioration (−) in predictive power of new model compared to reference model. Categories of obesity/FM% as independent variable.
^gNet reclassification of cases + net reclassification of non-cases. A positive number denotes increased predictive power for the new model.
^hLikelihood of net reclassification index to be 0, that is, the new model showing no improvement/deterioration over reference model.
ⁱNet reclassification of cases = percentage of cases reclassified by the new model into a higher risk category − percentage of cases reclassified by the new model into a lower risk category.
^jNet reclassification of non-cases = percentage of non-cases reclassified by the new model into a lower risk category − percentage of non-cases reclassified by the new model into a higher risk category.
^kIntegrated discrimination improvement (+) or deterioration (−) of new model compared to reference model. Categories of obesity/FM% as independent variable in an age-adjusted model.
^lMean difference in predicted individual probabilities between cases and non-cases for two models. A positive number denotes increased predictive power for the new model.
^mLikelihood of net reclassification index to be 0, that is, the new model showing no improvement/deterioration over reference model.
ⁿMeasures of obesity (BMI/FM%) as continuous variable in a logistic regression model predicting respective outcomes.
^oDifference in area under curve of receiver operating characteristic compared to reference method.
^pProbability of 0-hypothesis (no difference).
^qEstimation of FM% with bioimpedance device InBody (720) (Biospace, Korea).
^rAnthropometry-based estimate; arithmetic mean of FM% estimations according to prediction methods Deurenberg et al. [12], Gallagher et al. [15], and Larsson et al. [14].
^sDefinition of metabolic syndrome suggested by the common task force from the IDF and the American Heart Association/National Heart, Lung and Blood Institute (AHA/NHBLI) [17].