Table 5.

Statistical analysis, results and conclusions of the included articles.

Reference	Statistical Analysis	Results	Conclusions
Bragança et al., 2018 [4]	Accuracy: technical error of measurement (TEM) and relative technical error of measurement (%TEM). Reliability: Relative: interclass correlation coefficient (ICC) and reliability coefficient (R). Absolute: standard error of measurement (SEM) and coefficient of variation (CV).	Accuracy: TEM values < 2 cm. Higher manual technical accuracy (slightly lower values). %TEM: Only chest length obtained a value > 1.5% using the manual technique, while seven measurements did so using the 3D technique. Reliability: Relative (manual: ICC 0.80–0.99 and 3D: ICC 0.91–0.99). When comparing both methods, all the measurements, except neck circumference, presented slightly higher values using the manual technique. Very similar results for R (R > 0.95). Absolute: 3D technique results less reliable (higher SEM values), except neck circumference. According to CV, none of the methods performs well because for all measurements, the results were >5%.	Despite being considered sufficiently accurate and reliable for certain applications, the 3D scanner showed, for almost all measurements, a different result than obtained using the manual technique.
Adler et al., 2017 [10]	Validity: Pearson correlation coefficient and Bland-Altman plots. Q-Q plots to examine differences between 3D and ADP for body volume. Reliability: differences in 3D measurements, calculated as scan1 and scan2 and ICC.	Validity: 3D body volume and ADP strongly correlated (R = 0.99). ADP body volume 72.2 L and 3D body volume higher by 1.1 L (p < 0.001), 1.0 L (p < 0.001) and 2.5 L (p < 0.001) in standard, relaxed and exhaled positions, respectively. %MG 3D and ADP well correlated (R = 0.79), %MG ADP 23.75 and %MG 3D higher by 7.0% (p < 0.001), 6.6% (p < 0.001) and 16.6% (p < 0.001) for standard, relaxed and exhaled positions. Reliability: high for body volume, with a mean difference of 0.1 L, 0.2 L and 0.2 L for standard, relaxed and exhaled positions, respectively, and ICC > 0.98. %MG, standard position, mean difference of -0.4%, relaxed position 0.2%, and exhaled 0.3%, with ICCs of 0.982, 0.983 and 0.945, respectively.	Good agreement between 3D and ADP. Good validity and excellent reliability of the 3D scan.
Bourgeois et al., 2017 [11]	Comparison of measurements between methods: paired t-tests. Associations between methods: linear regression analysis. Bland-Altman plots.	Hip circumference: significant difference between conventional anthropometry and 3D scan (#1 and #2) (p < 0.0001). Waist: significant difference between conventional anthropometry and 3D scan (#2 and #3) (p < 0.0001). Arm: significant difference between conventional anthropometry and 3D scan (#1 and #3) (p < 0.0001). Thigh: significant difference between conventional anthropometry and 3D scan (#1 and 2) (p < 0.0001). Significant correlations between methods (R = 0.71–0.96; p < 0.0001 for all). Total body volume: significant difference between ADP and 3D measurements (p < 0.0001). Body volume measured by the three 3D scans were highly correlated with ADP volumes (R = 0.99 for all). Significant bias (p < 0.05) of −3.4 L, −2.4 L and −9.1 L for 3D scan (#1), (#2) and (#3), respectively; 3D systems underestimate body volume.	Reproducible measurements correlate well with reference methods.
Medina-Inojosa et al., 2016 [12]	Reproducibility: intraobserver and interobserver variability and paired t-test. Comparison between methods: unpaired t-test. ICC and Bland-Altman.	Intraobserver variations (reproducibility): 3.1 cm waist and 1.8 cm hip. Interobserver variations (precision): 3.9 cm waist and 2.4 cm hip. 3D scanner variability: 1.3 cm waist and 0.8 cm hip. Significant difference between methods (p < 0.05). ICC > 0.95 for all.	A 3D scanner is a more reliable and reproducible way to measure waist and hip circumference.
Ng et al., 2016 [13]	Agreement between methods: univariate linear regressions. Measurement biases between methods: Student t-test. % CV RMSE for paired test-retest measurements of the 3D scanner. R2.	Strong associations between methods for waist and hip circumference (R = 0.95 and 0.92, respectively). Significant differences of 1.75 cm for waist and 3.17 cm for hip between 3D and conventional anthropometry. Strong associations between 3D scan and ADP and DXA for total body volume (R = 0.99 and 0.97, respectively), with significantly lower volume measured by 3D scan compared to ADP (−4.15 L).	This study supports the use of 3D scanning as an accurate, reliable and automated surrogate for other methods.
Ng et al., 2019 [14]	Model accuracy/precision: R2 and RMSE. Measurement precision: RMSE and CV (%).	Precision of body composition comparing 3D scanner was DXA was as follows: fat mass, R = 0.88 male, 0.93 female; visceral fat mass, R = 0.67 male, 0.75 female. The test precision (test-retest) of the 3D scan for body fat was as follows: mean square error = 0.81 kg male, 0.66 kg female. Visceral fat according to 3D scan was as accurate (% CV = 7.4 for males, 6.8 for females) as that obtained using DXA (% CV = 6.8 for males, 7.4 for females).	The 3D estimates may be somewhat less accurate than DXA estimates.
Brooke-Wavell et al., 1994 [15]	Intraobserver and interobserver variability: standard error of measurement. Means, standard error of the mean and t-tests.	Comparison between methods (reliability): Women: significant differences (p < 0.05) between conventional anthropometry and 3D scan for neck, chest, waist width, waist depth and waist height. Men: significant difference between conventional anthropometry and 3D scan for neck circumference, chest and waist depth. Good agreement between methods (r = 0.964–0.998). Intraobserver Diff: 7 mm (larger) for waist circumference (manual). Only neck circumference, larger for 3D scan (5.3 mm vs. 3.0 mm). Interobserver difference (accuracy): 3.0–13.1 mm manual and 1.3–8.5 mm 3D scan.	3D measurements and anthropometry were generally similar. Larger interobserver differences for manual technique, lower precision.
Weiss et al., 2009 [16]	- 1	Intraobserver variations (reproducibility): researcher 1: 0.37 cm between repetitions, researcher 2: 0.406 cm and 3D: 0.171 cm. Very high correlations (r > 0.99), although higher 3D scan correlations (researcher 1 and 2 = 0.995 vs. 3D = 0.9988). Interobserver variations (precision): thigh circumference, variance 20.5% higher than variance for 3D scan. Abdominal circumference, variance 231.3% greater than the variance for the 3D scan.	Greater precision and reproducibility of the measurement with the use of the 3D scanner.
Pepper et al., 2010 [17]	Reproducibility: ICC and CV. Paired t-tests, correlation coefficients and Bland-Altman plots.	ICC > 0.99 for all circumferences measured by 3D. CVs showed little difference between intraindividual measurements, showing high agreement between repeated measurements (CV 0.53%-1.68%). No significant difference between methods for waist and hip (3D: 87.87 cm and 104.15 cm vs. conventional anthropometry: 87.73 cm and 104.39 cm, respectively p > 0.05). Highly correlated measurements (waist: r = 0.998 and hip: r = 0.984; p < 0.01).	3D scanner reliable and valid technique compared to conventional anthropometry.
Harbin et al., 2017 [18]	Level of agreement between methods: Bland-Altman graphs. Mean differences in %MG estimation: multivariate ANOVA.	Significant difference (p < 0.001) between %MG measured by 3D scan and the other methods (3D %MG: 18.1%; hydrostatic weighing %MG: 22.8%; bioelectrical impedance %MG: 20.1%; folds %MG: 19.7%; circumferences %MG: 21.2%). Bonferroni post hoc analysis revealed that the %MG estimated by 3D scan was significantly lower than that estimated by all other techniques.	Advances must be made before 3D scans can be designated as an accurate method.
Bragança et al., 2017 [3]	Comparison between methods: paired t-test.	Significant difference between various 3D measurements and conventional anthropometry (p < 0.001): shoulder width, back length, waist circumference, hip circumference, thigh circumference, knee circumference and ankle circumference.	Reliability and accuracy depend on the ability to remain static.
Vonk & Daanen, 2015 [19]	Repeatability: (ICC, ICC < 0.80: measurements with low repeatability. Accuracy: SEM, SEM > 10mm: not accurate enough. Validity: paired t-tests.	SizeStream scan: 120 measurements: ICC > 0.90 and 20 measurements: ICC < 0.80. Mean SEM: 10.1 mm. Validity: 6 measurements by 3D and conventional anthropometry: significant difference (p < 0.001) (chest, waist, hip, wrist, neck-bust distance and arm length). However, strong and significant correlations for chest (r2= 0.95; p < 0.001), waist (r2 = 0.92; p < 0.001) and hip (r2 = 0.96; p < 0.001). Poikos scanner: 14 measurements: ICC < 0.80 and 2 measurements: ICC > 0.90. Mean SEM: 54.5 mm. Significant difference only for waist (p < 0.001), but weak correlations (R2 < 0.60).	Only three of the six measurements compared could be validated (SizeStream scanner). Poikos is promising but less repeatable and valid than the SizeStream scanner.
Tinsley et al., 2019 [20]	Accuracy: ICC and RMS-%CV. Validity (regional and total volumes only): one-way ANOVA. Coefficient of determination (R2). RMSE. Bland-Altman with linear regression to evaluate the degree of proportional bias.	Accuracy: circumferences (ICC from 0.974 to 0.999) and volumes (ICC from 0.952 to 0.999). Average of four scans for RMS-%CV: circumferences (1.1% to 1.3%) and body volume (1.9% to 2.3%). Circumference highest accuracy: hip (RMS-%CV < 1% for all), waist (0.7–1.6%), thigh (0.8–1.4%) and arm (1.4–2.8%). Volume highest accuracy: total (RMS-%CV < 1% for all), torso volume (approx. 1.2%), leg (approx. 2.5%) and arm (3–5%). Validity: very strong linear relationships between methods for total body volume (R: 0.98–1.0), but SizeStream significantly overestimated it, and Styku underestimated it. Stronger relationship between 3D and DXA for torso volume (R: 0.96–0.97) than arm and leg volume (R: 0.65–0.93). However, all 3D scans significantly overestimated torso volume and underestimated arm and leg volume.	Excellent accuracy; however, relatively poor validity for total and regional body volume.
Ladouceur et al., 2017 [21]	Concurrent validity between methods: Pearson product moment correlation coefficient (PPMC) and paired t-test. Systematic error between the two methods: paired t-test. Bland-Altman.	Significant difference between conventional anthropometry and 3D measurements (p = 0.000).	The results of this study have shown promise for the future.
Ramos-Jiménez et al., 2018 [22]	Differences between methods: t-test for independent samples. Significance of finding differences, was analyzed using Cohen’s d. Linear regression for strength of associations.	3D measurements highly correlated with those of conventional anthropometry and plestimography. (R ≥ 0.75) but significantly different for all (p < 0.01).	Valid and reliable measurements when evaluating adult individuals; however, it is important to minimize body motion.
Kuehnapfel et al., 2016 [23]	Concordance of paired measurements: overall concordance correlation coefficient (OCCC). Illustration of results: scatter and Bland-Altman plots.	Validity: excellent for height (OCCC = 0.995), weight (OCCC = 1.00), waist (OCCC = 0.982), hip (OCCC = 0.938) and calf (OCCC = 0.988); good for arm (OCCC = 0.720); moderate for thigh (OCCC = 0.557). Notable bias between anthropometry and 3D measurements.	Reliability of 3D measurements was generally excellent or good, with some exceptions.
Koepke et al., 2017 [24]	Repeatability and agreement between repeated measurements within each method: mean differences, ICC, precision, and paired t-tests. Agreement between methods: mean differences (mSM, mMM), correlation coefficients, and paired t-tests. In addition, Lin’s coefficient of concordance.	3D: no significant difference between repeated measurements and strong correlations: chest: 0.981; p = 0.486; waist: 0.993; p = 0.397; buttocks: 0.997; p = 0.052; hip: 0.994; p = 0.280. Manual: chest: 0.968; p < 0.001; waist: 0.990; p < 0.001; buttocks: -0.955; p = 0.018; hip: 0.972; p = 0.186. Precision higher than 2.50 cm, up to 8.19 cm, indicating high disagreement. CCC remains high (>0.94) for height and waist. CCC = 0.781 for chest, 0.784 for hip and 0.258 for buttocks. However, significant difference between methods, (chest: +3.88cm p < 0.001); (waist: +1.17 cm p < 0.001); (buttocks: +12.62 p < 0.001); (hip: +4.37 p < 0.001).	Better accuracy and repeatability for 3D scanner. Highly correlated data, but important systematic differences. Therefore, the two techniques are not directly equivalent.
Lu & Wang., 2010 [25]	Paired t-test and MAD (mean absolute difference) between scan-derived measurement and manual measurement for each dimension as a measure of accuracy performance.	Accuracy: significant difference between methods for chest circumference (p = 0.0008) and waist circumference (p = 0.0090) but not hip circumference (p = 0.5974). Most MADs between scan-derived and manual measurements exceeded ISO 20685 criteria. Accuracy: MADs of all repeated measurements were less than 7 mm. When compared to the maximum allowable interobserver error reported in ANSUR, the accuracy of the 3D measurements was higher than that of the manual measurements.	3D measurements more accurate than manual measurements.

¹ Information not reported in the article.