Development and Internal Validation of a Prognostic Model for 4-Year Risk of Metabolic Syndrome in Adults: A Retrospective Cohort Study

Figures & data

Table 1 Characteristics of Participants

Candidate predictor	Derivation Cohort (N=6793) n (%)/Mean±SD	Missing Values n (%)
Sex
Female	2377 (34.99%)	0
Male	4416 (65.01%)	0
Age	41.488±10.411	0
WBC	5.794±1.459	0.147
Hb	143.245±14.858	0.147
LC	33.296±7.288	0.147
NGC	3.317±1.088	0.147
TC	4.419±0.846	0.471
UA	303.305±86.971	0.501
Weight	63.28±10.538	30.561
Height	166.225±7.722	30.59
BMI	22.802±2.726	30.59
AST	19.754±9.33	38.216
ALT	21.453±20.271	0.118
MCV	91.757±4.967	0.147
MCH	31.486±1.962	0.147
HCT	41.752±4.211	0.147
Outcome Indicator	Participants with Events (N=1750) n (%)/Mean±SD	Missing Values n (%)
WC	86.445±6.972	16.971
TG	1.969±1.383	0.114
HDL-c	1.151±0.29	0.114
SBP	127.648±14.457	4.914
DBP	78.301±10.582	4.914
FPG	5.306±0.993	0.114
Medication for hypertension	269 (15.37%)	0
Medication for FPG	39 (2.23%)	0
Medication for TG or HDL-c	617 (35.26%)	0

Figure 1 Participant flow.

Figure 2 Predictors selection using the least absolute shrinkage and selection operator (LASSO) binary logistic regression model. (A) LASSO coefficient profiles of the 14 texture features. A coefficient profile plot was produced against the log (λ) sequence. (B) Hyperparameter (λ) selection in the LASSO model used 10-fold cross-validation via minimum criteria. Dotted vertical lines were drawn at the optimal values by using the minimum criteria and the 1 standard error of the minimum criteria (the 1-SE criteria).

Figure 3 Calibration plot.

Notes: Apparent refers to apparent performance for calibration; bias-corrected refers to optimism-corrected in internal validation; bootstrap=100.

Figure 4 Decision-curve analysis.

Notes: The horizontal lines (labeled “none”) refer to the expected net benefit without any treatment or intervention which indicates no benefits (net benefit>0) or harms (net benefit<0) from this strategy. The slanted vertical line (labeled “all”) refers to treatment or intervention provided for all patients. False positives are weighted more, and the net benefit becomes negative for providing treatment or intervention for all patients.

Figure 5 A clinical example of the application of the calculator 4-year risk of 60% based on the prognostic prediction model for a man, age 50, total cholesterol of 7 mmol/l, serum uric acid of 400 μmol/l, alanine transaminase of 30 U/L, body mass index (22.2 kg/m²).