TABLE 2.

Correlations between the MDS and metabolite scores in the derivation and test sets: the Fenland Study, 2005–2015¹

Derivation set			Test set2
Strata	n	ρ 3	Strata	n	ρ
Metabolite score
Random half 1	5382	0.47	Random half 2	5424	0.43
Random half 2	5424	0.45	Random half 1	5382	0.44
			Pooled4		0.43
Metabolite score from backwards regression
Random half 1	5382	0.45	Random half 2	5424	0.42
Random half 2	5424	0.43	Random half 1	5382	0.42
			Pooled		0.42
Metabolite score stratified by age, y
≤48.5	5424	0.48	>48.5	5382	0.42
>48.5	5382	0.46	≤48.5	5424	0.43
			Pooled		0.42
Metabolite score stratified by sex
Men	5031	0.46	Women	5775	0.40
Women	5775	0.43	Men	5031	0.43
			Pooled		0.43
Metabolite score stratified by BMI, kg/m2
≤26.2	5470	0.47	>26.2	5336	0.37
>26.2	5336	0.40	≤26.2	5470	0.43
			Pooled		0.43
Metabolite score stratified by smoking status
Former or current	4934	0.50	Never	5874	0.42
Never or current	7236	0.46	Former	3572	0.42
Never or former	9446	0.44	Current	1362	0.40
			Pooled		0.43
Metabolite score stratified by study site5
Cambridge and Ely	7794	0.46	Wisbech	3012	0.38
Cambridge and Wisbech	6817	0.48	Ely	3989	0.40
Ely and Wisbech	7001	0.43	Cambridge	3805	0.41
			Pooled		0.42

¹Total n = 10,806. MDS, Mediterranean diet score.

²Scoring weights were derived from a derivation set, then applied to independent samples (a test set) within the Fenland Study. Then, the mass Spearman's ρ between the metabolite score and the MDS was calculated in the test set.

³ ρ represents Spearman's ρ of the metabolite scores with the MDS based on the Mediterranean dietary pyramid within the derivation set (subject to an overfitting problem).

⁴Spearman's ρ between the metabolite score and the MDS was calculated by pooling 2 or 3 independent test sets.

⁵Three study sites of the Fenland Study (Cambridge, Ely, and Wisbech in Cambridgeshire, United Kingdom).