TABLE 3.

PCAs of iron homeostasis and inflammatory variables in men and women at HIV diagnosis and Cox regression analyses of all-cause mortality¹

					HR (95% CI)
							Iron homeostasis index5
PCA model	Grouping pattern identified by PCA	Total variance explained by pattern, %	PCs contributing to grouping pattern	PC eigenvector2	PC score,3 unadjusted	PC score,3 adjusted4	Unadjusted	Adjusted4
PCA16	Pattern 1 (PCA1.1)	43.4	Hepcidin	0.54	1.72 (1.49, 1.99)	1.31 (1.07, 1.61)	1.37 (1.26, 1.50)	1.13 (1.00, 1.27)
			Ferritin	0.54
			Transferrin	0.53
	Pattern 2 (PCA1.2)	24.7	Iron	0.63	1.08 (0.93, 1.27)	1.00 (0.83, 1.20)	1.15 (1.03, 1.29)	1.08 (0.80, 1.45)
			sTfR	0.54
			Hemoglobin	0.49
PCA27	Pattern 1 (PCA2.1)	44.8	Hepcidin	0.48	1.64 (1.44, 1.87)	1.26 (1.05, 1.51)	1.28 (1.19, 1.37)	1.08 (0.99, 1.89)
			Ferritin	0.48
			Transferrin	0.46
			ACT	0.45
	Pattern 2 (PCA2.2)	21.3	Iron	0.61	1.06 (0.90, 1.24)	0.98 (0.82, 1.18)	1.158 (1.03, 1.29)	1.088 (0.80, 1.45)
			sTfR	0.56
			Hemoglobin	0.46

¹Unadjusted models sample size, n = 196; unadjusted with hemoglobin, n = 140; adjusted, n = 168, adjusted with hemoglobin, n = 119, due to missing data. ACT, α₁-antichymotrypsin; PC, principal component; PCA, principal components analysis; sTfR, soluble transferrin receptor.

²Ordered by relative loading (impact) of eigenvector where more important variables are assigned greater weights. Only meaningful (e.g., eigenvector >0.40) variables are presented.

³PC scores represent a linear algebraic combination of all variables in the model and are weighted by eigenvectors.

⁴Adjusted for HIV type (HIV-1/HIV-dual, HIV-2), age, gender, BMI, absolute CD4 cell count (>500, 200–500, or <200 cells/μL).

⁵The iron homeostasis index represents the combined impact on mortality of PCA-identified patterns. Observed concentrations were classified into tertiles: a value of 1 was assigned to tertiles with the lowest HR (in Table 2) including the lowest hepcidin, ferritin, ACT, and sTfR tertiles and the highest transferrin, iron, and hemoglobin tertiles. Intermediate tertile = 2; otherwise = 3 (e.g., the greatest mortality HR in Table 2). Values were summed for individuals representing an overall iron homeostasis index, which was modeled as a single continuous variable.

⁶Six-dimensional model including all iron homeostasis variables entered in tertiles.

⁷Seven-dimensional model including all iron homeostasis variables plus ACT entered in tertiles.

⁸The iron homeostasis index for PCA1.2 and PCA2.2 is equivalent because the same PC pattern (e.g., iron-sTfR-hemoglobin) was identified in both PCA models.