TABLE 4.

GRADE assessment for each study group and subgroup.

N of studies (n of participants)	Factors that may decrease certainty of evidence					Certainty	Pooled AUC (95% CI)
	Risk of bias	Indirectness	Inconsistency	Imprecision	Publication bias
Nonpregnant (inflammation-adjusted ferritin <15 μg/L)
9 (3,377)	Not serious1	Not serious	Not serious2	Not serious	None	⊕⊕⊕⊕⊕ High1,2	0.77 (0.74, 0.79)
Pregnant (unadjusted ferritin <15 μg/L)
2 (120)	Not serious1	Serious3	Not serious	Very serious4	None	⊕◯◯◯ Very low1,3,4	0.58 (0.45, 0.70)
Pregnant (unadjusted ferritin <30 μg/L)
2 (120)	Not serious1	Serious3	Not serious	Serious5	None	⊕◯◯◯ Low1,3,5	0.57 (0.47, 0.68)
Subgroup: blood source = capillary blood specimen
4 (1998)	Not serious1	Not serious	Serious6	Not serious	None	⊕⊕⊕◯ Moderate1,6	0.66 (0.63, 0.70)
Subgroup: blood source = venous blood specimen
8 (2932)	Not serious1	Not serious	Not serious2	Serious7	None	⊕⊕⊕◯ Moderate1,2,7	0.77 (0.74, 0.80)
Subgroup: analytical method = hemoglobinometer
7 (2722)	Not serious1	Not serious	Serious6	Serious8	None	⊕⊕◯◯ Low1,6,8	0.69 (0.66, 0.72)
Subgroup: analytical method = analyzer
5 (2208)	Not serious1	Not serious	Serious6	Serious7	None	⊕⊕◯◯ Low1,6,7	0.77 (0.74, 0.80)

Abbreviations: AUC, area under the curve; CI, confidence interval; GRADE, Grading of Recommendations Assessment, Development, and Evaluation.

Explanations for each GRADE assessment are as follows:

¹Not serious: most information was from studies at low risk of bias for each of the 4 domains in the QUADAS-2 tool [17].

²Not serious: there was inconsistency driven by 1 study (Whitfield et al. 2017 [26]), but sensitivity analysis with the removal of that study did not change the effect estimate and 95% CIs, and statistical heterogeneity (I²) in the sensitivity analysis was 0%.

³Serious indirectness: there was a low prevalence of iron deficiency within and across the studies.

⁴Very serious imprecision: the 95% CI crossed multiple thresholds of AUC, from no discrimination to “fair” as per study definitions [16].

⁵Serious imprecision: the 95% CI crossed a threshold of AUC, from no discrimination to “poor” as per study definitions [16].

⁶Serious inconsistency: point estimates and 95% CIs from individual studies differed in terms of their AUC values.

⁷Serious imprecision: the 95% CI crossed a threshold of AUC, from “fair” to “good” as per study definitions [16].

⁸Serious imprecision: the 95% CI crossed a threshold of AUC, from “poor” to “fair” as per study definitions [16].