Diagnostic test accuracy of serum ferritin and prevalence of iron deficiency in pregnant and menstruating individuals: A systematic review and meta analysis

Abstract Background Serum ferritin is widely used to diagnose iron deficiency (ID) with or without anemia due to its correlation with iron stores. However, its diagnostic sensitivity may be limited in pregnant and menstruating women, as ferritin is an acute-phase reactant and could mask true prevalence of ID. This study systematically reviews the prevalence of ID with and without anemia in pregnant and menstruating individuals to inform baseline risk and assess the diagnostic test accuracy (DTA) of serum ferritin against the reference standard of bone marrow iron. Methods We systematically reviewed DTA results in patients who underwent BM biopsy suspected of having ID, presenting to inpatient or outpatient settings. For prevalence studies, systematically reviewed prevalence results in patients in pregnant and menstruating individuals above 12, presenting to inpatient or outpatient settings. We searched Pubmed, Cochrane Central, and EMBASE from inception till January 2025 for eligible studies. We included studies that reported data on DTA (cohort studies, cross sectional studies) for serum ferritin and prevalence of ID with or without anemia in pregnant or menstruating individuals over the age of 12. Two reviewers screened title and abstract, and full text to assess for eligibility, and abstracted data. We assessed risk of bias for DTA using QUADAS-2 and using a risk of bias tool assessing sampling, condition definition, and statistical analysis domains. Certainty of evidence was assessed using the GRADE approach. We performed meta-analysis using Stata 19.6 software, and pooled estimates were reported as sensitivity and specificity with 95% confidence interval. Prevalence was analyzed per trimester, World Bank Country Income, inflammation adjustment, and each at the different serum ferritin (SF) threshold cut-offs of 15, 30, and 50 ng/mL. Results Out of 26710 references, 128 studies were included for the final analysis. 8 studies (n=2385) addressed DTA for SF 15, with a pooled estimated of sensitivity and specificity of 0.41 (0.28, 0.56) and 0.99(0.97, 1.00) respectively. 9 studies (n=4046) addressed DTA for SF 30, with a pooled estimated of sensitivity and specificity of 0.67 (0.48, 0.81) and 0.96 (0.93, 0.98) respectively. 9 studies (n=1152) addressed DTA for SF 50 with a pooled estimate of sensitivity and specificity of 0.89 (0.75, 0.95) and 0.86 (0.81, 0.90) respectively. The certainty of evidence ranged from low to moderate due to the high risk of bias related to the reference standard, which in most cases was based on diagnostic criteria established by international societies. Regarding prevalence in pregnant individuals, a total of 31 studies (n=93782) showed a pooled prevalence of 38% (27-50). Analysis per SF cut-off was done. SF <15: 24 studies (n=83063) showed ID prevalence was 23% (17-30). Subgroup analyses revealed statistically significant differences in the prevalence between trimesters (10% in first; 24% in second; up to 40% in third). SF <30: 11 studies (n=41536) showed ID prevalence was 43% (27-61). Similarly, subgroup analyses revealed statistically significant differences in the prevalence between trimesters (27% in first; 56% in second; up to 69% in third). For SF <50, one study reported on prevalence across trimesters showing ID prevalence of 69.8%. No significant differences were detected when analyzed by country income or inflammation status. Regarding prevalence in menstruating individuals, a total of 79 studies (n=288465) showing a pooled prevalence of 27%(27-27), and 31%(31-32) when adjusted for inflammation by CRP. SF <15: 71 studies (n= 270907) showed ID prevalence was 22%(22-22) overall and 27%(27-28) in 38 studies (n=110721) that adjusted for inflammation. SF<30: 14 studies (n=96455) showed ID prevalence was 28% (28-29) overall and 33% (33-34) in 8 studies (n=39738) that adjusted for inflammation. 2 studies (n=8447) reported on prevalence at ID at SF<50 at 58% (57-59). No significant differences were detected when analyzed by country income. Conclusion This systematic review and meta-analysis is the first study to inform a clinical guideline and provide accuracy estimates for SF for the diagnosis of ID with and without anemia, and the prevalence of said conditions in pregnant and menstruating individuals. The results show that a higher cut-off of SF has promising diagnostic utility and could potentially offer diagnostic and real-life benefits for these individuals.