Introduction
Iron deficiency is the most common nutritional deficiency worldwide, and is estimated to affect a quarter of the pregnant population.1 Low systemic iron levels have been associated with several maternal pregnancy complications, including infection and preeclampsia.2 However, these results from traditional observational studies may be affected by confounding. A systematic review of trials found that daily iron supplementation in pregnancy reduced iron deficiency anemia, but there was no clear effect on other maternal complications, which in large part may be due to low statistical power.3
To robustly evaluate whether systemic iron status affects risk of maternal pregnancy complications, we conducted a Mendelian randomization (MR) study. By leveraging genetic instruments for systemic iron status, the approach greatly reduced the risk of confounding and allowed for a large study sample.
Methods
In this two-sample MR study, we evaluated the association between genetically-predicted iron homeostasis biomarkers and the following five maternal pregnancy complications: abruptio placenta, genitourinary tract infection, postpartum hemorrhage, premature rupture of membranes, and preeclampsia.
The genetic instruments for the iron homeostasis biomarkers were collected from a recent genome-wide association study of 84,328 premenopausal women (Supplementary Methods).4 Our primary exposure of interest was systemic iron status; measured through a composite of multiple iron homeostasis biomarkers.5 As our instruments for systemic iron status we therefore used the four uncorrelated single-nucleotide polymorphisms that were robustly associated with each of serum iron, iron binding capacity, transferrin and ferritin, as has previously been done.5 In sensitivity analyses, we conducted separate analyses for each of the four iron homeostasis biomarkers, using all uncorrelated genetic variants robustly associated with the respective biomarker.
The genetic associations with the outcomes of interest were collected from relevant genome-wide association studies (Supplementary Methods).
The main analysis was the inverse-variance weighted method, which assumes all genetic instruments to be valid.6 MR estimates may be biased by pleiotropy (the genetic instruments affect the outcome other than through the exposure) and weak instruments: Weighted mode, weighted median, and MR Egger sensitivity analyses were conducted to address the former,6 while weak instruments were accounted for using robust adjusted profile scores.6 To account for multiple testing, we used p-value < 0.01 as threshold for statistical significance. All MR analyses were conducted using the TwoSampleMR package (version 0.5.6) in R (version 3.6.2).
Publicly available data with relevant ethical approvals were used.
Results
Genetically-predicted systemic iron status was significantly associated with risk of genitourinary tract infection in pregnancy; odds ratio 0.65 (95% confidence interval 0.47 to 0.89, p-value = 0.0085) for each standard deviation increase in serum iron (Figure). There was a tendency of a protective effect of increasing genetically-predicted systemic iron status on abruptio placenta, but low precision yielded inconclusive results. Genetically-predicted systemic iron status was not clearly linked to the three other complications.
MR sensitivity analyses supported the findings from the main analyses (Figure), as did the sensitivity analyses of transferrin saturation and serum iron (Table).
Finally, we conducted a post hoc analysis of genetically-predicted systemic iron status and risk of urinary tract infection among non-pregnant women (Supplementary Methods) which yielded no strong association (inverse-variance weighted analysis; odds ratio 1.00, 95% confidence interval 0.69 to 1.46).
Discussion
We found evidence supporting a protective effect of increasing systemic iron status on risk of genitourinary tract infection in pregnancy. Considering this finding together with previous observations,2 it is plausible that the reported association is causal. Given the increased demands of iron in pregnancy, and that a quarter of the pregnant population are iron deficient,1 our findings support efforts in public health and antenatal care to ensure adequate iron status among pregnant women.
Figure. Systemic iron status and risk of maternal pregnancy complications.
Unit of exposure is one standard deviation (7.1 μmol/L) increase of genetically-predicted serum iron using the four uncorrelated genetic instruments from the three genes (DUOX2, HFE, and TMPRSS6) linked to all iron homeostasis biomarkers.4 The number of cases are reported for each complication (number of controls presented in Supplementary Methods). CI, confidence interval; OR, odds ratio; SNPs, single-nucleotide polymorphisms.
Table. Sensitivity analysis of iron homeostasis biomarkers and risk of maternal pregnancy complications.
| Abruptio placenta | Genitourinary tract infection | Postpartum hemorrhage | Premature rupture of membranes | Preeclampsia | ||||||
|---|---|---|---|---|---|---|---|---|---|---|
| OR (95% CI) | P-value | OR (95% CI) | P-value | OR (95% CI) | P-value | OR (95% CI) | P-value | OR (95% CI) | P-value | |
| Transferrin saturation | ||||||||||
| Inverse variance weighted | 0.53 (0.21-1.28) | 0.161 | 0.60 (0.42-0.84) | 0.003 | 0.94 (0.78-1.13) | 0.558 | 0.93 (0.70-1.24) | 0.644 | 0.84 (0.69-1.02) | 0.093 |
| Weighted mode | 0.36 (0.12-1.06) | 0.101 | 0.63 (0.42-0.95) | 0.029 | 0.99 (0.78-1.25) | 0.942 | 0.90 (0.65-1.23) | 0.541 | 0.88 (0.71-1.09) | 0.297 |
| Weighted median | 0.37 (0.12-1.10) | 0.076 | 0.67 (0.43-1.02) | 0.067 | 0.97 (0.77-1.21) | 0.810 | 0.85 (0.59-1.22) | 0.396 | 0.89 (0.70-1.13) | 0.353 |
| MR Egger | 0.58 (0.16-2.05) | 0.427 | 0.65 (0.40-1.06) | 0.126 | 0.98 (0.75-1.28) | 0.915 | 0.93 (0.62-1.40) | 0.752 | 0.85 (0.64-1.12) | 0.293 |
| Robust adjusted profile score | 0.53 (0.21-1.34) | 0.183 | 0.59 (0.41-0.85) | 0.005 | 0.94 (0.77-1.14) | 0.560 | 0.93 (0.69-1.25) | 0.632 | 0.85 (0.69-1.04) | 0.127 |
| Iron | ||||||||||
| Inverse variance weighted | 0.55 (0.25-1.17) | 0.124 | 0.62 (0.45-0.84) | 0.002 | 0.94 (0.79-1.10) | 0.474 | 0.95 (0.74-1.21) | 0.712 | 0.90 (0.76-1.07) | 0.265 |
| Weighted mode | 0.46 (0.18-1.17) | 0.132 | 0.66 (0.46-0.93) | 0.015 | 0.92 (0.76-1.12) | 0.443 | 0.94 (0.71-1.23) | 0.670 | 0.94 (0.77-1.15) | 0.604 |
| Weighted median | 0.50 (0.19-1.32) | 0.164 | 0.66 (0.45-0.97) | 0.035 | 0.89 (0.72-1.09) | 0.275 | 0.89 (0.66-1.20) | 0.471 | 0.94 (0.76-1.17) | 0.611 |
| MR Egger | 0.67 (0.21-2.07) | 0.506 | 0.66 (0.41-1.06) | 0.115 | 0.91 (0.71-1.16) | 0.487 | 1.02 (0.71-1.47) | 0.889 | 0.85 (0.65-1.11) | 0.281 |
| Robust adjusted profile score | 0.54 (0.25-1.19) | 0.131 | 0.61 (0.45-0.83) | 0.002 | 0.93 (0.79-1.11) | 0.452 | 0.95 (0.74-1.23) | 0.727 | 0.91 (0.76-1.09) | 0.346 |
| Iron binding capacity | ||||||||||
| Inverse variance weighted | 0.98 (0.67-1.43) | 0.922 | 1.06 (0.92-1.23) | 0.388 | 1.00 (0.92-1.09) | 0.915 | 1.05 (0.93-1.18) | 0.409 | 1.05 (0.95-1.17) | 0.293 |
| Weighted mode | 0.98 (0.64-1.48) | 0.927 | 1.05 (0.89-1.23) | 0.561 | 0.98 (0.90-1.07) | 0.753 | 1.06 (0.93-1.21) | 0.365 | 1.03 (0.94-1.14) | 0.459 |
| Weighted median | 1.00 (0.65-1.55) | 0.980 | 1.00 (0.84-1.18) | 0.994 | 0.97 (0.89-1.07) | 0.630 | 1.04 (0.91-1.20) | 0.506 | 1.04 (0.94-1.15) | 0.428 |
| MR Egger | 0.94 (0.58-1.53) | 0.825 | 1.04 (0.86-1.26) | 0.639 | 0.98 (0.88-1.10) | 0.824 | 1.02 (0.87-1.19) | 0.780 | 1.02 (0.89-1.17) | 0.749 |
| Robust adjusted profile score | 0.98 (0.66-1.44) | 0.924 | 1.06 (0.91-1.23) | 0.443 | 0.99 (0.91-1.08) | 0.988 | 1.05 (0.92-1.19) | 0.427 | 1.05 (0.94-1.16) | 0.361 |
| Ferritin | ||||||||||
| Inverse variance weighted | 1.74 (0.75-4.01) | 0.191 | 0.87 (0.61-1.24) | 0.457 | 0.92 (0.77-1.10) | 0.376 | 1.08 (0.84-1.39) | 0.517 | 0.82 (0.64-1.05) | 0.119 |
| Weighted mode | 1.35 (0.43-4.23) | 0.608 | 0.68 (0.41-1.12) | 0.148 | 0.87 (0.65-1.17) | 0.385 | 0.99 (0.67-1.45) | 0.962 | 0.85 (0.65-1.11) | 0.250 |
| Weighted median | 1.13 (0.37-3.41) | 0.820 | 0.69 (0.43-1.13) | 0.149 | 0.96 (0.72-1.28) | 0.813 | 1.09 (0.76-1.56) | 0.622 | 0.79 (0.59-1.06) | 0.121 |
| MR Egger | 0.48 (0.14-1.64) | 0.257 | 0.76 (0.43-1.34) | 0.358 | 0.84 (0.63-1.11) | 0.247 | 0.99 (0.67-1.47) | 0.990 | 0.81 (0.55-1.20) | 0.317 |
| Robust adjusted profile score | 1.70 (0.72-4.04) | 0.224 | 0.86 (0.61-1.21) | 0.408 | 0.93 (0.77-1.12) | 0.472 | 1.08 (0.83-1.40) | 0.550 | 0.83 (0.66-1.04) | 0.107 |
Estimates represent odds ratios per standard deviation increase in each exposure, which is 11.7 %, 7.1 μmol/L, 14.9 μmol/L, and 70.5 μg/L for transferrin saturation, serum iron, iron binding capacity and ferritin, respectively. The following maximum number of single-nucleotide polymorphisms were included in the analyses: Transferring saturation, n = 10; iron, n = 14; iron binding capacity, n = 16; and ferritin, n = 37. CI, confidence interval; OR, odds ratio
Footnotes
Conflict of interest
DG is employed part-time by Novo Nordisk, outside of the submitted work. The other authors report no conflict of interest.
Contributor Information
Tormod Rogne, Email: tormod.rogne@yale.edu.
Stephen Burgess, Email: sb452@medschl.cam.ac.uk.
Dipender Gill, Email: dipender.gill@imperial.ac.uk.
References
- 1.Barroso F, Allard S, Kahan BC, et al. Prevalence of maternal anaemia and its predictors: A multi-centre study. Eur J Obstet Gynecol Reprod Biol. 2011;159(1):99–105. doi: 10.1016/j.ejogrb.2011.07.041. [DOI] [PubMed] [Google Scholar]
- 2.Smith C, Teng F, Branch E, Chu S, Joseph KS. Maternal and Perinatal Morbidity and Mortality Associated with Anemia in Pregnancy. Obstet Gynecol. 2019;134(6):1234–1244. doi: 10.1097/AOG.0000000000003557. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Peña-Rosas JP, De-Regil LM, Garcia-Casal MN, Dowswell T. Daily oral iron supplementation during pregnancy. Cochrane Database Syst Rev. 2015;2015(7):1–527. doi: 10.1002/14651858.CD004736.pub5. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Bell S, Rigas AS, Magnusson MK, et al. A genome-wide meta-analysis yields 46 new loci associating with biomarkers of iron homeostasis. Commun Biol. 2021;4(1):156. doi: 10.1038/s42003-020-01575-z. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Gill D, Benyamin B, Moore LSP, et al. Associations of genetically determined iron status across the phenome: A mendelian randomization study. Rahimi K, editor. PLOS Med. 2019;16(6):e1002833. doi: 10.1371/journal.pmed.1002833. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Burgess S, Davey Smith G, Davies NM, et al. Guidelines for performing Mendelian randomization investigations. Wellcome Open Res. 2020;4:186. doi: 10.12688/wellcomeopenres.15555.2. [DOI] [PMC free article] [PubMed] [Google Scholar]