We excluded seventeen women (17/115 or 14.8%) who had given their consent for participation in the study, from the analysis, because they refused part of the sampling procedure, or because of incomplete or inadequate sampling, including delay between vaginal and blood sampling, or because of early (missed) abortion. As a group, the subjects excluded did not differ significantly from the remainder of women in terms of baseline characteristics (mean maternal age, mean body mass index, smoking habits, mean gestational age, and median parity and gravidity).
From each patient included for the final analysis (98/115 or 85.2%) a vaginal swab and venous blood samples were obtained at a single point in time at a mean gestational age of 9.2 ± 2.6 weeks. Eighty women had normal or grade I microflora on Gram stain and eighteen women presented with disturbed or vaginosis-like microflora, i.e. intermediate (grade II) microflora or overt bacterial vaginosis (grade III) microflora.
Basic clinical characteristics of study participants, who were all of white Caucasian origin, are displayed in table . Women with disturbed vaginal flora in the index pregnancy were significantly more likely to have delivered a child previously and tended to have a higher body mass index (Table ).
Basic clinical characteristics of the study population.
Traditional indicators of iron deficiency in relation to vaginal microflora status
In this study sample, 10.2% of subjects (10/98) had depleted iron stores during early pregnancy according to the conventional criterion of serum ferritin ≥ 12 g/L. 4.1% of women (4/98) presented with anaemia defined as Hgb < 11 g/dL and 10.2% with a Hgb concentration below 12 g/dL (10/98), though only one of these (1/98 or 1.0%) had true iron-deficient anaemia when accounting for both serum ferritin (≥ 12 g/L) and haemoglobin (< 11 g/dL). Any other traditional compound indices of iron-deficiency, e.g. by accounting for mean cell volume, mean corpuscular haemoglobin, and transferrin saturation did not identify any additional cases of clinical overt iron deficiency.
There was no statistically significant correlation between vaginal microflora status as assessed by Gram stain and any of the conventionally assessed iron and red blood cell indices, including red blood cell counts, serum haemoglobin, haematocrit, mean corpuscular volume, mean corpuscular haemoglobin, serum iron, serum ferritin, serum transferrin and transferritin saturation (Table ). Similarly, we observed no significant association between vaginal microflora status and maternal CRP concentration during early pregnancy. In this cohort there were no cases of overt systemic inflammation according to maternal CRP, nor were ferritin and transferrin concentrations significantly correlated with CRP.
Table 2 Maternal serum iron and red blood cell indices according to vaginal microflora status. Healthy vaginal flora is defined as grade I or lactobacilli-dominated microflora on Gram stain (corresponding to a Nugent score 0 – 3) and disturbed vaginal (more ...)
Soluble transferrin receptors in relation to vaginal microflora status
In contrast to the above, we observed a trend by which maternal serum transferrin receptor (sTfR) concentrations during early pregnancy were negatively correlated with lactobacillary grading and hence positively correlated with the degree of vaginal microflora alteration (R = 0.26, p = 0.01). When vaginal microflora status was handled as a dichotomous variable ('healthy' versus 'disturbed'), women with healthy vaginal microflora (n = 80) during early pregnancy had a mean sTfR concentration of 1.15 ± 0.30 mg/L as compared to a mean sTfR of 1.37 ± 0.38 mg/L among women (n = 18) with disturbed microflora (p = 0.008) (Figure ).
Figure 1 Distribution of sTfR concentrations according to vaginal microflora status. Box-and-whisker plots of the sTfR distributions according to vaginal microflora status during early pregnancy. The thick line represents the median sTfR value, the horizontal (more ...)
Given the significant overlap in sTfR distributions for the healthy and disturbed vaginal microflora groups of women, classification plots were constructed and the sTfR value with the highest discriminative value between both groups was chosen as the sTfR cut-off level. Serum transferrin receptor concentrations > 1.45 mg/L were associated with a prevalence or risk ratio of 3.0 (95% CI: 1.4 – 6.7, p = 0.014) for disturbed vaginal flora. The accuracy of the sTfR assay at this threshold was 79% (95% CI: 0.69–0.86). Sensitivity and the positive predictive value (PPV) were however low and estimated at 39% (95% CI: 0.18–0.64) and 41% (95% CI: 0.19–0.67), respectively. In contrast, specificity and consequently the negative predictive value (NPV) of the assay were as high as 88% (95% CI: 0.78–0.94) and 86% (95% CI: 0.77–0.93), respectively.
Of the parameters entered into in the multivariable analysis only smoking, body mass index and gravidity/parity were expected to act as true confounders of the association under study considering these variables impinge both on serum transferrin receptor concentrations and on vaginal microflora status. We also controlled for maternal age, gestational age at sampling, and CRP as covariates in the model. Yet, the only significant variable retained from the multivariable analysis was the maternal sTfR concentration, suggesting that raised sTfR concentrations during early pregnancy are independently associated with vaginal microflora alteration. The adjusted odds ratio of an sTfR-concentration > 1.45 mg/L for vaginosis-like vaginal microflora was 4.5 (95%CI: 1.4–14.2, p = 0.011).
The log10 [sTfR/ferritin] iron deficiency index in relation to vaginal microflora status
Since sTfR concentrations may reflect both rates of erythropoiesis as well as cellular iron needs, several combined measures following log transformation of sTfR and/or ferritin have been used as highly specific and sensitive measures of iron deficiency in particular, including sTfR/log10
[ferritin], and log10
The logarithm of the ratio of the soluble transferrin receptor to ferritin concentration (log10
[sTfR/ferritin]) is currently the most precise measure of body storage iron available [32
]. We found a highly significant difference (difference in means = 0.49, 95% CI 0.30–0.68, p < 0.0001) between mean values of the log10
[sTfR/ferritin] index (1.57 ± 0.30 versus 1.08 ± 0.56) between both groups (Table ) further indicating that the observed association between maternal sTfR concentrations and vaginal microflora status indeed relates to depleted available body iron stores and cellular iron avidity.
Though our sample size was rather small, post hoc analysis revealed that the above difference in the mean iron deficiency index values (log10[sTfR/ferritin]) between both groups was demonstrated at a two-sided significance level of α = 0.05 with a statistical power of more than 90% (1-β = 92.7) under the parametric assumption and by accounting for unequal variances.
Supplemental iron intake as a potential determinant of iron indices
Of note is that 65.3% of women in this cohort (64/98) were already taking oligo-elements or vitamin supplements at the time of sampling, most often as combined preparations (35/98 or 35.7%) or as folate supplements (27/98 or 27.6%). We found no significant association between iron supplementation (37/98 or 37.8%) and maternal sTfR concentrations (p = 0.95) or with the combined sTfR-ferritin indexes following log transformation (p = 0.20 to 0.97), while there was a marginally significant association between serum ferritin concentrations and supplemental iron intake (p = 0.053).