International values for haemoglobin distributions in healthy pregnant women

Background: Anaemia in pregnancy is a global health problem with associated morbidity and mortality. Methods: A secondary analysis of prospective, population-based study from 2009 to 2016 to generate maternal haemoglobin normative centiles in uncomplicated pregnancies in women receiving optimal antenatal care. Pregnant women were enrolled <14 weeks’ gestation in the Fetal Growth Longitudinal Study (FGLS) of the INTERGROWTH-21st Project which involved eight geographically diverse urban areas in Brazil, China, India, Italy, Kenya, Oman, United Kingdom and United States. At each 5 ± 1 weekly visit until delivery, information was collected about the pregnancy, as well as the results of blood tests taken as part of routine antenatal care that complemented the study's requirements, including haemoglobin values. Findings: A total of 3502 (81%) of 4321 women who delivered a live, singleton newborn with no visible congenital anomalies, contributed at least one haemoglobin value. Median haemoglobin concentrations ranged from 114.6 to 121.4 g/L, 94 to 103 g/L at the 3rd centile, and from 135 to 141 g/L at the 97th centile. The lowest values were seen between 31 and 32 weeks’ gestation, representing a mean drop of 6.8 g/L compared to 14 weeks’ gestation. The percentage variation in maternal haemoglobin within-site was 47% of the total variance compared to 13% between sites. Interpretation: We have generated International, gestational age-specific, smoothed centiles for maternal haemoglobin concentration compatible with better pregnancy outcomes, as well as adequate neonatal and early childhood morbidity, growth and development up to 2 years of age. Funding: Bill & Melinda Gates Foundation Grant number 49038. This work is reported following the STROBE guidelines [35]. A secondary analysis of prospective, population-based, longitudinal, observational cohort study from 2009 to 2016 to generate maternal haemoglobin normative centiles in uncomplicated pregnancies in women receiving optimal antenatal care. The INTERGROWTH-21st Project consisted of several interrelated studies with the principal aim of evaluating growth, health, nutrition and development from less than 14 weeks’ gestation to 2 years of age, using the same conceptual framework as the WHO Multicentre Growth Reference Study (MGRS) [36]. The INTERGROWTH-21st Project was carried out between 2009 and 2016 across eight diverse geographically delimited urban areas in: Pelotas (Brazil), Turin (Italy), Muscat (Oman), Oxford (UK), Seattle (USA), Beijing (China), Nagpur (India), and Nairobi (Kenya) [25]. The selection criteria at the cluster level were: the areas had to be located at an altitude <1600 m above sea level with a low risk of fetal and infant growth and developmental disturbances, as well as an absence or low levels of major, known, non-microbiological contamination. Within each area, all institutions classified locally as “private” or “corporation” hospitals and/or serving the middle to upper socio-economic population were selected, provided that most institutional deliveries from the target population took place there. Women receiving antenatal care had to plan to deliver in these institutions or in a similar hospital located in the same geographical area. The participants were selected based upon criteria for optimal health, nutrition, education and socioeconomic status, needed to construct international standards [24]. At each study site, we recruited women with no clinically relevant obstetric, gynaecological or medical history, who initiated antenatal care in early pregnancy i.e., <14+0 weeks’ gestation by menstrual dates, and met the entry criteria of optimal health, nutrition, education and socio-economic status. A detailed description of the entry criteria and definitions has been published previously [25]. For example, adequate nutritional status was defined in the first trimester according to maternal height (≥153 cm), body mass index (BMI, ≥18.5 and <30 kg/m2), Hb level (≥110 g/L), and not receiving treatment for anaemia or following any special diets (e.g., vegetarian with no animal products). This resulted in a group of educated, affluent, clinically healthy women with adequate nutritional status, who by definition were at low risk of adverse maternal and perinatal outcomes. The FGLS exclusion criteria included hypertension (defined as systolic ≥140 mmHg or diastolic ≥90 mmHg) in a past pregnancy or in the first trimester of the present pregnancy; chronic hypertension on treatment, and a past history of preeclampsia, eclampsia or Haemolysis Elevated Liver enzymes and Low Platelets (HELLP) syndrome. FGLS also excluded women if their pregnancies became complicated by criteria specified a priori, including fetal death, congenital anomaly, severe or catastrophic medical morbidity not evident at enrolment (such as cancer or HIV), severe unanticipated conditions related to the pregnancy (such as severe preeclampsia or eclampsia), and those identified during the study who no longer fulfilled the entry criteria (e.g. women who started smoking during pregnancy or had an episode of malaria) [25]. Gestational age was calculated from the date of the last menstrual period provided: the woman had a regular 24–32-day menstrual cycle, she had not been using hormonal contraception or breastfeeding in the preceding 2 months, and any discrepancy between the gestational ages based on last menstrual period and crown-rump length, measured by ultrasound between 9+0 and 13+6 weeks’ gestation, was 7 days or less. The dating scan was undertaken using standard study criteria for measuring crown-rump length [37]. Dedicated research staff then performed an ultrasound scan every 5 weeks (± 1 week) until delivery to assess fetal growth. At each visit, information was collected about the pregnancy, as well as the results of blood tests (including Hb) taken as part of routine antenatal care that was provided separately to the study's requirements. The gestational age at which those tests were taken varied depending on local protocols as this was a pragmatic study that aimed to mimic routine clinical practice in the different settings. The primary objective of this analysis of the FGLS data was two-fold: (a) to describe Hb ranges and trajectories in a population of optimally healthy women with good pregnancy, perinatal and neonatal outcomes, whose children had satisfactory postnatal growth and development up to 2 years of age so as to establish gestational age-specific distributions and populations thresholds for normal Hb in pregnancy, and (b) to define prevalence thresholds to diagnose adequate Hb concentrations in pregnancy at the individual level and the prevalence of anaemia at population level. The Hb tests were taken as part of routine antenatal care, i.e., in relation to laboratory tests, 1) we relied on collecting the results of available routine blood tests; 2) the commercially available instruments for assessing Hb were not standardised across sites; and 3) information on the use of preventive or therapeutic iron and folic acid-containing supplements or calcium supplements, was collected from medical records. Although the eight study sites were not asked to follow a specific protocol, we have documented carefully the biochemical methods of Hb determination they used. All sites assessed Hb concentration from venous blood samples using commercially available methods (automatised colorimetry, automatised turbidimetry, high efficiency liquid chromatography, sysmex autoanalyser, automated flow fluorescent analyser, photometric method using automated cell counter, high-efficiency liquid chromatography and cyanide-free sodium lauryl sulphate) of Hb assessment that are widely used in routine patient care and considered highly reliable [38]. Our overall aim was to produce Hb centiles that change smoothly with gestational age and maximise simplicity without compromising model fit. We followed the same statistical methodology and approach previously described [39,40] for the analyses of already published international standards [29,30,33,41]. The first step was to assess the variation in maternal Hb across different study sites to determine whether we could pool the data to estimate international normative values. The criteria used to judge similarities among study sites was based on WHO recommendations for analysing human growth data [42]. In brief, we first inspected the data visually comparing patterns across sites. We then applied variance component analysis (analysis of variance (ANOVA)) to calculate the percentage of variance in the longitudinal maternal Hb concentrations from variance between sites and the estimated variance in individuals within a site (within-site variance). We treated gestational age as a fixed effect, whereas sites and individuals were treated as random effects in a multi-level linear regression model. Having satisfied the criteria for pooling, we used the pooled data to construct smoothed centiles of maternal Hb according to gestational age using fractional polynomial regression that models the mean and standard deviation (SD) separately as a smooth function of gestational age. The best fitting powers for the mean and SD of maternal Hb according to gestational age were provided by the second and first-degree fractional polynomials, respectively. Goodness of fit of the resultant models was assessed as previously described for the INTERGROWTH-21st data by Ohuma and Altman [39], i.e., visual inspection of the overall model fit by comparing empirical centiles (calculated per completed week of gestation, e.g. 38 weeks = 38+0 - 38+6 weeks’ gestation) to the fitted centiles, a plot of the residuals (observed values minus fitted values) according to gestational age, a quantile-quantile (Q‐Q) plots of the residuals to assess normality, and a plot of fitted z-scores across gestational ages. We then conducted various sensitivity analyses: 1) comparing the fitted smoothed centiles of longitudinal Hb data (n = 3502 women, 9954 observations) to a cross-sectional random sample of Hb data between 14+0 and 40+0 weeks’ gestation for each woman included in the study (n = 3502 observations) to evaluate whether multiple Hb values per women resulted in reduced error variance and consequently reduced variance of the estimated centiles; 2) comparing the total pooled sample (n = 3502 women) with the sample (n = 3364 women) that excluded those mothers who delivered preterm, i.e. less than 37 weeks’ gestation (n = 138 women), and then superimposing the two sets of fitted centiles to evaluate whether there were differences in maternal Hb among women delivering preterm compared to term newborns; and 3) excluding each site's Hb data one at a time, refitting the centiles (seven sites’ data), and comparing the fitted (i.e., 3rd, 50th and 97th centiles) on the basis of fractional polynomial regression between the pooled data (eight sites) and the reduced datasets (one site excluded at a time) to establish whether there was any site-specific influence to the derived smoothed pooled maternal Hb centiles. Descriptive analyses were used to summarise data on supplementation information that was available and collected as part of routine care. The supplementation was provided as prophylaxis as applied in routine practice following country-specific guidelines. In addition, for each site, we calculated empirical Hb centiles (specifically, 3rd, 5th, 10th, 50th, 90th, 95th and 97th centiles) and then computed the median across all eight sites to obtain Hb centiles for situations where gestational age is unknown. All analyses were performed in STATA, version 15, software (StataCorp LP, College Station, TX, USA). The next step was to decide the approach for establishing thresholds for Hb concentration to define anaemia at individual level. Such thresholds are applied to judge the location of a single value in relation to the median of the normative distribution, i.e. to assess an individual's status. The definition of “normality” is conventionally set at 2 SD below (or above) a standard or normative median, but this is frequently rounded up to the 3rd (or 97th) centile in most international guidelines and literature [43]. We used this value to define the most severe threshold of Hb concentration in this population of healthy pregnant women. The INTERGROWTH-21st Steering Committee included voluntary lay member representation during the design and implementation of the project [44]. We plan to involve pregnant women in the dissemination of results through publication in peer-reviewed journals, presentation at national conferences and involvement of maternity groups associated with the Nuffield Department of Women's & Reproductive Health, University of Oxford, and the WHO constituted Guideline Development Group – Anaemia: use and interpretation of Hb concentrations for assessing anaemia status in individuals and populations. The INTERGROWTH-21st Project was approved by the Oxfordshire Research Ethics Committee “C” (reference: 08/H0606/139), the research ethics committees of the individual institutions and the regional health authorities where the project was implemented. All women provided informed written consent to participate in the study. The funders of the study had no role in study design, data collection, data analysis, data interpretation, or writing of the report. The corresponding author had full access to all the data in the study and had final responsibility for the decision to submit for publication.