Abstract
BACKGROUND: Pregnancy is associated with changes in hematological and biochemistry values, yet there are no African reference intervals for clinical management of pregnant women. We sought to 1) develop laboratory reference intervals during pregnancy and up to 24 weeks postpartum and 2) determine the proportion of women in a previous clinical trial who would be misclassified as having out-of-range values using reference intervals from a United States (U.S.) population. METHODS AND FINDINGS: This was a longitudinal sub-study of 120 clinically healthy, HIV-uninfected, self-selected pregnant women seeking antenatal care services at either of two public hospitals in western Kenya. Blood specimens were obtained from consented women at gestational ages 28 and 36 weeks and at 2, 6, 14 and 24 weeks postpartum. Median and 95% reference intervals were calculated for immune-hematological and biochemistry parameters and compared to reference intervals from a Kenyan and United States (U.S.) population, using Wilcoxon tests. Differences with p≤0.05 were considered significant. Some hematological parameters, including hemoglobin and neutrophils showed significant variations compared to reference intervals for non-pregnant women. Hemoglobin values were significantly lower during pregnancy but were comparable to the values in non-pregnant women by 6 weeks postpartum. CD4, CD8 and platelets were significantly elevated in early postpartum but declined gradually, reaching normal levels by 24 weeks postpartum. Using the new hemoglobin reference levels from this study to estimate prevalence of 'out of range' values in a prior Kisumu research cohort of pregnant/postpartum women, resulted in 0% out of range values, in contrast to 96.3% using US non-pregnant reference values. CONCLUSION: There were substantial differences in U.S. and Kenyan values for immune-hematological parameters among pregnant/postpartum women, specifically in red blood cell parameters in late pregnancy and 2 weeks postpartum. Use of U.S. reference intervals markedly increases likelihood of out of range values, highlighting the need for suitable locally developed reference intervals.