SEARCH

SEARCH BY CITATION

Keywords:

  • post-partum haemorrhage;
  • anaemia;
  • parturition;
  • pregnancy
  • hémorragie post-partum;
  • anémie;
  • accouchement;
  • grossesse
  • hemorragia posparto;
  • anemia;
  • parturienta;
  • embarazo

Abstract

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Objectives  Anaemia is a potential long-term sequel of obstetric blood loss, but the increased risk of anaemia in women who experience a haemorrhage compared to those who do not has not been quantified. We sought to quantify this risk and explore the duration of increased risk for these women.

Methods  Systematic review of articles published between 1990 and 2009. Data were analysed by high- and low-income country groupings. Prevalence and incidence ratios, and mean haemoglobin levels were compared.

Results  Eleven of 822 studies screened were included in the analysis. Most studies showed a higher prevalence or incidence of anaemia in women who had experienced haemorrhage than in those who did not, irrespective of the timing of measurement post-partum. In high-income countries, women who had a haemorrhage were at 5.68 (95% CI 5.04–6.40) times higher risk of post-partum anaemia than women who did not. In low-income countries, the prevalence of anaemia was 1.58 (95% CI 0.96–2.60) times higher in women who had a haemorrhage than in women who did not, although this ratio was greater when the study including mild anaemia in its definition of anaemia was excluded (1.93, 95% CI 1.42–2.62). Population-attributable fractions ranged from 14.9% to 39.6%. Several methodological issues, such as definitions, exclusion criteria and timing of measurements, hindered the comparability of study results.

Conclusions  Women who experience haemorrhage appear to be at increased risk of anaemia for many months after delivery. This important finding could have serious implications for their health care and management.

Objectifs:  L’anémie est une séquelle potentielle à long terme à la suite de perte obstétrique de sang, mais l’augmentation du risque d’anémie chez les femmes qui font une hémorragie par rapport à celles qui ne la font pas n’a pas été quantifiée. Nous avons cherchéà quantifier ce risque et à investiguer la durée du risque accru chez ces femmes.

Méthodes:  Revue systématique des articles publiés entre 2000 et 2009. Les données ont été analysées par pays à revenus élevés et pays à faibles revenus. Les taux de prévalence et d’incidence, et les taux moyens d’hémoglobine ont été comparés.

Résultats:  Onze sur 822 études de dépistage ont été incluses dans l’analyse. La plupart des études ont montré une prévalence ou une incidence plus élevée de l’anémie chez les femmes qui avaient subi une hémorragie par rapport à celles qui ne l’avaient pas subi, quel que soit le moment de la mesure au cours du post-partum. Dans les pays à revenus élevés les femmes qui avaient fait une hémorragie étaient 5,68 (IC95%: 5,04 à 6,40) fois plus à risque d’anémie post-partum que celles qui ne l’avaient pas fait. Dans les pays à faibles revenus la prévalence de l’anémie était 1,58 (IC95%: 0,96 à 2.60) fois plus élevée chez les femmes qui avaient fait une hémorragie que chez celles qui l’avaient pas fait, bien que ce rapport était plus élevé lorsqu’une étude, prenant en compte l’anémie légère dans sa définition de l’anémie a été exclue (1,93; IC95%: 1,42 à 2,62). Les fractions attribuables dans la population variaient de 14,9%à 39,6%. Plusieurs questions méthodologiques, telles que les définitions, les critères d’exclusion et le moment des mesures, entravaient la comparabilité des résultats de l’étude.

Conclusions:  Les femmes qui subissent une hémorragie semblent être à risque accru d’anémie durant plusieurs mois après l’accouchement. Cette importante observation révélerait des implications sérieuses pour leur santé et sa prise en charge.

Objetivos:  La anemia es una posible secuela a largo plazo de la pérdida de sangre por motivos obstétricos, pero hasta ahora no se ha cuantificado el riesgo aumentado de anemia de las mujeres que han experimentado una hemorragia obstétrica en comparación con aquellas que no la han tenido. Hemos buscado cuantificar el riesgo y explorar la duración del riesgo aumentado para estas mujeres.

Métodos:  Revisión sistemática de artículos publicados entre el 2000 y el 2009. Los datos se analizaron en grupos de países con ingresos altos y con ingresos bajos. Se compararon las tasas de prevalencia e incidencia, y los niveles medios de hemoglobina.

Resultados:  Se incluyeron en el análisis once de los 822 estudios revisados. La mayoría de los estudios mostraban una mayor prevalencia o incidencia de anemia en mujeres que habían experimentado una hemorragia, en comparación con aquellas que no la habían tenido, independientemente del momento de realizar la medición postparto. En países de renta alta, las mujeres que habían tenido una hemorragia tenían 5.68 (IC 95% 5.04-6.40) veces más riesgo de anemia en el postparto que las mujeres que no la habían tenido. En países de baja renta la prevalencia de anemia era 1.58 (IC 95% 0.96-2.60) veces mayor en mujeres que habían sufrido una hemorragia que en mujeres que no la habían tenido, aunque esta proporción era mayor cuando se excluían los estudios que incluían la anemia leve en su definición de anemia (1.93, IC 95% 1.42-2.62). Las fracciones atribuibles poblacionales estaban entre el 14.9% y el 39.6%. Varias cuestiones metodológicas, tales como las definiciones, los criterios de exclusión y el momento de las mediciones interferían en la comparabilidad de los resultados de los estudios.

Conclusiones:  Las mujeres que experimentaron una hemorragia parecían tener un mayor riesgo de anemia durante muchos meses después del parto. Este importante hallazgo podría tener implicaciones serias en los cuidados sanitarios y el manejo de este tipo de pacientes.


Introduction

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Obstetric haemorrhage is a leading cause of maternal mortality worldwide (Khan et al. 2006) and also has serious health consequences for women who survive. Significant bleeding reduces haemoglobin levels and hence leads to anaemia. Women experiencing a haemorrhage from an abortion, during delivery or post-partum are therefore at increased risk of post-abortum or post-partum anaemia. Globally, it is estimated that post-partum haemorrhage of blood loss 1000 ml or more has a case fatality rate of 1%; a further 12% survive but with severe anaemia (AbouZahr 2003).

Although the proportion of women experiencing anaemia after post-partum haemorrhage is known (AbouZahr 2003), the increased risk in these women compared to women who do not have a haemorrhage is unknown. After delivery, women may be anaemic for several reasons unrelated to obstetric haemorrhage, for example, because of prior anaemic conditions relating to nutrition, parasitic infections or because of the normal blood loss experienced during an uncomplicated delivery. What this review seeks to quantify is the excess risk of anaemia relating to obstetric haemorrhage. Where possible, we examined evidence for the duration of anaemia after haemorrhage. If a woman has post-partum anaemia for several months after the end of pregnancy, it is important that awareness is raised about such a long-term impairment to her health.

This systematic review examines the epidemiological evidence of an association between obstetric haemorrhage and post-partum anaemia, summarises the magnitude of effect and discusses the methodological issues encountered. The term ‘post-partum’ is used throughout the rest of this study to refer to the period both after abortion and after delivery.

Methods

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Search strategy

We searched the Pubmed, Embase, Popline and Lilacs databases using MeSH terms and free-text words for obstetric haemorrhage, its causes and its sequelae in July 2009 (online-only Appendix S1). This search covered articles published between 1990 and 2009. There were no country or language restrictions. We also searched reference lists of relevant papers. The search included terms related to haemorrhage, blood loss, causes of haemorrhage such as placental abruption and placenta praevia, as well as terms related to sequelae such as anaemia and hysterectomy.

All abstracts were reviewed by KW and VF, and articles likely to contain information on the burden of post-partum anaemia after obstetric haemorrhage were obtained. Data were requested directly from the authors for studies where measurements were described but results not detailed in the publication itself using a standardised form with a cut-off point for anaemia of 12 g/dl.

It was decided a priori to analyse two data sets from prospective cohort studies undertaken in Benin and Burkina Faso. These unique studies compared the post-partum health of women with a near miss and those with a normal delivery in health facilities over 1 year, including women with a live birth, women with a still birth and women who died perinatally. These data sets included women who had a life-threatening haemorrhage as a complication of abortion or birth. Women in Benin were recruited between September 2004 and January 2005; in Burkina Faso, the study was conducted between November 2004 and March 2006. Full methods and main study analyses were reported by Filippi et al. (2007, 2010). Haemoglobin measurements from haemocue test results at 6 and 12 months post-partum were available for analysis from these data sets.

Inclusion criteria

To estimate the effect of haemorrhage on post-partum anaemia, studies were restricted to those including a comparison group. Eligible study designs included cohort, cross-sectional and case–control studies. Randomised control trial (RCT) study designs were also included if data were available for secondary analyses allowing calculation of the prevalence or incidence of anaemia in women who had a haemorrhage compared to those who did not. Acceptable reports of haemorrhage were objective measurement, clinically estimated and a woman’s self-report.

Objective measurement involves the use of a calibrated collection drape, collection of blood and blood clots in a measuring jug and/or weighing of soaked linen and swabs. Anaemia was defined by means of haemoglobin or haematocrit measurements; the levels indicating anaemia were those decided by the researchers in each study, or by other laboratory measurement (anaemia defined as blood loss was not acceptable). Anaemia measurements from a few hours or days after delivery through to 12 months post-partum were included. Where available, anaemia information was captured as both binary (anaemic/non-anaemic) and mean haemoglobin.

Data extraction

Data were extracted on location of study, study dates, study design, participation, definition of haemorrhage, blood loss measurement method, incidence of haemorrhage, case definition of anaemia, anaemia status of women entering the study, timing of anaemia measurement post-partum, number of post-partum anaemia cases, odds ratios, crude and adjusted incidence ratios when available or prevalence ratios if not, 95% confidence intervals, covariates adjusted for and mean haemoglobin measurements. Study populations described in more than one paper were included only once.

Assessment of the quality of studies

Assessment was made of the following aspects of each study: level of participation, loss to follow up in cohort studies, choice of comparison group in case–control studies, method of assessing amount of haemorrhage and control for confounding, following the approach used by The Cochrane Collaboration (Higgins & Green 2011). Each quality ‘domain’ was classified as having a low risk or high risk of generating bias (for example, a clinical estimate or self-report of blood loss would be considered high risk, whereas an objective measurement was categorised as low risk) or an unclear risk if insufficient information was reported to assess this. If all domains were assessed as low risk, the study was classified as at low risk of bias, otherwise it was classified as at high risk.

Data analysis

Forest plots were produced of incidence ratios or prevalence ratios/odds ratios if no information was available regarding antepartum haemoglobin measurements. Between-study heterogeneity was evaluated using I-squared and Cochrane’s Q statistic for heterogeneity. Meta-analyses were conducted using fixed effect or random effect models depending on the results of the heterogeneity tests. Data were analysed using Stata 11.

The studies were divided into high- and low-income country settings as defined by the World Bank, April 2010 (World Bank 2010). Summary estimates were also obtained for different time points (6 and 12 months) for available data.

The Benin and Burkina Faso data sets were analysed to compare women who experienced a near-miss obstetric haemorrhage during delivery (live or dead infant) with women who had an uncomplicated delivery. Haemoglobin levels at 6 months post-partum were utilised (antepartum measurements were not available), as these were considered to represent more accurately the presence of anaemia because of obstetric haemorrhage. Anaemia was defined as <10 g/dl in these data. Women who became pregnant during the 6-month follow-up period were excluded. Prevalence ratios were adjusted for socio-economic status (SES) and parity. SES quintiles were calculated by principal component analysis using a series of household indicators.

When available, mean post-partum haemoglobin levels for women experiencing obstetric haemorrhage were subtracted from mean post-partum haemoglobin levels for women without haemorrhage; 95% confidence intervals were calculated for the difference in the means.

Population-attributable risk was calculated by subtracting the incidence of post-partum anaemia amongst the women who did not haemorrhage from the incidence of post-partum anaemia amongst all women in the study. This was then divided by the incidence of post-partum anaemia amongst all women in the study to arrive at the population-attributable fraction (PAF).

Results

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

The search process is outlined in Figure 1. Data were extracted from nine papers (with authors providing additional data for two of these) (Chan et al. 2001; Walraven et al. 2005) and the two unpublished data sets. Table 1 describes the nine eligible papers and two data sets evaluating the increased risk of post-partum anaemia after obstetric haemorrhage. Study populations were from the United States (n = 4), Hong Kong (n = 2), Singapore (n = 1), Uganda (n = 1), Gambia (n = 1), Benin (n = 1) and Burkina Faso (n = 1). Study populations ranged from women delivering at tertiary referral centres (high-risk women) to specially selected low-risk populations.

image

Figure 1.  Search process for post-partum anaemia articles.

Download figure to PowerPoint

Table 1. Studies included in the systematic review: study populations and results
Author, year of publicationStudy location, study datesStudy designStudy populationNumber of women in studyDefinition of haemorrhageIncidence of haemorrhage in study population (%)Definition of anaemiaAntepartum anaemia statusTiming of anaemia measurement post-partumOverall prevalence of anaemia post-partum (%)Ratio (95% CI) of anaemia incidenceAnaemia prevalence ratio (95% CI)Odds ratio (95% CI)PAF (%)
  1. Hb, haemoglobin; Hct, haematocrit; PAF, population-attributable fraction; RCT, randomised controlled trial; ANC, antenatal care; NA, not available; SES, socio-economic status.

  2. *OR adjusted for iron supplementation, bleeding during pregnancy, malaria, parity and age.

  3. †Prevalence ratio adjusted for SES and parity.

  4. ‡OR adjusted for other complications, history of complications, nulliparity, macrosomia.

  5. §OR adjusted for race, dizziness, laceration degree, forceps and age.¶we use the incidence at discharge for estimating PAF in this study. the prevalence ratio refers to the 6 weeks data.

Sserunjogi et al. (2003)Uganda, 7.1998Cross-sectionalMothers attending health units for immunisation of children aged ≤1 year348A lot of blood31.9Hb <10.0 g/dlNot recorded6 weeks to 12 months post-partum15.8NA1.78 (1.10, 2.88)2.26* (1.09, 4.68)NA
Filippi et al. (2010) (data from this study)Benin, recruited 09.2004–01.2005CohortWomen who delivered in six referral facilities, and live within 25–30 km of the hospitals344 (46 hem, 298 normal delivery)Near miss (shock, hysterectomy or blood transfusion)NAHb <10.0 g/dlNot recorded6 months post-partumNANA3.30† (0.80, 13.63)NANA
Filippi et al. (2007) (data from this study)Burkina Faso, 11.2004–03.2006CohortWomen who delivered at seven public urban and rural hospitals and who lived within 30 km of these hospitals677 (108 hem, 569 normal delivery)Near miss (shock, hysterectomy or blood transfusion)NAHb <10.0 g/dlNot recorded6 months post-partumNANA1.96† (1.30, 2.97)NANA
Lao et al. (1996)Hong Kong, dates not specifiedCohortLow-risk mothers booked before 20-week gestation were recruited at the 28- to 30-week visit to hospital467Not defined5.8Hb <10.0 g/dlWomen with Hb <10 g/dl at the time of booking excluded3 days post-partum12.03.98 (2.34, 6.78)NA6.03 (2.68, 13.61)14.9
Chan et al. (2001)Hong Kong, 02.1998–12.1998Cross-sectionalWomen 20–40 years old with healthy singleton delivered ≥37 weeks gestation in hospital, excluded women with specific medical disorders/history of smoking/drinking.47≥500 ml12.8Hb <10 g/dl in post-natal ward or Hb <12 g/dl at 6 weeksWomen with Hb <11 g/dl at the 1st or 2nd antenatal check-up excludedAt recruitment in post-natal ward for PAF and 6 weeks post-partum for all other measurements14.9NA2.73 (0.68–11.06)3.60 (0.62–22.40)38.2
Nicol et al. (1997)USA, 01.1978–03.1994Cross-sectionalWomen recorded in the University of California Perinatal Database with a vaginal delivery at ≥20 weeks gestation12 397≥500 ml13.8Hct <27% at discharge or women receiving a transfusion regardless of discharge hctWomen with hct <33% at presentation for delivery excludedAt discharge7.05.79 (5.12, 6.55)NA4.59‡ (3.88, 5.42)39.6
Singh et al. (1998)Singapore, 01.1993–12.1993Cross-sectionalWomen delivering at the National University Hospital3728>500 ml9.1Hb <11 g/dl15% of women were anaemic at deliveryAt dischargeNANA1.96 (1.83, 2.11)5.13 (3.96, 6.66)NA
Bodnar et al. (2002) USA, 1997–1999Case–control nested in RCTEnglish-speaking women (mainly low SES) attending a public prenatal clinic. Live singleton birth, return for post-partum visit.497Not defined4.8Hb <11.8 g/dl for aged 12–15 years, Hb <12.0 g/dl for aged ≥15 years4% of women were anaemic at entry to prenatal care4–26 weeks, mean 6.8 weeks post-partum19.1NA1.92 (1.07, 3.45)2.45 (1.02, 5.90)NA
Swaim et al. (1999)USA, dates not specifiedCross-sectionalWomen with uncomplicated vaginal delivery in hospital who had a post-partum Hb measurement because of antepartum Hb <9.0 g/dl, blood loss >500 ml, abnormal orthostatic vital signs or physician’s discretion358>500 ml3.0Hb <9.0 g/dlRecorded but not reported in sufficient detail for the 358 women under study12 h post-partumNANA2.39 (1.40, 4.08)NANA
Petersen et al. (2002)USA, 02.1997–07.1998Case–controlWomen delivering at tertiary referral centre117 anaemic, 198 controls≥500 mlNAHct <26%Not recorded1st day post-partumNANANA5.35§ (2.71, 10.55)NA
Walraven et al. (2005)Gambia,RCTLow-risk women in 26 villages, delivering with traditional birth attendants (active management of labour with misoprostol or ergometrine); a very high proportion of women (above 98%) received iron supplementation in ANC1228 (141 who haemorrhaged and 1087 with normal childbirth)≥500 ml11.5Hb <12 g/dlWomen ineligible if Hb <8 g/dl at antenatal clinic3–5 days post-partum97.9NA1.01 (0.99, 1.03)NANA

Table 1 describes the study populations and definitions of exposure and outcome. Only one study (Walraven et al. 2005) described objectively measuring the amount of blood lost, although Chan et al. (2001) described estimating the volume from a standard kidney dish. In the studies by Chan et al. (2001), Nicol et al. (1997), Singh et al. (1998), Swaim et al. (1999), Petersen et al. (2002) and Walraven et al. (2005); haemorrhage was defined as ≥500 or >500 ml, and for the studies by Lao et al. (1996) and Bodnar et al. (2002) it was not defined. For the studies in Benin and Burkina Faso, the haemorrhage was sufficient for the woman’s state of health to be defined as ‘near miss’ (haemorrhage accompanied by shock, hysterectomy or blood transfusion) (Filippi et al. 2007, 2010). In the Ugandan study, haemorrhage was recorded as the woman’s self-report of ‘a lot of blood’ (Sserunjogi et al. 2003).

Post-partum anaemia was most frequently defined as <10 g/dl of haemoglobin (Lao et al. 1996; Chan et al. 2001; Sserunjogi et al. 2003; Filippi et al. 2007, 2010), with other definitions ranging from <9 to <12 g/dl haemoglobin, haematocrit levels were also used. There was much variation of timing of anaemia measurements post-partum, ranging from in the post-natal ward to 12 months after delivery (Table 1).

Four studies excluded women with antenatal anaemia (Lao et al. 1996; Nicol et al. 1997; Chan et al. 2001; Walraven et al. 2005). Other studies recorded antenatal anaemia status but did not appear to adjust for this in their results (Singh et al. 1998; Swaim et al. 1999; Bodnar et al. 2002).

All studies had at least one quality domain considered to have a potentially ‘high risk’ of bias (Table 2). In many situations, there was insufficient information available so the risk of bias was unclear. High risk of bias related to loss to follow up in two studies (Filippi et al. 2007, 2010), possible bias in the selection of the study population in two studies (Swaim et al. 1999; Walraven et al. 2005), methods of ascertaining blood loss in five studies (Nicol et al. 1997; Chan et al. 2001; Bodnar et al. 2002; Petersen et al. 2002; Sserunjogi et al. 2003) and no adjustment for confounders in the incidence or prevalence ratios in seven studies (Lao et al. 1996; Nicol et al. 1997; Singh et al. 1998; Swaim et al. 1999; Chan et al. 2001; Bodnar et al. 2002; Sserunjogi et al. 2003). Only three published studies adjusted their final odds ratio estimate for confounders (Nicol et al. 1997; Petersen et al. 2002; Sserunjogi et al. 2003). Prevalence ratios were adjusted for confounding (SES and parity) only for the Benin and Burkina data sets.

Table 2. Summary of quality and other methodological issues
Author, year of publicationLoss to follow up/participation in studySelection of comparison group for incidence/prevalence ratio calculationMethod of assessment of haemorrhageAdjustment for confounders
  1. N/A, not applicable; SES, socio-economic status.

Sserunjogi et al. (2003)Only mothers who agreed to participate had measurements takenComparison group comprised the study participants who did not experience haemorrhageMothers’ reportOR adjusted for iron supplementation, bleeding during pregnancy, malaria, parity and age but prevalence ratio unadjusted
QualityUnclear riskLow riskHigh riskHigh risk
Filippi et al. (2010) (data from this study)709 women recruited into study, 509 met criteria of near-miss haemorrhage or uncomplicated delivery, 344 (67%) included in analysis; 163 did not have Hb measurement at 6 months, two further women excluded as pregnant at 6 monthsWomen were sampled from those with uncomplicated hospital deliveriesNear-miss criteriaPrevalence ratio adjusted for SES and parity
QualityHigh riskLow riskLow riskLow risk
Filippi et al. (2007) (data from this study)1025 women recruited into study, 821 met criteria of near-miss haemorrhage or uncomplicated delivery, 677 (82%) included in analysis; 129 did not have Hb measurement at 6 months because they did not return to the clinic for examination (even though they may have been interviewed), 15 further women excluded as pregnant at 6 monthsWomen were sampled from those with uncomplicated hospital deliveriesNear-miss criteriaPrevalence ratio adjusted for SES and parity
QualityHigh riskLow riskLow riskLow risk
Lao et al. (1996)549 women were recruited into the study, 467 (85%) were included in analysis; seven records were incomplete or not available for analysis, 30 women delivered elsewhere, 45 were subsequently found to have anaemia or other blood diseasesComparison group comprised the study participants who did not experience haemorrhageNot knownNone
QualityUnclear riskLow riskUnclear riskHigh risk
Chan et al. (2001)63 women were recruited into the study; 47 (75%) had complete data at both 0 and 6 weeks post-partum. No significant differences in demographic data detected between those excluded and those who remained in the studyComparison group comprised the study participants who did not experience haemorrhageEstimated by midwife/doctor (vaginal) or anaesthetist (caesarean) by viewing blood, clots, and placenta in a standard kidney dish, and accounting for number of blood-soaked gauze swabsNone
QualityLow riskLow riskHigh riskHigh risk
Nicol et al. (1997)30 617 women delivered at ≥20 weeks during the study period; 11 332 had incomplete haematocrit or admission data. Of those with complete data 14 313 delivered vaginally and of these 12 397 (40% overall) had admission haematocrit of ≥33%Comparison group comprised the study participants who did not experience haemorrhageClinical estimateOR adjusted for other complications, history of complications, nulliparity, macrosomia, but incidence ratio unadjusted
QualityUnclear riskLow riskHigh riskHigh risk
Singh et al. (1998)All deliveries includedComparison group comprised the study participants who did not experience haemorrhageNot knownNone
QualityLow riskLow riskUnclear riskHigh risk
Bodnar et al. (2002)867 women were randomised in trial; 497 (57%) delivered a live infant, returned for a post-partum visit and had blood drawn at that visit for determination of Hb concentrationComparison group comprised the study participants who did not experience haemorrhagePhysician’s subjective opinion documented in medical recordHb concentration was adjusted for smoking
QualityLow riskLow riskHigh riskHigh risk
Swaim et al. (1999)Only those with measurements included in this study (see study population, Table 1)Control group is women with Hb taken at the physician discretionNot knownNone
QualityHigh riskHigh riskUnclear riskHigh risk
Petersen et al. (2002) No loss to follow up as retrospective analysis of databaseN/AClinical estimateOR adjusted for race, dizziness, laceration degree, forceps and age
QualityLow riskN/AHigh riskLow risk
Walraven et al. (2005)910 women were enrolled during antenatal care; 120 of these women were excluded because they delivered at home without a traditional birth attendant, and 53 women were excluded because they delivered in a health facility; 492 women were additionally recruited at the time of delivery (they were advised to deliver in hospital but delivered with study)Comparison group comprised the study participants who did not experience haemorrhageObjective measurement of blood loss was used which involved collecting blood loss during delivery in a basin and weighing itNone
 High riskLow riskLow riskHigh risk

All studies except one (Walraven et al. 2005) showed an higher prevalence or incidence of anaemia in women who had experienced haemorrhage than in those who had not, irrespective of the timing of measurement post-partum. The study by Walraven et al. provided data (on request) defining anaemia as <12 g/dl Hb, which includes women with mild anaemia. Combining results for all studies with prevalence or incidence ratios available gave a summary ratio of 2.35 (95% CI 1.44–3.84) although the I-squared statistic was 99.2%, P < 0.001, indicating considerable heterogeneity. The prevalence ratios for the four low-income studies (all African) are shown in Figure 2a; the summary prevalence ratio was 1.58 (95% CI 0.96–2.60), with evidence of heterogeneity (I-squared statistic 83.2%, P < 0.001). However, when these results are stratified by anaemia definition (Figure 2b), the studies excluding mild anaemia from their definition have a higher prevalence ratio (1.93, 95% CI 1.42–2.62) than the study including mild anaemia in its definition. Figure 3 summarises data from Benin and Burkina Faso measured at two time points. There was no evidence that the summary prevalence ratios at 6 and 12 months were different (P = 0.85).

image

Figure 2.  (a) Forest plot of three African studies showing the prevalence ratio of post-partum anaemia following obstetric haemorrhage. (b) Forest plot of four African studies showing the prevalence ratio of post-partum anaemia following obstetric haemorrhage stratified by definition of anaemia.

Download figure to PowerPoint

image

Figure 3.  Prevalence ratios for anaemia following obstetric haemorrhage at 6 and 12 month time points in data from Benin and Burkina Faso.

Download figure to PowerPoint

Figure 4 shows summary estimates for high-income countries, separating prevalence and incidence ratios. The summary prevalence ratio for high-income countries was 1.97 (95% CI 1.83–2.11; I-squared statistic 0%, P = 0.769). The summary incidence ratio for high-income countries was 5.68 (95% CI 5.04–6.40; I-squared statistic 44.8%, P = 0.178), i.e. the risk of anaemia post-partum was nearly six times higher in women who had a haemorrhage than in those who had not. For one study only an odds ratio was available; the odds of having had a haemorrhage were 5.35 times higher in women with post-partum anaemia than in those without (Petersen et al. 2002).

image

Figure 4.  Forest plot of studies from high-income countries showing the prevalence or incidence ratios of post-partum anaemia following obstetric haemorrhage.

Download figure to PowerPoint

Only one study provided the data to calculate the prevalence and incidence ratios for the same population. In the study by Nicol et al. (1997), excluding the 1916 (13%) women with haematocrit at admission of <33% resulted in an incidence ratio of 5.8 (95% CI 5.1–6.6) (Table 3). Including these women in the analysis resulted in a prevalence ratio of only 2.0 (95% CI 1.9–2.2).

Table 3. Estimates of incidence ratio and prevalence ratio from study by Nicol et al. (1997)
 Study participants excluding 1916 women with antepartum anaemiaStudy participants including 1916 women with antepartum anaemia (calculated data)*
Anaemic post-partumNon-anaemic post-partumTotalAnaemic post-partumNon-anaemic post-partumTotal
  1. *This section assumes that women with antepartum anaemia are all anaemic post-partum and have the same risk of haemorrhage as women without anaemia.

Haemorrhage4151298171368012981978
Non-haemorrhage44710 23710 864209810 23712 335
Total86211 53512 397277811 53514 313
 Incidence ratio = 5.8Prevalence ratio = 2.0
95% confidence interval(5.1–6.6)(1.9–2.2)

Population-attributable fractions were calculated for the three studies for which incidence ratios were available (Table 1, final column); these ranged from 14.9% to 39.6%. However, these estimates should be interpreted with caution as they were based on estimates of the incidence of anaemia in the study populations, rather than the population at large, and some of these study populations were very small.

Mean haemoglobin differences are summarised in Figure 5. Four of five studies with available data showed a significantly lower mean haemoglobin level in women who had experienced a haemorrhage compared to women who had not. The overall summary mean difference was 0.68 g/dl (95% CI 0.53–0.83; I-squared statistic 54.3%, P = 0.053). The study reported by Swaim et al. (1999) measured haemoglobin at 12 h after delivery, whereas the other studies took measurements between 3 days and 12 months after.

image

Figure 5.  Forest plot of studies that recorded the difference in mean post-partum haemoglobin level in women who did not experience a haemorrhage compared with those who did. Mean haemoglobin levels among women who did not experience haemorrhage were 10.8 g/dl (Swaim et al. 1999), 11.5 g/dl (Sserunjogi et al. 2003), 12.4 g/dl (Filippi et al. 2010; Benin) and 11.8 g/dl (Filippi et al. 2007; Burkina Faso).

Download figure to PowerPoint

Discussion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

The studies confirm that women who experience an obstetric haemorrhage, particularly when severe, have a higher prevalence or incidence of anaemia post-partum than those who do not have a haemorrhage. Women with obstetric haemorrhage were 2.35 times more likely to be anaemic post-partum than women without obstetric haemorrhage. Excluding women with antenatal anaemia from the study populations resulted in a sixfold increase in post-partum anaemia associated with obstetric haemorrhage. There is evidence that the increased anaemia prevalence can persist for many months after delivery, as demonstrated in Benin and Burkina Faso.

The high incidence of anaemia a few days after obstetric haemorrhage is not surprising, because acute blood loss directly depletes the haemoglobin in the blood. The summary estimates for high-income countries are strongly influenced by two large studies in the USA and Singapore (Nicol et al. 1997; Singh et al. 1998), which measured haemoglobin at discharge, and it is uncertain whether these women would have remained anaemic in the long term, even in the absence of treatment. A declining blood volume with a rise in haematocrit is usually seen 3–7 days after a normal vaginal delivery (Decherney & Pernoll 1994), but women generally recover spontaneously. The very high prevalence of anaemia 6 months after delivery in women in Benin and Burkina Faso, however, cannot be explained on physiological grounds only. These women experienced a massive blood loss resulting in near miss, and they would have needed treatment to restore their haemoglobin levels. However, not all women can afford hospital treatment of obstetric complications, and many of the women in these African samples were discharged too early or whilst still unwell, as demonstrated by the high maternal mortality post-discharge in the Burkina Faso sample (Filippi et al. 2007; Storeng et al. 2008). Little is known about the health status of women at discharge from hospital in poor countries, particularly after an obstetric complication (World Health Organization 2010).

Our review has some limitations. Only a small number of studies were available, and the majority were not designed to explore the increased risk of post-partum anaemia in women experiencing haemorrhage. Our search strategy may have missed articles that found no effect of blood loss on anaemia because these studies may not have mentioned anaemia in the abstract. Most of the studies relied on visual estimation of blood loss, which underestimates blood loss compared to objective measurement. This could have resulted in underestimation of the risks presented (Prasertcharoensuk et al. 2000; Benchimol et al. 2001). In addition, women’s self-reporting as used by the study by Sserunjogi et al. (2003) has previously been found to have poor agreement with medical diagnosis (Ronsmans et al. 1997). Variations in the definitions of anaemia used were a further source of heterogeneity and very few studies adjusted for confounders in the prevalence or incidence ratio estimates. Some women may have presented with clinical signs and symptoms of post-partum anaemia and may have received treatment for this prior to measurement of their haemoglobin level for the study, which could have reduced the observed effect of haemorrhage. However, our findings are consistent with evidence from RCTs of the effect of the management of the third stage of labour on post-partum haemorrhage. All RCTs showing a reduction in the incidence of post-partum haemorrhage after active management (Begley et al. 2010) or uterotonics (Cotter et al. 2001) consistently also found a reduction in the incidence of post-partum anaemia.

Evaluating the increased risk of post-partum anaemia following obstetric haemorrhage requires the measurement of both antepartum and post-partum haemoglobin levels. Table 3 demonstrates the difficulty in relying on prevalence ratio estimates, in that inclusion of women who were anaemic prior to delivery leads to an underestimate of the magnitude of effect of haemorrhage on risk of anaemia. This underestimation becomes more pronounced as the prevalence of antepartum anaemia increases (data not shown). Measuring antepartum haemoglobin may therefore be particularly important for populations with several additional causes of anaemia in the population, such as malaria, hookworm and poor nutrition, where the contribution of maternal haemorrhage as a cause of anaemia will be less pronounced when studying prevalence ratios. On the other hand, measuring the difference in mean haemoglobin estimates post-partum instead of a binary measure of anaemia may avoid the need for antepartum haemoglobin estimates. Because antepartum haemoglobin levels can be assumed to be the same for women who subsequently haemorrhage and those who do not, the effect of haemorrhage can be assessed by comparing actual haemoglobin levels post-delivery in the two groups. A further advantage of using mean haemoglobin measurements is that they overcome some of the criticisms of using ‘cut-off’ points to define anaemia. Using a fixed cut-off such as <10 g/dl to categorise women as anaemic or non-anaemic does not enable us to see an effect when both groups are categorised as anaemic post-partum, for example, a change from 9.5 g/dl (measured in the non-haemorrhage group) to 7 g/dl (in the haemorrhage group), or when both groups are non-anaemic post-partum, such as a change from 12 to 10 g/dl (which may still be physiologically important). The impact of selecting either a < 10 or <12 g/dl cut-off point can be clearly seen in Figure 2b. The substantially lower mean haemoglobin in women experiencing a haemorrhage in the study by Walraven et al. suggests that the prevalence of moderate anaemia will have increased in these women. However, the inclusion of women with mild anaemia in the definition could have obscured this difference.

Population-attributable fractions were only calculated for the four studies for which incidence ratios were available. These estimates are wide ranging and should be interpreted with caution because some of the studies involved are small and the methods and definitions differ. The largest PAF was calculated from the results of the study by Nicol et al. (1997) and suggests that more than a third of women with post-partum anaemia could have experienced anaemia as a consequence of haemorrhage. PAFs are driven in large part by the prevalence of obstetric haemorrhage in the population. The prevalence of post-partum haemorrhage is thought to range from 1.05% in Western Asia to 18.67% in Middle Africa (Carroli et al. 2008).

Conclusion

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

In conclusion, this review shows that whilst the majority of studies show a clear relationship between post-partum haemorrhage and an increased risk of anaemia, there are several factors relating to study design which can have a dramatic impact on the final result. The control of maternal anaemia is a very important public health topic in developing countries. However, most of the focus has been on anaemia during pregnancy, and on iron and folic acid deficiency, other nutrient deficiencies, hookworm infection and malaria infection. Similarly, the content of post-partum care has only been recently revisited by the World Health Organization, with the recognition that the evidence base for the care of discharged women who have had an obstetric complication is very limited (World Health Organization 2010). It is particularly important to know how long women may suffer from untreated post-partum anaemia after a haemorrhage in the design of post-natal care programmes in low-income countries; this review suggests that the duration of post-partum anaemia could be many months. We therefore urge more cohort studies to quantify the duration and risk factors for post-partum anaemia, as well as other post-partum complications.

Acknowledgements

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

This research was funded through a grant made to the Child Health Epidemiology Reference Group (CHERG) by the Bill and Melinda Gates Foundation. The authors thank Guillermo Carroli, Eugene Oteng-Ntim, Lale Say, Herbert Peterson, Doris Chou, Gretchen Stevens and Robert Black for their comments and suggestions. The authors also thank Ruth Chan, Laurie Swaim, Jennifer Blum and Gijs Walraven for providing additional data to the World Health Organization for this review.

References

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information
  • AbouZahr C (2003) Global burden of maternal death and disability. British Medical Bulletin67, 111.
  • Begley CM, Gyte GML, Murphy DJ, Devane D, McDonald SJ & McGuire W (2010) Active versus expectant management for women in the third stage of labour. Cochrane Database of Systematic Reviews (7), CD007412. DOI: 10.1002/14651858.CD007412.pub2.
  • Benchimol M, Gondry J, Mention JE, Gagneur O & Boulanger JC (2001) [Role of misoprostol in the delivery outcome]. Journal de Gynécologie Obstétrique et Biologie de la Reproduction (Paris)30, 576583.
  • Bodnar LM, Siega-Riz AM, Miller WC, Cogswell ME & McDonald T (2002) Who should be screened for postpartum anemia? An evaluation of current recommendations. American Journal of Epidemiology156, 903912.
  • Carroli G, Cuesta C, Abalos E & Gulmezoglu AM (2008) Epidemiology of postpartum haemorrhage: a systematic review. Best Practice & Research Clinical Obstetrics & Gynaecology22, 9991012.
  • Chan SM, Nelson EA, Leung SS & Li CY (2001) Postnatal iron status of Hong Kong Chinese women in a longitudinal study of maternal nutrition. European Journal of Clinical Nutrition55, 538546.
  • Cotter AM, Ness A & Tolosa JE (2001) Prophylactic oxytocin for the third stage of labour. Cochrane Database of Systematic Reviews (4), CD001808.
  • Decherney AH & Pernoll ML (1994) Current Obstetric & Gynecologic Diagnosis & Treatment. Appleton & Lange, East Norwalk, CT.
  • Filippi V, Ganaba R, Baggaley RF et al. (2007) Health of women after severe obstetric complications in Burkina Faso: a longitudinal study. Lancet370, 13291337.
  • Filippi V, Goufodji S, Sismanidis C et al. (2010) Effects of severe obstetric complications on women’s health and infant mortality in Benin. Tropical Medicine & International Health15, 733742.
  • Higgins JPT & Green S (eds) Cochrane Handbook for Systematic Reviews of Interventions Version 5.1.0 [updated March 2011]. The Cochrane Collaboration, 2011. Available at: http://www.cochrane-handbook.org.
  • Khan KS, Wojdyla D, Say L, Gulmezoglu AM & Van Look PF (2006) WHO analysis of causes of maternal death: a systematic review. Lancet367, 10661074.
  • Lao TT, Lee CP & Mak WP (1996) Postpartum anaemia is not related to maternal iron status in the third trimester. European Journal of Obstetrics & Gynecology and Reproductive Biology64, 710.
  • Nicol B, Croughan-Minihane M & Kilpatrick SJ (1997) Lack of value of routine postpartum hematocrit determination after vaginal delivery. Obstetrics & Gynecology90, 514518.
  • Petersen LA, Lindner DS, Kleiber CM, Zimmerman MB, Hinton AT & Yankowitz J (2002) Factors that predict low hematocrit levels in the postpartum patient after vaginal delivery. American Journal of Obstetrics & Gynecology186, 737744.
  • Prasertcharoensuk W, Swadpanich U & Lumbiganon P (2000) Accuracy of the blood loss estimation in the third stage of labor. International Journal of Gynaecology and Obstetrics71, 6970.
  • Ronsmans C, Achadi E, Cohen S & Zazri A (1997) Women’s recall of obstetric complications in south Kalimantan, Indonesia. Studies in Family Planning28, 203214.
  • Singh K, Fong YF & Arulkumaran S (1998) Anaemia in pregnancy – a cross-sectional study in Singapore. European Journal of Clinical Nutrition52, 6570.
  • Sserunjogi L, Scheutz F & Whyte SR (2003) Postnatal anaemia: neglected problems and missed opportunities in Uganda. Health Policy and Planning18, 225231.
  • Storeng K, Baggaley R, Ganaba R, Ouattara F, Akoum M & Filippi V (2008) Paying the price: the costs and consequences of emergency obstetric care in Burkina Faso. Social Science & Medicine66, 545557.
  • Swaim LS, Perriatt S, Andres RL, Paradissis J & Watson MN (1999) Clinical utility of routine postpartum hemoglobin determinations. American Journal of Perinatology16, 333337.
  • Walraven G, Blum J, Dampha Y et al. (2005) Misoprostol in the management of the third stage of labour in the home delivery setting in rural Gambia: a randomised controlled trial. BJOG: An International Journal of Obstetrics & Gynaecology112, 12771283.
  • World Bank List of Economies (2010) April 2010. Available at: http://siteresources.worldbank.org/DATASTATISTICS/Resources/CLASS.XLS Accessed July 2010.
  • World Health Organization (2010) WHO Technical Consultation on Postpartum and Postnatal Care. WHO, Geneva.

Supporting Information

  1. Top of page
  2. Abstract
  3. Introduction
  4. Methods
  5. Results
  6. Discussion
  7. Conclusion
  8. Acknowledgements
  9. References
  10. Supporting Information

Appendix S1. Pubmed anaemia search strategy: July 2009.

FilenameFormatSizeDescription
tmi2883_sm_appendixs1.docx12KSupporting info item

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.