Background

“Macrosomia” is a term that describes a very large fetus or neonate. The condition may be caused by constitutional/genetic factors, maternal obesity and/or excessive gestational weight gain, or maternal hyperglycemia due to pre-existing diabetes or gestational diabetes that were not adequately controlled [1]. In low-and middle- income countries (LMICs) or settings where antenatal care is sub–optimal, poorly controlled diabetes or undiagnosed gestational diabetes may be a more important cause for macrosomia than in high-income countries, where antenatal care is better. In high-income countries, the prevalence of macrosomia has been increasing in the last two to three decades [2, 3]. But in many LMICs, macrosomia is still not perceived to have the same priority as other public health problems (e.g., HIV) [4]. However, with the increasing prevalence of maternal obesity and diabetes [5, 6] a parallel increase in macrosomic infants might be expected in LMICs. Complicated deliveries related to macrosomia could lead to more severe adverse outcomes in resource-poor settings due to limited availability of obstetric care. Thus, a precise definition of macrosomia that is more predictive of maternal and perinatal mortality and morbidity is needed. In this study, we aimed to explore a definition through an outcome-based approach and comparing commonly used definitions currently.

Currently, no consensus definition exists among researchers and obstetricians. The most commonly used definition is based on birthweight cut-off points (e.g., 4000 g or 4500 g) [3, 7–9]. As it is increasingly recognized that racial variation in birthweight is substantial, more and more studies are using specific birthweight percentiles as cut-off points at a given gestational week (e.g., P90 or P97) based on the concept of large-for-gestational-age (LGA) [2, 10]. Furthermore, most studies on macrosomia have focused on Caucasian populations in high-income countries, and very few studies on the topic focus on LMICs [2].

In this study we analyzed data from 23 LMICs in Africa, Asia, and Latin America that participated in the World Health Organization (WHO) Global Survey on Maternal and Perinatal Health (2004–2008). We assessed commonly used definitions of the term ‘macrosomia’ through an outcome-based approach. Two types of definitions were compared: one based on empirical absolute birthweight and the other on the country-specific birthweight percentile at each gestational week. We aimed to identify a definition that was more predictive of maternal and perinatal mortality and morbidity in term pregnancies in LMICs, which also takes into account regional variation.

Methods

Study design and data extraction

The general objective of the WHO Global Survey on Maternal and Perinatal Health was to create a global database on health services and outcomes for maternal and perinatal health, which concentrated on the relationship between mode of delivery and perinatal outcomes [2, 11]. This survey has previously been described in detail elsewhere [2, 11, 12]. A total of 373 facilities in 24 countries in Africa, Asia and Latin America participated in this survey. Data collection was carried out in 2004–05 in Africa and Latin America, and in 2007–08 in Asia. Trained data collectors extracted data from medical records and completed standardized forms. Gestational age was calculated based on the difference between the estimated and actual delivery date in the medical records. Data related to outcomes were obtained until discharge from the hospital. Maternal weight was defined differently as described previously: in Africa and Latin America, maternal weight was the weight recorded at the first antenatal care visit, while in Asia it was the weight at the last visit before delivery [2, 11].

This is a secondary data analysis using data from the WHO Global Survey on Maternal and Perinatal Health. The protocol of this survey was approved by the ethics committees at the WHO and in all participating centres [11]. We obtained permission to use this data from Department of Reproductive Health and Research, WHO. An individual informed consent was not obtained because in this survey data were extracted from medical records without individual identification [11].

For the purpose of this analysis, the study sample was restricted to: 1) LMICs; 2) singleton pregnancies; 3) live births or fresh stillbirth; 4) birthweight ≥ 1000 g; and 5) term births (gestational age 37–42 weeks). As the analysis focused on LMICs, Japan was excluded. Macerated stillbirths were also excluded, as we were interested in mortality associated with delivery, but not pre-delivery mortality. Infants who had missing information on birthweight or gestational age were also excluded. The sample selection process is shown in Fig. 1.

Fig. 1
figure 1

Flow chart of inclusion and exclusion of study subjects

Full size image

Statistical analysis

Previous studies have shown that the perinatal mortality rate decreases with increasing birthweight or birthweight percentile until it reaches an inflection point, after which a further rise in birthweight increases the mortality, i.e., a reversed J-shaped mortality curve [13–17]. We applied this principle to our analysis: macrosomia was defined as a birthweight or birthweight percentile that exceeds the nadir of the mortality curve, and at which point the relative risk of perinatal mortality is twice or greater than that of the nadir. We used a priori odds ratio of 2 · 0 as a criterion for clinical significance, as per previous studies [18, 19].

We used the birthweight range of 3000–3499 g or at 50th–74th percentiles as the reference groups because our exploratory analyses showed that the nadir of the birthweight-specific mortality rate fell into these ranges. According to the exploratory analyses, there was no significant difference of risks of neonatal mortality and morbidity among subgroups of every 100 g between 3500 and 4000 g. Thus, infants with birthweights in the range of 3500–3999 g were combined as one subgroup. Birthweights between 4000 and 4500 g were categorized into subgroups of every 100 g in order to explore the cut-off point for the definition of macrosomia.

To calculate weight percentiles, we adopted the global reference for fetal and birth weight percentile [20]. Briefly, based on the mean birthweight and variation at 40 weeks of gestation at each country, country-specific, equation-derived (i.e., not empirical) birthweight reference percentiles at 75th, 90th, 95th and 97th for each gestational week were generated. Infant birthweight was categorized according to these references.

The adverse maternal outcomes included maternal mortality and severe morbidity defined as any of the following: admission to an intensive-care unit (ICU), referral to a higher level or special care unit, blood transfusion, hysterectomy, vesico-vaginal/recto-vaginal fistula and third or fourth degree perineal laceration. The adverse perinatal outcomes were still-birth, early neonatal death (neonatal death within 7 days of birth) and severe neonatal morbidity defined as any of the following: admission to an Neonatal ICU for 7 days or more, referral to a higher level or special care unit and 5-min Apgar score less than four. As maternal and perinatal mortality and severe morbidity are rare outcomes, the composite maternal mortality and morbidity index (MMMI) and perinatal mortality and morbidity index (PMMI) were computed [2]. The MMMI and PMMI were coded as an event if mortality or any of the corresponding severe morbidity occurred. We also compared the risks of caesarean section and assisted vaginal delivery (forceps/vacuum extraction) for subgroups of infants by birthweight and birthweight percentile.

Two-level logistic regression models were used to estimate odds ratios (ORs) of maternal and perinatal mortality and morbidity by absolute birthweight and birthweight percentile for gestational age. Facilities represented units at level two and individuals within facilities were observations at level one. To take into account the large variations of anthropometry among regions, we stratified the analysis by region (Africa, Asia and Latin America). We adjusted for country, maternal age, marital status, education (total years of school attendance), obesity (maternal body mass index, BMI ≥30 kg/m2), pre-existing diabetes, parity, infant sex, and gestational age as potential confounders according to previous literature [7, 10]. All analyses were conducted with SAS version 9 · 2 (SAS Institute Inc, Cary, NC).

Maternal weight and/or height, and consequently maternal BMI, were missing in more than 10 % of the study population in Kenya (85 %), Brazil (67 %), Angola (43 %), Argentina (33 %), Uganda (21 %) and Peru (13 %). A missing category within “maternal obesity” (BMI ≥ 30 kg/m2) was therefore created and included in all regression models. In our sensitivity analysis, we restricted our calculations to countries with less than 10 % of the missing value for BMI. The “(pre-existing) diabetes mellitus” variable was missing in less than 1 % of the study population of each region. For the estimation of risks of MMMI and PMMI by birthweight or birthweight percentile, we performed a sensitivity analysis by excluding the variables “maternal obesity” and “diabetes” from the regression models. In another sensitivity analysis, we compared the adjusted odds ratios (aORs) of MMMI and PMMI in subgroups using birthweight percentile of the study population as cut-off points, the results were essentially unchanged (data not shown). In another sensitivity analysis, we compared the aORs of MMMI and PMMI in subgroups using the empirical country- and gestational-age-specific birthweight percentile of the study population as cut-off points, the results were essentially unchanged (not shown).

Results

A total of 246,659 deliveries at 363 facilities in 23 LMICs were included in this analysis (Fig. 1). Large variations in birthweight distribution were observed among infants in the three regions. Mean birthweight was 3037 g, 3225 g and 3253 g in Asia, Africa and Latin America, respectively, and the proportion of infants with a birthweight greater than 4500 g was 0 · 3 %, 1 · 2 %, and 0 · 7 % across these three regions. The prevalence of birthweight greater than the 97th percentile was 7.2 % (Latin America), 10.1 % (Africa) and 10 · 5 % (Asia) in the three regions (Table 1). In all three regions, the proportion of maternal age older than 35 years, maternal obesity (BMI ≥30 kg/m2), maternal preexisting gestational diabetes, multiparas and infant male sex was positively associated with higher birthweight (all p values <0 · 0001, Additional file 1: Table S1).

Table 1 Country-specific birthweight distribution of singleton term births
Full size table

The overall rates of caesarean deliveries (and intrapartum rate) in Africa, Asia and Latin America were 12 · 0 % (6 · 2 %), 28 · 0 % (12 · 5 %) and 34 · 1 % (12 · 0 %) (p < 0 · 0001), respectively. Table 2 shows that compared with the reference group (3000–3499 g), aORs of elective and intrapartum caesarean section exceeded 2 · 0 when birthweight was greater than 4000 g in all three regions, while the risks of forceps or vacuum extraction did not rise significantly in most subgroups of infants in Asia and Latin America. In Africa, the aORs of forceps or vacuum extraction reached 2.0 when birthweight exceeded 4500 g. There was a large variation of elective caesarean section rate among populations. In Africa, only around 5 % suspected macrosomic babies were born through elective caesarean section while in Asia and Latin America, the responding rate was as high as around 40 and 20 %, respectively. The association between caesarean section and birthweight seemed less pronounced when infants were categorized by birthweight percentile and aORs hardly reached 2.0 (Table 3). The most important indication for CS for suspected macrosomic cases in all populations were cephalopelvic disproportion, followed by previous caesarean and fetal distress. Around 40 % of suspected macrosomic cases born through caesarean had an indication of previous caesarean section. The other two indications accounted for around 30 and 20 % caesarean deliveries, respectively (Additional file 2: Table S2).

Table 2 Prevalence and odds ratios of elective caesarean, intrapartum caesarean section and forceps or vacuum extraction by birthweight
Full size table
Table 3 Prevalence and odds ratios of elective caesarean, intrapartum caesarean section and forceps or vacuum extraction by birthweight percentile
Full size table

The birthweight-specific risks of MMMI and PMMI are presented in Fig. 2. The association is expressed in a reverse “J”- or “U”-shaped curves. In all three regions, the lowest risk of MMMI corresponded to the birthweight range of 3000–3500 g, which was used as the reference group. When the birthweight exceeded 3500 g, aORs of MMMI increased gradually. The aORs of MMMI reached 2 · 0 when birthweight was greater than 4000 g, 4500 g and 5000 g in Asia, Africa and Latin America, respectively. The rise in PMMI risks lagged behind that of MMMI in all three regions (Fig. 2). The aORs of PMMI reached 2 · 0 when birthweight was greater than 4200 g in Asia and 5000 g in Africa. In Latin America, birthweight of 4500–4999 g corresponded to the aOR of PMMI 1 · 78 (95 % CI: 1 · 16, 2 · 74). When birthweight was greater than 5000 g, aOR rose dramatically to 7 · 40 (95 % CI: 3 · 5, 15 · 66) (Table 4).

Fig. 2
figure 2

Adjusted odds ratios of maternal and perinatal mortality and morbidity by birthweight in singleton term births. a: Africa b: Asia c: Latin America reference category: 3000 –3499 g

Full size image
Table 4 Prevalence and odds ratios of maternal and perinatal mortality and morbidity by birthweight
Full size table

When we excluded prelabor caesarean deliveries or restricted the analyses to vaginal deliveries, the results remained essentially unchanged (Additional file 3: Tables S3 and S4).

Table 5 shows that the risks of MMMI and PMMI in infants with a birthweight greater than the 95th percentile increased slightly compared with that of birthweight at the 50th–75th percentiles in all three regions. When birthweight was at the 97th percentile or higher, the aORs of MMMI and PMMI increased significantly, but none exceeded 2 · 0 in any region. In addition to the equation-derived global reference birthweight percentiles, we also used the empirical country- and gestational-age-specific birthweight percentile of the study population as cut-off points. The results were essentially unchanged (not shown).

Table 5 Prevalence and odds ratios of maternal and perinatal mortality and morbidity by birthweight percentile
Full size table

We conducted sensitivity analyses using mortality and morbidity as two separate outcomes. The mortality included maternal and perinatal deaths. For morbidity, the occurrence of any components of MMMI and PMMI (excluding maternal or perinatal death) was considered as a positive event. As maternal and perinatal mortality were rare, no significant differences in mortality were found in the subgroups of infants with a birthweight greater than 4500 g, or the 97th birthweight percentile compared with the reference groups. However, results were similar for maternal and perinatal morbidity after excluding maternal or perinatal death (data no shown). Similar trends for the risks of MMMI and PMMI by birthweight and birthweight percentile were found when we restricted the analyses to countries with less than 10 % of missing value for BMI (Additional file 3: Tables S5 and S6). When maternal obesity and diabetes were excluded from the regression models, aORs became larger for most of the subgroups, but the patterns remained the same in the three regions (Additional file 3: Tables S7 and S8).

Discussion

Our results indicate that there is a significant increase in adverse maternal and perinatal outcomes when the birthweight of term infants (37–42 weeks) reaches 4500 g in African and Latin America, and 4000 g in Asia. These cut-offs could, therefore, be used to define ‘macrosomia’ in these settings. Our findings do not support using LGA as a new definition for macrosomia because LGA was less predictive of adverse outcomes.

The use of the cut-off point of 4500 g to define macrosomia is consistent with results of previous studies of Caucasians populations [1, 9], and supports the definition of the American Congress of Obstetricians and Gynecologists [1]. Based on analyses of national datasets of the United States, Zhang et al. [7], found that infants with birthweight of 4500–4999 g were at significantly increased risks of stillbirth, neonatal mortality (especially because of birth asphyxia), morbidity, and caesarean delivery as a consequence of either slow labor progress or non-reassuring cardiotocography. Ye et al. [9] used the same database also found that risks of MMMI or PMMI did not increased significantly until birthweight was at the 97th percentile or higher. A birthweight cut-off points irrespective of gestational age (4500 g in Whites, 4300 g in Blacks and Hispanics) is more predictive of mortality and morbidity outcomes than the 97th percentile for a given gestational age.

We used an OR of 2 · 0 for either MMMI or PMMI as a priori criterion to identify clinically important macrosomia. Though arbitrary, this cut-off point was also used in a study by Boulet et al. [18] in defining clinically important fetal growth restriction. In the randomized trial of the Twin Birth Study Collaborative Group, a relative risk of 0 · 5 was also used to justify the smallest clinically important difference between the planned caesarean delivery group and control group [19], equivalent to two-fold increase (or decrease) of risks.

Using the concept of birthweight percentile at a given gestational age (i.e., LGA) as the definition of macrosomia has been proposed in recent years [2, 10, 21]. However, our study shows that it has a poor prediction of adverse maternal and perinatal outcomes. This may be partially attributable to the imprecise estimation of gestational age, which has resulted in misclassifications of macrosomia. Thus, a definition based on birthweight would be more practical, especially in settings where accurate estimation of gestational age may be difficult.

We observed that risks of elective and intrapartum caesarean section increased significantly for infants with a birthweight greater than 4000 g in the selected facilities of all three regions. However, the risks of or vacuum extraction did not increase prominently, which suggested that obstetricians may have a tendency for operative delivery when faced with a suspiciously large fetus and slow labor progress. Therefore, an estimated birthweight of 4000 g may be a useful indicator for difficult labour. We also found that risks of MMMI and PMMI did not change substantially after excluding elective caesarean deliveries, suggesting that suspected macrosomic cases may not benefits from elective caesarean section in a meaningful way. This was also demonstrated in other studies [22, 23].

We also observed that risks of MMMI increased prior to that of PMMI in three regions. This was consistent with the three-level definition of macrosomia proposed by Boulet et al. [3]: Grade 1 (>4000 g) to identify increased risks of labour and newborn complications; Grade 2 (>4500 g) to predict neonatal morbidity; and Grade 3 (>5000 g) to predict infant mortality. The major causes of maternal mortality and morbidity for mothers of macrosomic babies included uterine atony, prolonged labour, haemorrhage, vesico-vaginal/recto-vaginal fistula and severe perineal laceration [21]. These complications contributed to MMMI in our study. The most frequently severe adverse outcome was admission to an intensive-care unit (5.0 %), followed by blood transfusion (2.0 %) and third or fourth degree perineal laceration (1.5 %) for suspected macrosomia infants.

Macrosomia represents a significant obstetric challenge. The definition of macrosomia has important clinical, medicolegal and cost implications. Therefore, it should be evidence-based, particularly in resources limited countries where skilled birth attendants and caesarean delivery may not be readily available, and hospital transfer and special care are costly. To the best of our knowledge, this is the first study to search for an evidence-based definition of macrosomia in LMICs.

However, our study has several limitations. First, the significance of the definition of macrosomia using birthweight cut-off point in obstetric management is limited by the inaccuracy in birthweight estimation. Either ultrasound or clinical prediction of birthweight is not accurate enough to serve as the basis for obstetric decision making. But obstetricians are prone to operative delivery when faced with a suspected large fetus, which may lead to unnecessary cesarean [24].

Second, despite the standardization of data collection, participating facilities may have different labour management protocols. Inter-institutional variability is inevitable due to the nature of a multinational study. For example, screening for diabetes in pregnancy is not available in all facilities, particularly in Africa [2], and therefore underestimation of gestational diabetes is likely. However, our sensitivity analysis, which excluded the variable “diabetes” from the regression models, showed a similar result to that of the fully adjusted models.

Third, information on maternal height and weight was problematic in two aspects. More than 10 % of data for height and/or weight were missing for some countries. In addition, maternal weight was defined differently across regions: in Africa and Latin America it was referred to as the first recorded weight at the first antenatal care visit while in Asia it was defined as the last recorded weight before delivery. In theory, the last recorded maternal weight would better control for the potential confounding effects of gestational weight gain in Asia [25]. However, the results of the sensitivity analysis, which excluded those countries or the variable “obesity” from the regression model indicated that the limitation did not affect the conclusion.

Finally, the sample used in the survey was selected from facilities with more than 1000 deliveries per year and where caesarean sections were available [11]. Selection bias is possible, especially in certain LMICs where an institutional delivery rate is low. Therefore, the generalizability of our findings may be limited.

Conclusions

A population-specific definition of macrosomia using birthweight cut-off points (4500 g in Africa and Latin America, 4000 g in Asia) for term infants at 37–42 gestational weeks is more closely associated with maternal and perinatal mortality and morbidity. This definition is also easier to apply than that based on birthweight percentile for a given gestational age. The use of an evidence-based definition of macrosomia may improve obstetric and perinatal care, especially in resource-limited settings in LMICs.