With a new symphysis-fundus height (SFH) growth chart, based on Mozambican women with ultrasound-dated singleton pregnancies, the aim was to examine the possibility to enhance sensitivity of predicting small for gestational age (SGA) newborns by attempts to adjust the chart for parity and for mid-upper-arm circumference (MUAC).
methods Two antenatal clinics were chosen in the suburban area of Maputo City. A cohort of 904 consecutively recruited antenatal clients was followed until birth. Gestational age was determined by ultrasound at enrolment. The growth of the SFH was measured every 2–3 weeks. Women with multiple pregnancy or with gestational age >21 weeks at enrolment were excluded. Attempts were made to adjust SFH measurements for parity and MUAC by developing a mathematical model to increase sensitivity of the SFH method to predict a foetus being SGA.
results Parous women had on average 0.5–1 cm higher SFH readings than nulliparous women. Women with a body mass index (BMI) <19 and women with BMI >27 had approximately 1 cm lower and 1 cm higher readings, respectively, compared with women of normal BMI. There was a significant correlation between BMI and MUAC (r = 0.621; P < 0.001). The usefulness of SFH measurements to predict SGA newborns was analysed. The sensitivity was 49%, the specificity was 66%, the positive predictive value was 14% and the negative predictive value, 93%. By using the correlation between BMI and MUAC we tried to find a simple and useful method to improve the sensitivity of SFH to detect SGA foetuses. By reducing the SFH measurement by 1 cm for women with MUAC >29 and by 1 cm for multiparous women the sensitivity raised to 65% at the expense of reducing the specificity to 51%. Using a linear function of BMI, MUAC and parity to adjust the SFH measurement for each individual woman, it was possible to get a sensitivity of 70% with a corresponding specificity of 56%.
conclusion By using BMI, MUAC and parity, it might be possible to improve the sensitivity of the SFH growth chart in predicting newborn being SGA but mostly at the expense of specificity.
The rates of perinatal morbidity and mortality are high in Mozambique (Axemo 1995). Interventions with the purpose of reducing these rates in low-income countries should focus upon early detection and management of conditions associated with recognizable risk factors affecting maternal and foetal health (Conde-Agudelo et al. 2000). Having low birth weight (<2500 g) or being small for gestational age (SGA) newborn is the most significant single determinant of perinatal morbidity and mortality in these countries (Hofvander 1982; Mavalankar et al. 1994; Read & Clemens 1995; Conde-Agudelo et al. 2000). To detect a foetus being SGA in advance is still difficult, even in high-income countries where ultrasound is available. The sensitivity of ultrasound to detect the SGA foetus is about 60% (Warsof et al. 1986; Laurin 1987; Marsal 1992).
In low-income countries, where ultrasound technology is mostly unavailable, foetal growth can only be monitored by serial measurements of the symphysis-fundus height (SFH). However the sensitivity of detecting SGA is poor and varies from 26% to 56% (Westin 1977; Wallin et al. 1981; Persson et al. 1986). An SFH growth chart is a simple and cheap screening method for detection of multiple pregnancies and polyhydramnios (Westin 1977; Munjanja et al. 1987; Steingrimsdóttir et al. 1995), but its value to predict SGA foetuses and among them foetuses with intra uterine growth retardation (IUGR) has been and is still under discussion (Cnattingius et al. 1984; Persson et al. 1986; Azziz et al. 1988; Hakansson et al. 1995; Steingrimsdóttir et al. 1995).
Elaborating an SFH growth chart with follow-up of the pregnancy outcome offers the possibility of associating SFH growth patterns with different foetal growth outcomes, e.g. at birth being SGA, adequate for gestational age or large for gestational age. The growth charts resulting from our previous SFD study (Challis et al. 2002) gave different growth patterns depending on parity and BMI status, implying a potential to adjust for these two parameters.
The purpose of this study was to enhance the sensitivity of SFH measurements in predicting the SGA foetus by adjusting for parity and obesity. As BMI is laborious to calculate for each antenatal client, particularly in overcrowded clinics in low-income countries, we had the additional purpose to establish whether or not mid-upper-arm circumference (MUAC) is a practically useful proxy for body mass index (BMI) in performing this adjustment.
Subjects and methods
The SFH growth chart was created from measurements on 904 women with ultrasound-dated singleton pregnancies visiting two antenatal clinics (Malhangalene and Primeiro de Maio) in the suburban area of Maputo City during the period October 1993 to October 1994 (Challis et al. 2002). The areas covered by these clinics were considered representative of the vast majority of the pregnant city population (Bique Osman et al. 1999). The growth chart was drawn with the 5th, 50th and 95th percentiles per gestational week, by a cubic regression model (Challis et al. 2002).
In agreement with previous reports we knew that the SFH values of heavy women lie above those of normal and underweight women and that the mean SFH values of multiparous women lie above those of nulliparous women (Westin 1977; Munjanja et al. 1987; Hakansson et al. 1995; Steingrimsdóttir et al. 1995; Challis et al. 2002). We also knew that heavy women (BMI ≥ 27) delivered significantly heavier newborns, while thin women (BMI < 19) delivered significantly lighter newborns (Challis et al. 2002). In order to achieve a higher sensitivity of SFH measurements to predict SGA newborns in overweight and underweight pregnant women, SFH graphs were drawn for three groups of pregnant women with low, normal and high BMI, respectively. We also examined the relationship between BMI and MUAC with the intention to use MUAC as proxy for BMI, assuming that the tape measure for SFH measurement might be used to measure the MUAC instead of calculating the BMI (Liljestrand & Bergstrom 1991; Collins 1996).
The engagement of the foetal head is normally later in parous than in nulliparous women. This difference may be reflected in the position of the foetal body and the SFH. We therefore made different SFH graphs for multiparous and nulliparous women. Adjustment for multiparity was attempted in a way analogous to the way by which overweight was adjusted for. A similar mathematical model was sought.
The definition of a newborn being SGA used here implies birth weight <10th percentile for that particular gestational age. The standard was taken from our cohort study on the same population, and birth weight was measured within some hours to the nearest 50 g (Bique Osman et al. 1999). In the present study, at least one SFH value below the SFH graph for the 10th percentile was defined as an SFH value predicting an SGA newborn.
Women with at least two SFH measurements between 15 and 42 weeks (n = 817) contributed to the estimation of the growth chart percentiles. Each SFH measurement consisted of a data pair: day of pregnancy at measurement and SFH in cm. For each woman a straight line was drawn from the first measurement in time to the second, to the third and so on. With this method an estimate of the SFH was achieved for each day between the first and last measurement day. Then for each day in the gestational interval the SFH values for all women were sorted, from the lowest to the highest, and the 10th, 50th and 90th percentiles were calculated. Finally these three percentiles were fitted by smooth, continuous lines using a cubic regression model.
In the study population 770 women fulfilled the following requirements: (1) At least two SF-measurements recorded; (2) enough data to classify the newborn child as SGA or not SGA; (3) no missing values in the BMI, MUAC and parity measurements. Of these women, 71 (9.2%) gave birth to a neonate defined as SGA. By comparing each individual SF-chart with the SGA/not SGA classification, it was possible to estimate the sensitivity and specificity in the population.
To analyse the data we used the following software: epi-info version 6.02 (Centers for Disease Control, Atlanta, GA, USA), spss version 10.1 (Scandinavia AB, Stockholm, Sweden) and mathematica version 4.1 (© 2001 Wolfram Research, Inc., USA) software.
All women approached for enrolment were informed about the objectives of the study and that we followed the normal antenatal care programme, but that they were required to come every 2–3 weeks for SFH measurements (normally every 4 weeks) and to undergo an initial ultrasound examination. The ethical committees of the Faculty of Medicine of the Universidade Eduardo Mondlane and of the Maputo Central Hospital approved the project.
Of the study population (n = 904) 817 women enrolled had at least two SFH measurements and of these, 770 women qualified to participate in the analysis, because they had BMI and MUAC measurements. Mean BMI was 22.9 kg/m2 (n = 886) and mean MUAC was 26.2 cm (n = 872). There is a relatively strong correlation between BMI and MUAC, r = 0.621 (n = 865, P < 0.001). The sensitivity of at least one SFH measurement being below the 10th percentile in the SFH graph to predict birth weight below 10th percentile (SGA) was 49%, the specificity 66%, the sum of sensitivity and specificity 115%, the positive predictive value 13% and the negative predictive value 93%(Table 1).
|Outcome of screening||SGA verified||SGA not verified||Total|
|Positive for SGA||35||241||276|
|Negative for SGA||36||458||594|
With a reduction (adjustment) of 1 cm for heavy women (BMI > 29 kg/m2) and the same reduction of 1 cm for parous women the sensitivity was improved by 30%. In this manner the sensitivity was improved to 65% at the expense of a lower specificity, which decreased to 51% and with the sum of sensitivity and specificity 116%. Using as minimum criterion two values being below the 10th percentile reduced the sensitivity to 14% and increased the specificity to 91%.
We developed the following adjustment formula
where bi and mi, are the BMI and MUAC values, respectively, for the ith woman and pi is a simple mathematical function of parity for women i, defined as 0 if the parity is 0 and 1 if parity is ≥1. The constants 22.92 and 26.18 are the mean values for BMI and MUAC, respectively, in the study population.
Using a grid-search we tried to find the optimal combination of the parameters: α, β1, β2,β3. Here optimal is defined as the parameter combination which gives a high value when the sensitivity and specificity are added. By using the parameters: α = 0.02, β1 = 0.55, β2 = 0.82 and β3 = 2.15, the sensitivity became 0.70 and the specificity 0.56 and the sum of sensitivity and specificity 126%. Trying to keep the specificity high, the parameters: α = −1.40, β1 = −0.10, β2 = 0.85 and β3 = 1.9 gave an estimated sensitivity of 0.58 and a specificity of 0.68 (Table 2).
|Without adjustments||With optimal adjustment |
for maximal sensitivity
|With optimal adjustment |
for maximal specificity
|Sensitivity 0.493||Sensitivity 0.704||Sensitivity 0.577|
|Specificity 0.655||Specificity 0.558||Specificity 0.680|
|Total 1.148||Total 1.262||Total 1.257|
The main finding in this paper is that the use of simple and easily available anthropometrical data (like MUAC and BMI) and patient characteristics (like parity) may enhance the value of SFH growth chart. The observations indicate a possibility to further enhance the value of anthropometry in general rather than suggest a new practice of managing the SFH chart in resource-poor settings.
Appropriate technology is of vital importance in societies with scarce resources. In low-income countries ultrasound is largely unavailable. The SFH growth chart is the most reliable, simple, cost-effective and sustainable tool available for detection of multiple pregnancy and polyhydramnios. As a screening method to predict an SGA foetus it is not particularly sensitive, though often the only option (Cnattingius et al. 1984; Stuart et al. 1989). From our anthropometric data (Bique Osman et al. 1999) it appears that Mozambican women do not deviate significantly from other populations concerning weight, height, BMI and mid-upper-arm circumference (Kiserud 1986; Munjanja et al. 1987; Walraven et al. 1995). Other studies (Munjanja et al. 1987; Steingrimsdóttir et al. 1995) have shown that SFH may differ 2 cm in light and heavy women. We found a similar difference (Challis et al. 2002). In Zimbabwe, in accordance with these findings it has been recommended that SFH be reduced by 1 cm in pregnant women with body weight over 80 kg at booking.
Low sensitivity of existing anthropometrical methods for detecting the SGA/IUGR foetus is still a serious problem and therefore under continuous debate (Persson et al. 1986; Jahn et al. 1998). Efforts have been made to improve sensitivity by scoring risk factors for IUGR together with or without SFH measurements and such methods improve sensitivity (Wennergren et al. 1982; Cnattingius et al. 1984). Other authors conclude that the sensitivity is acceptable (73–86%) (Quaranta 1981; Belizán & Villar 1978) (Table 3).
|No. of subjects||428||139||138||6131||2919||817|
|Definition of SGA||<−1SD||<10%||<10%||<−2SD||<10%||<10%|
|Definition of low SF value||1 < −2SD||1 < 10%||2 consecutive |
or 3 isolated
|3 isolated |
<−2SD 3 static
|1 < −2SD||1 < 10%|
|Predictive value of positive screening (%)||Not calculated||79||60||13||18||14|
|Predictive value of negative screening (%)||Not calculated||93||88||99||92||93|
Our results correspond best with those of a previous Swedish study (Persson et al. 1986), showing that the specificity and positive predictive value are too low, below 50% and 20%, respectively. The main reasons for variation of results in specificity and positive predictive value are different definitions of SGA, different definitions of best cut-off SFH value to indicate SGA fetus and different study groups (sample size, sample selection, etc.) (Table 2).
Detecting the SGA foetus may enhance the possibilities of identifying the suspected IUGR suspect foetus by following the SFH growth or, if possible, the growth by ultrasound (Persson & Marsal 1978; Cnattingius et al. 1985; Marsal 1992; Marsal et al. 1996).
The best sensitivity of screening for SGA foetus by ultrasound we found in literature was 64% (Laurin 1987). The values of the SFH growth chart in our study for SGA screening in terms of sensitivity (49%), specificity (65%), positive (13%) and negative (93%) predictive value were low. In the current approach we studied ultrasound-dated pregnancies, which is essential for using the SFH growth chart as a tool for detecting SGA. In low-income countries there is normally no access to ultrasound; instead the last menstrual period (LMP) is used. In a forthcoming study (Challis et al. unpublished data) as well as in other studies it has been shown that LMP data frequently overestimate gestational length by about 1 week in comparison with ultrasound-dated gestational age (Grennert et al. 1978). In the absence of ultrasound, LMP may be used as a practical proxy when calculating gestational age.
The disadvantages of the SFH method are scarcity of standard SFH growth charts for a specific population in a specific region (Lindmark & Cnattingius 1992) and the inter- and intraobserver variations (Johnsen et al. 1988; Pattinson & Theron 1989; Jensen & Larsen 1991; Jacobsen 1992). In order to reduce these variations it is desirable that the same individual makes the measurements.
The definition of SGA is another dilemma. There are three different ones in the articles on the subject: <10%, <−1SD and <−2SD. It is difficult to detect the IUGR foetuses among the SGA foetuses. The vast majority of the IUGR foetuses are in the SGA group, about 30% of foetuses in the SGA group being IUGR foetuses (Cnattingius et al. 1987). Antenatal diagnosis in a third world setting is normally without ultrasound. There are only two possibilities: SFH measurements, scoring for risk factors for IUGR (previous obstetric history, maternal factors like low pregnancy weight, hypertension, etc.) or both together (Wennergren et al. 1982; Cnattingius et al. 1984).
In our study, enhancement of sensitivity was possible by adjustments but at the expense of reducing the specificity. The sum of the sensitivity and specificity is almost the same (116% and 115%, respectively). Mathematically, by using the linear function for optimal adjustment of the enhancement of sensitivity, the latter was improved to 70% with specificity of 59% and sum 129%. It may be argued that this formula had mere academic interest, but it shows that there is a possibility to enhance the sensitivity and specificity by adjustments. The importance of our findings does not lie in their immediate translation into antenatal care practice in resource-scare settings. Rather, they indicate a potentially useful new approach in obstetrical anthropometry, by which several available expressions of ‘tape measure obstetrics’ may be combined to enhance sensitivity of the only available appropriate and affordable screening tools.
As we already know the benefits of SFH chart screening for multiple pregnancy and polyhydramnios we excluded in order not to interfere with our study. However, it is important to express the obvious value of SFH chart screening for multiple pregnancy and polyhydramnios also in low-income countries.
In conclusion, in low-income countries ultrasound is a largely unavailable technology. As a screening method to predict SGA foetuses the sensitivity of the SFH chart is low but it might be improved by simple anthropometric and parity adjustments. We suggest that the SFH chart be assessed critically for its value in the diagnosis of IUGR. For the foreseeable future, simple and cheap clinical assessment will remain the main diagnostic tool in countries like Mozambique and the SFH growth chart belongs to that toolbox. There is a great need for further investigations within this field of simple anthropometric antenatal screening methods to be used particularly in low-income countries.
The study was made possible by grants from the Department of Research Co-operation with Developing countries (SAREC) at the Swedish International Development Authority (Sida) and from Mid Sweden Research and Development Centre (FoU). We acknowledge the valuable contribution to this field study by assistant medical officer and medical student Manuel Cotiro.