7/28/2019 16_5_2010_0553_0557

1/5

553

Estimation o birth weight by measurement o undalheight and abdominal girth in parturients at termF. Mortazavi1 and A. Akaberi2

ABSTRACT In a prospective descriptive study, the useulness o symphysisundal height and the product oabdominal girth and undal height in predicting birth weight < 2500 g and > 4000 g were examined. Fundal heightand abdominal girth were measured at the time o admission on a sample o 795 parturient women at a teachinghospital in the Islamic Republic o Iran. Receiver operating characteristics curve analysis was used to select thebest cut-o points. The product o abdominal girth undal height with the cut-o at 3900 g perormed betteror predicting birth weight > 4000 g, but or low birth weight, the regression model o undal height with cut-oat 3000 g was a better predictor.

1Department o Midwiery; 2Department o Health, Sabzevar Faculty o Medical Sciences, Sabzevar, Islamic Republic o Iran (Correspondence to F.Mortazavi: rmortazavi@yahoo.com).

Received: 15/02/08; accepted: 28/05/08

:. 4000 2500

. 7953900 .

3000 4000

.

Estimation du poids de naissance par la mesure de la hauteur utrine et du tour de taille de parturientes terme

RSUM Une tude prospective descriptive a permis dvaluer lutilit de la mesure de la distance entre la symphysepubienne et le ond utrin, et du produit du tour de taille par la hauteur utrine pour lestimation des poids de naissanceinrieurs 2500 g et suprieurs 4000 g. La hauteur utrine et le tour de taille ont t mesurs sur un chantillon de795 parturientes lors de leur admission dans un hpital universitaire en Rpublique islamique dIran. Lanalyse de lacourbe ROC (pourReceiver Operating Characteristics) a t utilise pour slectionner les meilleures valeurs de seuil. Leproduit du tour de taille par la hauteur utrine, avec une valeur de seuil de 3900 g, a permis dobtenir les meilleuresestimations pour les poids de naissance suprieurs 4000 g. En revanche, pour les aibles poids de naissance, lemodle de rgression de la hauteur utrine, avec valeur de seuil de 3000 g, produisait de meilleures estimations.

7/28/2019 16_5_2010_0553_0557

2/5

EMHJ Vol. 16 No. 5 2010 Eastern Mediterranean Health JournalLa Revue de Sant de la Mditerrane orientale

554

Introduction

Precise estimation of birth weight (BW)is one of the most important measuresat the beginning of labour. is is espe-

cially important in developing countrieswhere many births occur at home or atbirth centres without adequate facili-ties. In these circumstances diagnosis ofmacrosomic and light fetuses can resultin timely referral of diagnosed cases towell-equipped hospitals.

There are 2 common methodsof estimation of BW: sonographicevaluation and clinical palpation [1].In developing countries, ultrasonog-

raphy may be unavailable or may notbe aordable by patients. Physic ianestimates of BW by palpation are asreliable as, or superior to, those madefrom ultrasonographic measurementsof the fetus [2]. However, their ac-curacy depends on experience, whichmay be lacking in many obstetric carepersonnel in developing countries [1].Tat is why measurement of fundalheight (FH) using inexpensive and eas-

ily available non-elastic tapes has beenrecommended as a means of assessingBW in low-resource countries.

In some studies BW < 2500 g and >4000 g have been proposed as the cut-o points for predicting BW using FHmeasurement only [16]. Since the sizeof the fetus aects the abdominal girth(AG), a cut-o point for AG as a predic-tor of BW < 2500 g has been calculated[7]. Still other studies have developed

formulas based on the regression of BWon both FH and AG for predicting BW[8,9]. Dare et al. and Bothner et al. usedthe product of symphysisFH and AGat the level of the umbilicus to estimateBW at term in utero, and their estimatescorrelated well with BW [10,11]. ese2 studies did not consider BW < 2500g and > 4000 g. Shiu et al. comparedthe product of symphysisFH and AGwith sonographic est imation of BW

and found that the product formulaperformed as well as sonographic esti-mation, except in BW < 2500 g [12].

e aim of this study was to furtherexamine and compare the performanceof the product formula of FH AGwith that of the formula based on FHalone, to clarify whether taking an extra

measurement, i.e. AG, improves theprediction of BW. Since studies haveconcluded that FH [1,1214], theproduct formula [11,14] or formulasbased on both FH and AG [8] are notpowerful predictors of BW < 2500 gor > 4000 g [1,8,1114], we aimed totest the hypothesis that changing thecut-o points of the 2 formulas wouldbe improved in the case of BW < 2500g or > 4000 g.

Methods

A prospective descript ive study wasundertaken from 1 May to 1 August2003 on 795 consecutive parturientwomen hospitalized at Mobini teach-ing and maternity hospital aliated toSabzevar Faculty of Medical Sciences,Islamic Republic of Iran, which is the

only maternity hospital in this regionproviding standard maternity servicesto urban and rural women. e purposeof the original study was to estimategestational age by measurement of FHand AG in parturient women at term[13]. For this paper we reanalysed thedata collected to estimate BW by theproduct formula.

Sample

Based on receiver operating charac-teristics (ROC) curve power analysis,the sample sizes required for achieving80% power to detect a dierence of0.25 between the area under the ROCcurve of 0.5, using a signicance levelof 0.05, was 11 for the positive group(< 2500 g) and 500 from the negativegroup ( 2500 g). Also the sample sizesrequired for achieving 80% power todetect a dierence of 0.3 between the

area under the ROC curve of 0.5, using asignicance level of 0.05, was 7 from thepositive group (> 4000 g) and 500 from

the negative group ( 4000 g) [15,16].e dierences of 0.25 and 0.30 wereobtained by analysing the ROC curves,and the specied sample sizes were themaximum required for analysing the

power of each formula. e size of thesample in the present study in eachBW group was larger than the speciedmaximum required sample size.

e inclusion criteria were: alive,single and term fetuses with longitu-dinal lie. e exclusion criteria were:documented severe fetal congenitalanomalies; preterm labour; presenceof a thick deposited layer of fat at thelower abdomen; maternal weight > 91

kg; and clinical or ultrasonic evidenceof uterine broids, oligohydramnios orpolyhydramnios.

Data collection

e data were obtained by interviewand by means of clinical assessment.e background characteristics ofwomen were obtained by interv iew.Vaginal examination was carried outto determine the fetal station. As soon

as a woman meeting the above criteriawas admied for vaginal or abdominaldelivery, symphysisFH and AG at thelevel of umbilicus were measured.

FH was measured using a non-elastictape from the highest point on the uter-ine fundus to the midpoint of the up-per border of the symphysis pubis. ethumb was used to hold the tape whileaempting to reach the upper border ofthe symphysis pubis. Measurement was

made 3 times using the tape reverse-sideup to avoid any bias. e mean of the 3readings was then obtained to the near-est centimetre. e same precautionsfor preventing bias were made in thecase of AG. Abdominal measurementswere taken in the supine position withlile exion of legs, aer emptying thebladder and during uterine relaxationperiods. All the measurements weretaken by 2 members of the research

team who had been trained for the task.Informed wrien consent was obtainedfrom all the women.

7/28/2019 16_5_2010_0553_0557

3/5

555

To test the intra-rater reliability oftrained observers, we asked them totake FH and AG measurements in 30parturients. e mean dierence forAG and FH measurements between

2 trained observers was 1.35 mm and1.52 mm respectively. Also 98.0% of AGmeasurement dierences and 97.2%of FH measurement dierences werebetween 10 mm compar ing the 2observers. To further examine the intra-rater reliability we used BlandAltmanscaer plots. e dierences betweenthe trained observers measurementswere ploed against the mean of theobservers measurements. Intra-rater

scaer plots demonstrated that 96%of all AG and 92% of all FH measure -ment dierences were within 2 standarddeviations (SD) of acceptable levels ofagreement.

e actual BW of the baby wasmeasured in grams to the nearest 50 gby the midwife on duty within an hourof delivery using a weighing scale. emidwives who weighed the babies aerbirth were blind to the intrapartum esti-mates of BW.

Birth weight estimates

e 1st formula for estimating BW wasthe method of Dare et al. as the productof symphysisFH and AG at the level

of the umbilicus measured in cm [17]:BW = FH abdominal girth.

e 2nd formula for estimating BWwas obtained by regression of BW onFH using sample data from the original

study [13]: BW = (FH 87) + 515.

Data analysis

To predict BW < 2500 g and > 4000 g,the ROC curve was used to select cut-opoints with the best sensitivity and specif-icity. Levels of signicance in this studywere P< 0.05. e data were analysedusing Pearson correlation coecient, t-

test, covariance analysis and ROC curve,usingSPSS soware, version 15.

Results

e study group consisted of 795 women,442 (55.6%) of whom were primiparousand 353 multiparous. e mean andstandard deviation (SD) age and weightof the women were 25.0 (SD 5.28) yearsand 69.6 (SD 10.8) kg respectively.

e mean FH was 34.6 (SD 3.1)cm, range 2447 cm, and for AG was99.2 (SD 8.7) cm, range 75124 cm.e mean product of FH AG was3440 (SD 540) cm, range 21845640cm. All the mean values were higher in

multiparous women than in primipa-rous women (P< 0.001) (Table 1).

In 173 (21.8%) of the cases en-gagement had already occurred at thetime of admission. Covariance analy-

sis, adjusted for age and weight of thewoman, revealed signicant dierencesin the mean value of FH between the173 women presenting with the fetuspart-engaged and the 622 with the fetusunengaged [33.5 (SD 3.1) cm versus34.8 (SD 3.0) (P< 0.001)].

e mean actual BW of the infantswas 3212 (SD 421) g, range 16004450g. ere were 27 (3.4%) newborns withBW > 4000 g and 27 (3.4%) with BW

< 2500 g.All the correlation coecients be-

tween the measured parameters for boththe primiparous and multiparous groupwere signicantly dierent from zero (P< 0.001). e correlation coecientsbetween maternal indicators and BWwere higher in the primiparous than inmultiparous women. e correlationbetween FH and BW was stronger thanthe correlation between the product ofFH AG and BW (0.58 versus 0.56)(Table 2).

e best cut-o points of BW fordetecting BW < 2500 g and > 4000 gwere determined for each formula us-ing ROC curves. e cut-o point for

Table 1 Comparison o actual birth weight and pre-parturition birth weight estimated by 2 dierent ormulas by parity

Parameter Primiparous women(n = 442)

Multiparous women(n = 353)

All women(n = 795)

Mean (SD) Mean (SD) Mean (SD)

Mothers

Weight (kg) 68.0 (10.0) 71.6 (10.0) 69.6 (10.0)

Fundal height (cm) 34.2 (3.1) 35.0 (3.1) 34.6 (3.1)

Abdominal girth (cm) 97.3 (8.4) 101.6 (8.3) 99.2 (8.7)

Infants

Estimated birth weight (by undalheight abdominal girth) (g) 3331 (522) 3575 (533) 3440 (540)

Estimated birth weight (byregression model o birth weighton undal heighta) (g) 3487 (269) 3563 (271) 3520 (272)

Actual birth weight (g) 3153 (421) 3288 (410) 3213 (421)aEstimated birth weight = (undal height 87) + 515.

SD = standard deviation.

7/28/2019 16_5_2010_0553_0557

4/5

EMHJ Vol. 16 No. 5 2010 Eastern Mediterranean Health JournalLa Revue de Sant de la Mditerrane orientale

556

predicting BW > 4000 g was 3900 gbased on the FH AG formula and3450 g based on the regression modelof BW on FH. e sensitivity and spe-cicity of the obtained cut-o point forthe FH AG formula for the detectionof BW > 4000 g were 81.3% (95% CI:

80.2%82.4%) and 82.2% (95% CI:81.2%83.2%) respectively. e cor-responding values for the regressionmodel of BW on FH were 75.0% (95%CI: 73.7%76.3%) and 85.4% (95% CI:84.5%86.3%) (Table 3).

To predict BW < 2500 g, a singlecut-o point equal to 3000 g was ob-tained for both formulas. Sensitivity andspecicity of the obtained cut-o pointfor the FH AG formula were 70.4%

(95% CI: 68.9%71.8%) and 79.9%(95% CI: 78.8%81.0%) respectivelyand for the regression model of BW onFH formula, the corresponding values

were 77.8% (95% CI: 76.6%79.0%)and 85.5% (95% CI: 84.6%86.4%)(Table 4). e cut-o points were se-lected on the basis that we considereda satisfactory level of sensitivity to be atleast 70% and also sought a relativelyhigh level of specicity.

Discussion

In this study, 2 estimators of birth weight,that is a regression model of FH and theproduct of FH AG were evaluatedand compared, with particular emphasison BW < 2500 g and > 4000 g. Ourresults showed that the mean valuesfor weight of parturient, BW, FH and

AG were higher in multiparous thanin primiparous parturients. is is dueto the fact that multiparous womengenerally tended to be faer; this leads

to higher BW which in turn results inhigher values of FH and AG. In otherstudies this correlation coecient wasas follows: 0.56 [11], 0.91 [1], 0.59 [3],0.74 [4],0.87 [6], 0.74 [9] and 0.72 [5].e correlation between the product ofFH AG and BW was 0.56 in our study.

In other studies it was 0.74 and 0.57respectively [10,14].

As stated above, by using the ROCcurve we were able to nd cut-o pointsof BW for each formula in which theycould predict BW < 2500 g and > 4000g with maximum accuracy. e resultsindicate that using the product of FH AG the cut-o point 3900 g was abeer estimator of high BW (> 4000 g)than the regression model of FH at the

cut-o point 3450 g, and the regressionmodel of FH at cut-o point 3000 gwas a stronger predictor of low BW ( 4000 g

Formula Sensitivity Specifcity Positive predictive value Negative predictive value

% (CI) % (CI) % (CI) % (CI)

Fundal height abdominalgirtha 81.3 (80.282.4) 82.2 (81.283.2) 99.5 (99.599.5) 86.0 (81.091.0)

Regression model o birthweight on undal heightb 75.0 (73.776.3) 85.4 (84.586.3) 99.4 (99.499.4) 95.0 (89.0100.0)

aCut-o point = 3900 g; bCut-o point = 3450 g.CI = confdence interval.

Table 4 Distribution o the power o 2 ormulas in predicting birth weight < 2500 g

Formula Sensitivity Specifcity Positive predictivevalue

Negative predictivevalue

% (CI) % (CI) % (CI) % (CI)

Fundal height abdominal girth 70.4 (68.971.8) 79.9 (78.881.0) 98.7 (98.698.8) 11.0 (10.311.7)

Regression model o birthweight on undal height 77.8 (76.679.0) 85.5 (84.686.4) 99.1 (99.099.2) 18.9 (17.820.0)

Cut-o point = 3000 g.CI = confdence interval.

7/28/2019 16_5_2010_0553_0557

5/5

557

Onah He, Ikeme AC, Nkwo PO. Correlation between intrapar-1.tum undal height and birth weight.Arican journal o reproduc-tive health, 2002, 6(2):239.

Coningham FG et al. Fetal growth disorders. In:2. Williamss ob-stetrics. Section 7. Common complications o pregnancy, 21st ed.New York, McGrawHill, 2001:759.

Labrecque M, Boulianne M. Estimation du poids de nais--3.sance par la mesure de la hauteur uterine chez la parturienteen Republique Federale Islamique des Comores [Estimationof birth weight by measuring the height of the uterus in theparturient women in the Islamic Federal Republic of ComoroIslands]. Revue dpidmiologie et de sant publique, 1987,35:37885.

Mohanty C, Das BK, Mishra OP. Parturient undal height as4. a predictor o low birth weight. Journal o tropical pediatrics,1998, 44:2224.

Aghababaii S, Nahidi F. The use o tricolor measuring tape as a5.predictor o birth weight. Journal o medical sciences Pakistan,2005, 5(4):30710.

Ghaemmaghami F et al. Parturient undal height and birth weight6.estimation.Archives o Iranian medicine, 2002, 5(2):803.

Mohanty C, Das BK, Mishra OP. Parturient abdominal circum-7.erence as a predictor o low birth weight. Journal o tropicalpediatrics, 2000 46(6):3634.

Woo JS et al. Estimation o etal weight8. in utero rom symphysisundal height and abdominal girth measurements. Australianand New Zealand journal o obstetrics & gynaecology, 1985,25(4):26871.

Hamudu NA. Shafq M. Manji KP. Parturient symphysioundal9.height and abdominal girth measurements to predict birth

Reerences

weight at Muhimbili Medical Centre, Dar es Salaam, Tanzania.Tanzania medical journal, 2004, 19(1):1821.

Dare FO et al. The value o symphysio-undal height/abdomi-10.nal girth measurements in predicting etal weight. Internationaljournal o gynecology and obstetrics, 1990, 31:2438.

Bothner BK, Gulmezoglu AM, Homeyr GJ. Symphysis undus11.height measurements during labour: a prospective, descriptivestudy.Arican journal o reproductive health, 2000, 4(1):4855.

Shittu AS, Kuti O, Oriji EO. Comparison o clinical and ultra-12.sonographic estimation o etal weight. International journal ogynecology and obstetrics, 2003, 90(2):1401.

Mortazavi F, Rakhshani MH. Estimation o gestational age by13.measurement o undal height at the beginning o labor. Asrar

Scientifc Journal o Sabzevar Faculty o Medical Sciences, 2003,10(4):3944.

Kraiem J et al. Estimation clinique du poids oetal: interet dans14.la prediction de la macrosomie [Clinical etal weight estima-tion and prediction o macrosomia]. La Tunisie mdicale. 2004,82(3):2715.

Hanley JA, McNeill BJ. A method o comparing the areas under15.receiver operating characteristic curves derived rom the samecases. Radiology, 1983, 148:83948.

Obuchowski N, McClish D. Sample size determination or16.diagnostic accuracy studies involving binominal ROC curveindices.Statistics in medicine, 1997, 16:152942.

Shittu AS. Clinical versus sonographic estimation o oetal17.weight in southwest Nigeria. Journal o health, population andnutrition, 2007, 25(1):1423.

Berry M et al. Foetal growth parametersclinical versus ultra-18.sonographic. Indian journal o pediatrics, 1992, 59(1):91101.

A number of studies have found astrong correlation between BW and FH,BW and AG and BW and the productformula [1,314,18]. However, most ofthem concluded that these indicators

were not strong enough for the pur-pose of predicting BW < 2500 g or >4000 g [1,3,5,7,8,14,17]. Shiu et al., forexample, concluded that although theproduct of FH AG was as accurateas routine ultrasonographic estimation,it did not perform as satisfactorily incases of low BW [12]. Berry et al. hasconcluded that neither clinical nor ul-trasonographic parameters were satis-factory in identifying low-birth-weight

fetuses [18

]. Woo et al. reported that aformula based on the regression of BWon FH and AG [BW = 1.515 + (0.092 FH) + (0.016 AG)] was a powerfulpredictor of BW between 25003500 gbut was not accurate enough in predict-ing BW < 2500 g and > 3500 g [8].ey concluded that all the generatedequations obtained from their sample

similarly underestimated the BW in thelarger babies and overestimated it in thesmaller babies. Onah et al. also found astrong correlation (0.91) between FHand BW, but concluded that the regres-

sion model of FH was more useful inpredicting BW between 25003999 g[1]. Kraiems study, covering 400 casesof macrosomia, showed that the regres-sion model of FH was not strong enoughin predicting BW > 4000 g [12].

As can be seen, our corre lat ionswere lower than, or at best equal to,the above mentioned studies. Teseweaker correlations and the fact thatthe cited studies, despite their reported

higher correlations, produced unsatis-

factory results in the case of predictingBW < 2500 g and > 4000 g would haveled us to expect equally unsatisfactoryor worse results in such cases too. Infact, applying the usual cut-o points(i.e. 2500 g and 4000 g), we also ob-tained weak results which were in linewith such expectations. However, we

found that the shortcomings aectingthese indicators could be overcome byusing dierent cut-o points and em-ploying the formula which performsbeer in predicting BW < 2500 g and

> 4000 g.Our results indicate that in the caseof BW > 4000 g, the FH AG formulaand a cut-o point of 3900 g performedbeer in predicting BW, and in the caseof BW < 2500 g, the regression of BWon FH formula and a cut-o point of3000 g produced beer predictions.erefore we conclude that by selectingappropriate cut-o points specic toeach community and employing the

appropriate formula, it will be possibleto utilize the 2 formulae for predictingBW < 2500 g and > 4000 g.

Acknowledgements

We acknowledge funding from Sab-zevar Faculty of Medical Sciences.