Second-trimester prenatal screening for Down syndrome andthe relationship of maternal serum biochemical markers topregnancy complications with adverse outcome

Kevin Spencer*

Endocrine Unit, Clinical Biochemistry Department, Harold Wood Hospital, Gubbins Lane, Romford, Essex, RM3 0BE, UK

In a group of 26 524 control pregnancies and a group of 3728 pregnancies affected by one or more of thepregnancy complications of low birthweight, intra-uterine growth restriction (IUGR), preterm delivery andstillbirth, I have compared the relative risk of occurrence of these complications in pregnancies which had araised maternal serum AFP (>2.0 MoM), raised maternal serum free b-hCG (>2.0 MoM), low AFP(<0.5 MoM), low free b-hCG (<0.5 MoM), combined raised AFP and free b-hCG (>2.0 MoM), and inthose with an increased Down syndrome risk (1 in 250 or greater). In the low birthweight group, only anincreased AFP and decreased free b-hCG showed signi®cance with relative risks of 1.6 and 2.1. In the IUGRgroup, also only an increased AFP and decreased free b-hCG showed signi®cance with relative risks of 1.6and 2.3. In the preterm delivery group, raised AFP, reduced free b-hCG, and combined elevated AFP andfree b-hCG showed signi®cance with relative risks of 3.8, 1.8 and 6.2. In the stillbirth group, raised AFP,reduced free b-hCG, and combined elevated AFP and free b-hCG showed signi®cance with relative risks of4.5, 2,4 and 7.2. An isolated raised free b-hCG or an increased Down syndrome risk were not associatedwith an increased relative risk for any of the pregnancy complications investigated. However, apart from thesix to seven-fold increased risk when both AFP and free b-hCG are above 2.0 MoM, suggesting increasedrisk of preterm delivery or impending fetal death, the clinical utility of such signi®cant differences isprobably poor. Copyright # 2000 John Wiley & Sons, Ltd.

KEY WORDS: IUGR; low birthweight; prematurity; stillbirth; free b-hCG; AFP

INTRODUCTION

Since the beginning of maternal serum screening forneural tube defects (NTD) in the 1970s, it has beenwell documented that elevated levels of maternalserum alpha-fetoprotein (AFP), in the absence ofNTD or ventral wall defect, are associated with anincreased risk of pregnancy complications, such as lowbirth weight, intra-uterine growth restriction (IUGR),preterm labour, and neonatal and fetal death (Bern-stein et al., 1992; Brock et al., 1979; Crandall et al.,1991; Simpson et al., 1994; Wald et al., 1977). In thelate 1980s and early 1990s maternal serum screeningwas extended in many places to include screening forDown syndrome and other chromosomal anomaliesby including the measurement of hCG (either free b ortotal). Since the early period of Down syndromescreening, a number of publications have investigatedthe association between high levels of hCG and anincreased risk of complications such as those seen withAFP (Pergament et al., 1995; Chapman et al., 1997;Dungan et al., 1994; Onderoglu and Kabukcu, 1997;Jauniaux et al., 1996; Palacio et al., 1999; Blundellet al., 1999; Mikic and Johnson, 1999; Fejgin et al.,1997; Benn et al., 1996; Lieppman et al., 1993;Morssink et al., 1995; Walton et al., 1999). However,there appears to be con¯icting evidence as to the

association with pregnancy complications and out-come, and questions have been raised regarding theclinical utility for predicting the risk of complicationsand the outcome of pregnancy. The reasons for thevaried reports may lie with the study design and therelatively small sample sizes used in the majority of thestudies. In this study, outcome data from prenatalscreening within a four-year period in a large centrewere examined in order to ascertain any associationbetween abnormal maternal serum biochemistry andadverse pregnancy outcome.

MATERIALS AND METHODS

The Harold Wood centre has provided second-trimester prenatal screening services to a number ofobstetric units using maternal serum AFP and free b-hCG measured at 14 to 18 weeks (by menstrual datesor ultrasound) since 1991, and the performance of thisprogramme with respect to screening for Downsyndrome has been described previously (Spencer,1999). Pregnancy outcome information was obtainedfrom a variety of sources, including delivery roomrecords, obstetric databases and child health data-bases. All data from prenatal screening and outcomeinformation were stored in a fetal database. Thedatabase was interrogated for pregnancies in theperiod between 1995 and 1998. Pregnancies resultingin termination or spontaneous loss prior to 24 weekswere excluded. Pregnancies resulting in multiple

*Correspondence to: K. Spencer, Endocrine Unit, Clinical Biochem-istry Department, Harold Wood Hospital, Gubbins Lane, Romford,Essex, RM3 0BE, UK. E-mail: Kevin_Spencer@Compuserve.com

PRENATAL DIAGNOSIS

Prenat Diagn 2000; 20: 652±656.

Copyright # 2000 John Wiley & Sons, Ltd. Received: 19 February 2000Revised: 19 April 2000Accepted: 5 May 2000

pregnancy, neural tube defect, ventral wall defect,chromosomal abnormality or congenital anomalyidenti®ed through ICD coding were also excluded. Intotal 30 818 singleton pregnancies were assessed and566 were excluded on the above criteria. The remain-ing 30 252 pregnancies were classi®ed according topregnancy outcomes. The classi®cations were; lowbirthweight, de®ned as births under 2500 g; pretermdelivery, de®ned as delivery of a live-born infantbefore 37 weeks of gestation; early preterm delivery,de®ned as delivery of a live-born infant before 35weeks of gestation; intra-uterine growth restriction,de®ned as birthweight below the 10th centile forgestational age as de®ned by Yudkin et al., (1987);extreme intra-uterine growth restriction, de®ned asbirthweight below the 3rd centile for gestational age;and stillbirth, de®ned as fetal death at or after 24weeks of gestation. Pregnancies exhibiting more thanone adverse outcome were assigned to each relevantadverse outcome group. Pregnancies not exhibitingany of the pregnancy complications were treated as thecontrol group.

The prenatal screening data were analysed for eachoutcome group and compared with those in thenormal pregnancy group. The analysis includedcalculation of the median MoM maternal serum freeb-hCG and AFP, calculation of the relative proportionof cases with free b-hCG greater than 2.00 MoM, withAFP greater than 2.00 MoM, with free b-hCG lessthan 0.50 MoM, with AFP less than 0.50 MoM, withboth AFP and free b-hCG greater than 2.00 MoM,and those with a Down syndrome risk of 1 in 250 orgreater.

All statistical analyses were performed with Ana-lyse-It (Smart Software, Leeds, UK) a statisticalsoftware add-in for Microsoft Excel 7 using weightcorrected MoMs as previously described (Spencer,1999). Median differences were tested by unpairedt-tests for unequal variance after log transformationof the MoM. Relative risks were tested for signi®canceby Chi square analysis.

RESULTS

In total, some 26 524 singleton pregnancies unaffectedby congenital anomalies were not complicated by thepregnancy complications under investigation. Table 1summarizes the control population and the population

of 3728 pregnancies affected by one or more of thepregnancy complications.

Some 1372 women gave birth to a low-birthweightbaby, this represented 4.5% (1372 of 30 252) of thesingleton pregnancies. In this group the median AFPMoM was higher than in the control population (1.04versus 1.00) and this was statistically signi®cant( p=0.040). The median free b-hCG MoM was slightlylower than in the control group (0.98 versus 1.00) andthis was not signi®cant ( p=0.419). Table 2 shows therelative risks associated with the various marker andrisk categories and their signi®cance.

In the group of women having a preterm delivery,some 1552 women delivered prior to 37 weeks and 664women delivered prior to 35 weeks. This represented5.1% and 2.2% of the singleton pregnancies. Themedian AFP MoM was higher in both the preterm(1.14 versus 1.00) and the early preterm deliveries (1.14versus 1.00). This difference was highly signi®cant( p<0.0001) in both cases. For free b-hCG, the medianMoM was slightly increased in both the preterm (1.03versus 1.00) and the early preterm deliveries (1.05versus 1.00). This difference was not signi®cant ineither case (p=0.296 and 0.707). Table 3 shows therelative risks associated with the various marker andrisk categories, and their signi®cance.

In the group of women with a baby exhibitingIUGR, some 2174 women delivered a baby weighingless than the 10th centile for its gestational age andsome 1122 women had babies weighing less than the3rd centile for its gestational age. This represented7.2% and 3.7% of the singleton pregnancies. Themedian AFP MoM was higher in both the IUGRgroup (1.03 versus 1.00) and the severe IUGR group(1.07 versus 1.00). This difference, however, was notsigni®cant for the IUGR group ( p=0.207) and almostreached signi®cance in the severe IUGR group( p=0.051). For free b-hCG, the median MoM wasslightly reduced in both the IUGR (0.99 versus 1.00)and the severe IUGR groups (0.98 versus 1.00). Thisdifference was not signi®cant for either group( p=0.499 and 0.659). Table 4 shows the relativerisks associated with the various marker and riskcategories, and their signi®cance.

In the group of women with a stillbirth, 192 caseswere identi®ed, this represented 0.6% of the singletonpregnancies. The median AFP MoM was slightlyhigher (1.05 versus 1.00). This difference was statisti-cally signi®cant (p=0.017). For free b-hCG, themedian MoM was slightly reduced (0.99 versus 1.00).

Table 1ÐDescription of the normal pregnancy group and those affected by adverse outcome. Data are means

Normal pregnancy n=26 524 Adverse outcome n=3728

Maternal age (years) 29.00 29.00Maternal weight (kg) 64.6 65.2Infant birthweight (g) 3480 2650Gestational age at delivery (weeks) 40 39Male : female sex ratio 1 : 0.96 1 : 1.10Maternal smoker (%) 16.3 19.1Gestational age at screening (days) 107 107

MATERNAL SERUM MARKERS AND PREGNANCY COMPLICATIONS 653

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 652±656.

This difference was not signi®cant ( p=0.105). Table 5shows the relative risks associated with the variousmarker and risk categories, and their signi®cance.

DISCUSSION

In studying the ability of second-trimester maternalserum biochemical markers of Down syndrome toidentify those likely to result in pregnancy complica-

tion and adverse perinatal outcome, investigators havelargely used two different approaches. The majority ofstudies (Pergament et al., 1995; Chapman et al., 1997;Dungan et al., 1994; Onderoglu and Kabukcu, 1997;Jauniaux et al., 1996; Palacio et al., 1999; Blundellet al., 1999; Mikic and Johnson, 1999; Fejgin et al.,1997; Benn et al., 1996; Lieppman et al., 1993; Liuet al., 1999) have utilized small study groups selectedon the basis of various cut-off criteria (either raised orlowered AFP, raised hCG or increased Down syn-

Table 2ÐLow birthweight: relative risk at various marker and risk levels for women delivering a low birthweight babycompared with the normal pregnancy control group (* shows statistical signi®cance)

Relative risk Statistical signi®cance

Free b-hCG>2.00 MoM 0.986 p=0.8845AFP>2.00 MoM 1.578 p=0.0064*

Free b-hCG<0.50 MoM 2.129 p=<0.0001*

AFP<0.50 MoM 0.876 p=0.3084AFP and free b-hCG>2.00 MoM 0.668 p=0.4210Down syndrome risk>1 in 251 0.888 p=0.3578

Table 3ÐPreterm delivery: relative risk at various marker and risk levels for women delivering preterm or early pretermcompared with the normal pregnancy control group (* shows statistical signi®cance)

Relative risk (probability)

Preterm Early preterm

Free b-hCG>2.00 MoM 1.149 ( p=0.0345)* 0.935 ( p=0.5706)AFP>2.00 MoM 3.488 ( p=<0.0001)* 3.750 ( p<0.0001)*

Free b-hCG<0.50 MoM 1.932 ( p=0.0001)* 1.795(p=<0.0001)*

AFP<0.50 MoM 0.657 ( p=0.0017)* 0.713 ( p=0.0934)AFP and free b-hCG>2.00 MoM 5.897 ( p=<0.0001)* 6.204 ( p=<0.0001)*

Down syndrome risk>1 in 251 1.171 ( p=0.1287) 1.095 ( p=0.6188)

Table 4ÐIUGR: relative risk at various marker and risk levels for women delivering a baby with IUGR or severe IUGRcompared with the normal pregnancy control group (* shows statistical signi®cance)

Relative risk (probability)

IUGR Severe IUGR

Free b-hCG>2.00 MoM 0.952 ( p=0.4338) 1.036 ( p=0.6895)AFP>2.00 MoM 1.195 ( p=0.2673) 1.640 ( p=0.0062)*

Free b-hCG<0.50 MoM 1.980 ( p=<0.0001)* 2.296 ( p=<0.0001)*

AFP<0.50 MoM 0.821 ( p=0.0581) 0.909 (p=0.5144)AFP and free b-hCG>2.00 MoM 0.632 ( p=0.3471) 0.407 ( p=0.2846)Down syndrome Risk>1 in 251 1.087 ( p=0.3733) 1.134 ( p=0.3208)

Table 5ÐStillbirth: relative risk at various marker and risk levels for women with a stillbirth at or after 24 weeks comparedwith the normal pregnancy control group (* shows statistical signi®cance)

Relative Risk Statistical signi®cance

Free b-hCG>2.00 MoM 1.327 p=0.1098AFP>2.00 MoM 4.511 p=<0.0001*

Free b-hCG<0.50 MoM 2.403 p=<0.0001*

AFP<0.50 MoM 0.949 p=0.9909AFP and free b-hCG>2.00 MoM 7.151 p=<0.0001*

Down syndrome risk>1 in 251 1.704 p=0.0291*

K. SPENCER654

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 652±656.

drome risk) and compared the incidence of differentpregnancy complications with an equally small groupof control pregnancies, which did not meet theinclusion criteria. Two studies (Morssink et al., 1995;Walton et al., 1999) have adopted the whole popula-tion approach by analysing outcomes from largescreening populations and looking at the relativeincidence of various pregnancy complications ingroups with biochemical results greater than aspeci®ed cut-off. There is little wonder therefore, thatdifferent studies have produced different and con¯ict-ing ®ndings with respect to the relative risk of eitherraised hCG or increased Down syndrome risk inidentifying pregnancies more likely to suffer fromcomplications or adverse perinatal outcome.

Pergament et al. (1995) in a study of 58 women witha false-positive Down syndrome risk, and 116 mater-nal age, race and gestational age matched controls,found that the high-risk group had a relative risk forfetal death, IUGR and preterm delivery of 5.2, 5.2 and2.4, all said to be statistically signi®cant. Chapmanet al. (1997) however, in a study of 246 cases withincreased Down syndrome risk compared with 889controls found relative risks of 0.31, 1.09 and 0.79.They concluded that women with a false-positiveDown syndrome risk were not at an increased riskfor adverse pregnancy outcomes. Similarly, Dunganet al. (1994) studied 99 women with increased Downsyndrome risk matched for age, race, gravidity andparity with 99 control patients. They also found nosigni®cant difference in the relative risk for fetal death(2.0) and IUGR (1.78), concluding that there was noevidence to support increased antepartum surveillancein patients with increased Down syndrome risk and anormal karyotype. Onderoglu and Kabukcu (1997), ina study of 81 women with an hCG level greater than2.0 MoM compared with 481 women with a levelbelow 2.0 MoM, showed an increased relative risk of5.34 for IUGR and 5.66 for preterm delivery. Theyconcluded that unexplained elevations of hCG in thesecond trimester appeared to be associated withadverse pregnancy outcome. Jauniaux et al. (1996)and Palacio et al. (1999) suggested that elevated levelsof free b-hCG were related to increased incidence ofplacental-related pregnancy disorders, IUGR and fetaldeath with relative risks of the order of 2.3 when thefree b-hCG was greater than 5 MoM. Similar resultswere observed by Blundell et al. (1999) for total hCGlevels greater than 4 MoM. Mikic and Johnson (1999)however, could not con®rm these ®ndings for IUGRor other pregnancy complications in a study of free b-hCG levels greater than 3.5 MoM in 192 cases and 192matched controls. Fejgin et al. (1997) however, foundthat in a group of 298 women with a total hCG greaterthan 2.5 MoM, the incidence of pregnancy complica-tions (other than pre-eclampsia) were similar to that ina group of 229 controls. Benn et al. (1996) found thatpatients with an elevated hCG, with or withoutelevated AFP, had an increased risk of poor pregnancyoutcome, especially preterm delivery and IUGR.Lieppman et al. (1993) also found elevated totalhCG to be associated with increased relative risks for

low birthweight (4.0), preterm delivery (2.8) andIUGR (1.8). In studying placental pathology in agroup of 24 pregnancies with unexplained elevatedmaternal serum hCG levels compared with 48 controlsfrom pregnancies with normal hCG levels, Liu et al.(1999) found a higher prevalence of cases with pre-eclampsia, IUGR and preterm delivery in the testgroup. They also found an increased placental size inthis group, and abnormalities related to implantationand chronic ultraplacental in¯ammation wereincreased in the group with elevated hCG levels. Inaddition, placentas from this group had an increasedamount of immunoreactive hCG per gram of tissuecompared with the control group.

In large-scale population studies, con¯icting resultshave also been found. In a study of some 8000 or morewomen, Morssink et al. (1995) found that bothisolated or combined elevations of AFP and totalhCG were indicators of increased relative risk forsevere IUGR (4.5, 2.1, 10.9) and concluded thatisolated or combined elevations of AFP and totalhCG in the second trimester of pregnancy was anindication for extra vigilance during further prenatalcare. However, recently Walton et al. (1999), in alarge-scale study of 25 000 women, could ®nd verylittle evidence for such claims and found relative risksof only 1.4 for fetal death, 1.1 for IUGR and 1.1 forpreterm delivery in a group of women having a totalhCG above 2.0 MoM. They concluded that hCG levelsare of little value for predicting the risk of complica-tions and pregnancy outcome.

This current large-scale population study con®rmsthat there are statistically signi®cant increases in therelative risk for low birthweight when AFP values areabove 2.0 MoM and free b-hCG levels are below0.5 MoM. Similarly, in preterm and early pretermdelivery, when increased AFP and low free b-hCGindicate an increased risk, however, of even greatersigni®cance is when both are elevated. With IUGRand severe IUGR, low free b-hCG is associated with atwo-fold increased risk, and in stillbirth, after 23weeks, raised AFP, low AFP and combined raisedAFP and free b-hCG are indicators of increased risk.However, apart from the six to seven-fold increasedrisk when both AFP and free b-hCG are above2.0 MoM, suggesting increased risk of preterm deliv-ery or impending fetal death, the clinical utility of suchsigni®cant differences is probably poor, as we (Onget al., in press) and others (Morssink et al., 1998) havealso found for free b-hCG and pregnancy-associatedplasma protein-A in the ®rst trimester. However, insuch cases it may be advisable to monitor thepregnancy more closely than would normally be thecase.

REFERENCES

Benn PA, Horne D, Briganti S, Rodis JF, Clive JM. 1996. Elevatedsecond-trimester maternal serum hCG alone or in combinationwith elevated alpha-fetoprotein. Obstet Gynecol 87: 217±222.

Bernstein IM, Bart RA, Miller R, Capeless EL. 1992. Elevatedmaternal serum alpha-fetoprotein: association with placental

MATERNAL SERUM MARKERS AND PREGNANCY COMPLICATIONS 655

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 652±656.

sonolucencies, fetomaternal hemorrhage, vaginal bleeding andpregnancy outcome in the absence of fetal anomalies. ObstetGynecol 79: 71±74.

Blundell G, Ashby JP, Martin C, Shearing CH, Langdale-Brown B,Keeling J, Ellis PM, Shade M, Chambers SE, Wenham PR. 1999.Clinical follow-up of high mid-trimester maternal serum intacthuman chorionic gonadotrophin concentrations in singletonpregnancies. Prenat Diagn 19: 219±223.

Brock DJH, Barron L, Duncan P, Scrimgeour JB, Watt M. 1979.Signi®cance of elevated mid-trimester maternal plasma alpha-fetoprotein values. Lancet i: 1281±1282.

Chapman SJ, Brum®eld CG, Wenstrom KD, DuBard MB. 1997.Pregnancy outcomes following false positive multiple markerscreening tests. Am J Perinatol 14: 475±478.

Crandall BF, Robinson L, Grau P, Scrimgeour JB, Watt M. 1991.Risks associated with an elevated maternal serum a-fetoproteinlevel. Am J Obstet Gynecol 165: 581±586.

Dungan JS, Shulman LP, Phillips OP, Simpson JL, Meyer NL,Grevengood C, Elias S. 1994. Positive serum screening for fetalDown syndrome does not predict adverse pregnancy outcome inabsence of fetal aneuploidy. J Soc Gynecol Invest 1: 55±58.

Fejgin MD, Kedar I, Amiel A, Ben-Tovim T, Chen R, Petel Y,Tepper R. 1997. Elevated hCG as an isolated ®nding during thesecond trimester biochemical screen: genetic, ultrasonic, andperinatal signi®cance. Prenat Diagn 17: 1027±1031.

Jauniaux E, Gulbis B, Tunkel S, Ramsay B, Campbell S, Meuris S.1996. Maternal serum testing for alpha-fetoprotein and humanchorionic gonadotropin in high-risk pregnancies. Prenat Diagn 16:1129±1135.

Lieppman RE Williams MA, Cheng EY, Resta R, Zingheim R,Hockok DE, Luthy DA. 1993. An association between elevatedlevels of human chorionic gonadotropin in the midtrimester andadverse outcome. Am J Obstet Gynecol 168: 1852±1856.

Liu DF, Dickerman LH, Redline RW. 1999. Pathological ®ndings inpregnancies with unexplained increases in midtrimester maternalserum human chorionic gonadotropin levels. Am J Clin Pathol111: 209±215.

Mikic TS, Johnson P. 1999. Second trimester maternal serum b-human chorionic gonadotrophin and pregnancy outcome. BrJ Obstet Gynaecol 106: 598±600.

Morssink LP, Kornman LH, Beekhuis JR, De Wolf BTHM,

Mantingh A. 1995. Abnormal levels of maternal serum humanchorionic gonadotropin and alpha-fetoprotein in the secondtrimester : relation to fetal weight and preterm delivery. PrenatDiagn 15: 1041±1046.

Morssink LP, Kornman LH, Hallahan TW, Kloosterman MD,Beekhuis JR, De Wolf BTHM, Mantingh A. 1998. Maternalserum levels of free b-hCG and PAPP-A in the ®rst trimester ofpregnancy are not associated with subsequent fetal growthretardation or preterm delivery. Prenat Diagn 18: 147±152.

Onderoglu LS, Kabukcu A. 1997. Elevated second trimester humanchorionic gonadotropin levels associated with adverse pregnancyoutcome. Int J Gynaecol Obstet 56: 245±249.

Ong CYT, Liao AWJ, Spencer K, Nicolaides KH. First trimestermaternal serum free b-hCG and PAPP-A as predictors ofpregnancy complications. Br J Obstet Gynaecol (in press).

Palacio M, Jauniaux E, Kingdom J, Dell E, Sheldrake A, RodeckCH. 1999. Perinatal outcome in pregnancies with a positive serumscreening for Down's syndrome due to elevated levels of free b-human chorionic gonadotropin. Ultrasound Obstet Gynaecol 13:58±62.

Pergament E, Stein AK, Fiddler M, Cho NH, Kupferminc MJ. 1995.Adverse pregnancy outcome after a false-positive screen for Downsyndrome using multiple markers. Obstet Gynecol 86: 255±258.

Simpson JL, Shulman L, Dungan J, Phillips O, Elias S. 1994. Mid-trimester biochemistry in prediction of third trimester complica-tions. In Screening for Down's Syndrome, Grudzinskas JG, ChardT, Chapman M, Cuckle H (eds). Cambridge University Press:Cambridge; 325±338.

Spencer K. 1999. Second trimester prenatal screening for Down'ssyndrome using alpha fetoprotein and free beta hCG: a seven yearreview. Br J Obstet Gynaecol 106: 1287±1293.

Wald N, Cuckle H, Stirrat GM, Bennett MJ, Turnbull AC. 1977.Maternal serum alpha-fetoprotein and low birth-weight. Lancet ii:268±270.

Walton DL, Norem CT, Schoen EJ, Ray GT, Colby CJ. 1999.Second-trimester serum chorionic gonadotropin concentrationsand complications and outcome of pregnancy. N Engl J Med 341:2033±2038.

Yudkin PL, Aboualfa M, Eyre JA, Redman CWG, Wilkinson AR.1987. New birthweight and head circumference centiles forgestational ages 24 to 42 weeks. Early Human Develop 15: 45±52.

K. SPENCER656

Copyright # 2000 John Wiley & Sons, Ltd. Prenat Diagn 2000; 20: 652±656.