Original Research ajog.

org

1 OBSTETRICS 56
2 57
3 Risk of operative delivery for intrapartum fetal compromise 58
4
5
in small-for-gestational-age fetuses at term: an internally 59
60
6 validated prediction model 61
7 62
Q4 Erkan Kalafat, MD; Jose Morales-Rosello, MD; Basky Thilaganathan, MD, PhD, FRCOG; Fathema Tahera, BSc;
8 63
Asma Khalil, MD, MRCOG
9 64
10 65
11 66
12 BACKGROUND: Small-for-gestational-age fetuses are at an increased delivery beyond 39 weeks’ gestation, and the cerebroplacental ratio 67
risk of intrapartum fetal compromise requiring operative delivery. Factors multiples of median. The combined model (area under the curve, 0.76; 68
13
associated with the risk of intrapartum fetal compromise are yet to be 95% confidence interval, 0.72e0.80), using both the antenatal and
14 69
established, and a comprehensive model accounting for both the antenatal intrapartum risk factors, included the gestational age at delivery beyond 39
15 70
and intrapartum variables is lacking. weeks’ gestation (odds ratio, 1.62; 95% confidence interval, 1.14e2.56),
16 71
OBJECTIVE: We aimed to develop and validate a predictive model for the cerebroplacental ratio multiples of median (odds ratio, 0.38; 95%
17 the risk of operative delivery for presumed intrapartum fetal compromise in confidence interval, 0.18e0.79), parity (odds ratio 0.35; 95% confidence
72
18 fetuses suspected to be small for gestational age at term. interval, 0.22e0.54), induction of labor (odds ratio 1.63; 95% confidence 73
19 STUDY DESIGN: This was a single-center cohort study of small-for- interval, 1.11e2.40), augmentation using oxytocin (odds ratio, 1.84; 95% 74
20 gestational-age fetuses, defined as estimated fetal weight below the confidence interval, 1.23e2.73) and the use of epidural analgesia (odds 75
21 10th centile in singleton pregnancies at term. The variables included ratio, 2.80; 95% confidence interval, 1.94e4.04). The results indicate 76
22 known risk factors for operative delivery because of fetal compromise: that the model has good discrimination and, according to the Hosmer- 77
23 maternal characteristics, estimated fetal weight, abdominal circumfer- Lemeshow test, has good fit (P ¼ .591). 78
24 ence, Doppler parameters, gestational age at delivery, induction of labor, CONCLUSION: The prediction model demonstrates 6 important risk 79
25 and intrapartum risk factors (presence of meconium, augmentation of factors that are associated with the risk of operative delivery for fetal 80
26 labor using oxytocin, the use of epidural analgesia, intrapartum pyrexia, compromise in small-for-gestational-age fetuses at term. The model 81
27 and hemorrhage). The receiver-operating characteristics curve analysis shows good discrimination and fit and has the potential to be used for 82
28 was used to investigate the predictive accuracy. Internal validation of the clinical decision making and to counsel women about their individual 83
29 models was performed with bootstrapped data sets. intrapartum risk. 84
30 RESULTS: A total of 927 term pregnancies with 18.7% operative de- 85
31 liveries were included. The antenatal model (area under the curve, 0.69; Key words: cerebroplacental ratio, Doppler, emergency cesarean de- 86
32 95% confidence interval, 0.65e0.73) using only the antenatal risk factors livery, fetal distress, fetal growth restriction, forceps, operative delivery, 87
33 included parity, abdominal circumference centile, gestational age at small for gestational age 88
34 89
35 90
36
37
38
M anagement of fetuses presumed
to be small for gestational age
(SGA) at term continues to be the
Previous studies have identified
certain antenatal risk factors, in
particular SGA and low cerebroplacental
such a prediction model could help
in the identification of the pregnancies
at highest risk of intrapartum fetal
91
92
93
39 focus of much recent debate and ratio (CPR), which increase the risk of compromise, enabling appropriate man- 94
40 research. One of the main challenges operative delivery.3-6 Although these agement of at-risk fetuses as well as 95
41 arises from the fact that the estimated studies are informative about possible facilitating appropriate individualized 96
42 fetal weight (EFW) alone is not a good antenatal markers of fetal compromise, antenatal counseling. 97
43 proxy for the diagnosis of fetal growth the intrapartum events are likely to be The main aim of this study was to 98
44 restriction and is a poor predictor of equally, or even more, important in develop an integrated prediction model 99
45 intrapartum fetal compromise in SGA determining the mode of delivery and combining antenatal and intrapartum 100
46 pregnancies.1,2 neonatal outcome.5-7 Furthermore, it is parameters for the evaluation of their 101
47 possible that the combination of both predictive accuracy for the risk of 102
48 antenatal and intrapartum risk factors operative delivery for presumed intra- 103
49 Cite this article as: Kalafat E, Morales-Rosello J, may improve on their individual pre- partum fetal compromise. In some 104
50 Thilaganathan B, et al. Risk of operative delivery for dictive accuracy. circumstances, the decision regarding 105
51 intrapartum fetal compromise in small-for-gestational- Despite the need for tools to aid the timing and mode of delivery 106
age fetuses at term: an internally validated prediction clinical decision making, a comprehen- might have to rely on antenatal
52 model. Am J Obstet Gynecol 2017;xxx:xx-xx.
107
53 sive prediction model for assessing the parameters only. Therefore, we also 108
54 0002-9378/$36.00 risk of operative delivery for intrapartum aimed to develop an integrated pre- 109
ª 2017 Elsevier Inc. All rights reserved.
55 https://doi.org/10.1016/j.ajog.2017.10.022 fetal compromise in SGA fetuses is diction model combining antenatal 110
yet to be established. If validated, parameters only.

MONTH 2017 American Journal of Obstetrics & Gynecology 1.e1

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:38 pm
ce
 Original Research OBSTETRICS ajog.org

111 167
112 Materials and Methods the MCA PI and the UA PI.14,15 All mode of delivery and the variables. After 168
113 This was a retrospective cohort study in a Doppler indices were converted into adding all the significant parameters into 169
114 single tertiary referral center over a 16 multiples of median (MoM), correcting the model, the least contributory pa- 170
115 year period from 1999 through 2015. for GA using reference ranges, and rameters were removed one by one until 171
116 The ViewPoint database (ViewPoint birthweight values were converted into the model had both good precision and 172
117 5.6.8.428; ViewPoint Bildverarbeitung centiles.16-18 calibration. Multivariable models were 173
118 GmbH, Weßling, Germany) was used to Intrapartum data included whether built with a variable selection approach. 174
119 identify cases evaluated at the Fetal the labor was induced or spontaneous, After adding all variables with signif- 175
120 Medicine Unit (St George’s Hospital, presence or absence of meconium- icant associations in the univariate 176
121 London, United Kingdom). stained liquor (grade 2 or 3), CTG model, the final predictors were deter- 177
122 The inclusion criteria were singleton abnormalities (classified according to mined with a backwards elimination in 178
123 pregnancies diagnosed with an SGA National Institute for Health and Care the logistic regression model using 179
124 fetus, defined as EFW below the 10th Excellence guidelines), ST analysis ab- Akiake’s Information Criterion. Only 180
125 centile for gestational age at 36 weeks or normalities, use of oxytocin for slow the parameters that could be obtained 181
126 beyond. Pregnancies complicated by progress of labor, intrapartum pyrexia, prior to delivery were considered eligible 182
127 major structural fetal abnormalities, intrapartum hemorrhage, use of for inclusion in the multivariable model. 183
128 aneuploidy, or genetic syndromes were epidural analgesia for labor, and mode of The Hosmer-Lemeshow test was used to 184
129 excluded from the analysis. In addition, delivery.19 Data on the maternal baseline test the goodness of fit of models. Q2
185
130 pregnancies that had an elective cesarean characteristics and the pregnancy out- The validation of the model was per- 186
131 delivery and those that had operative comes were collected from the hospital formed using 10,000 bootstrap replicates 187
132 delivery (cesarean delivery or instru- obstetric records. of the original cohort. Separate data sets 188
133 mental delivery) because of failure to The main outcome in this study was of the same sizes were constructed using 189
134 progress in labor or any cause other than the operative delivery for presumed fetal a bootstrapping technique. The variables 190
135 presumed fetal compromise were also compromise. Operative delivery for were chosen at random with equal 191
136 excluded from the analysis. presumed fetal compromise included sampling probability and with replace- 192
137 Gestational age (GA) was calculated both cesarean delivery and instrumental ment. The predictive value of the final 193
138 from the crown-rump length measure- delivery. The diagnosis of fetal compro- model was assessed with the area under 194
139 ment at 11e13 weeks and only one (the mise was based on CTG abnormalities, the receiver-operator curve (AUC). 195
140 last) examination per pregnancy was STanalysis abnormalities, abnormal fetal Multiple bootstrapped data sets were 196
141 included in the analysis.8 For pregnan- scalp blood sample pH, or a combina- analyzed to prevent the overestimation 197
142 cies in which the first ultrasound tion of these.20 The study was exempt of the regression coefficients and to 198
143 was performed in the second trimester from review by Wandsworth Research prevent overfitting. After determining 199
144 (>14 weeks’ gestation), the pregnancy Ethics Committee. Some of the preg- the discrimination power and goodness 200
145 was dated according to the head nancies reported in this study were of fit of the final model receiver- 201
146 circumference.9 included in a previous study.5 operating characteristic curves and 202
147 Routine fetal biometry was performed Hosmer test, the probabilities of some 203
148 according to a standard protocol and the Statistical analysis clinical examples were calculated to 204
149 EFW was calculated using the Hadlock Continuous variables were presented as provide a better relationship to the 205
150 formula.10 The umbilical artery (UA) median with interquartile range, while practical uses of the model. The statisti- 206
151 and middle cerebral artery (MCA) categorical variables were presented as a cal analysis was performed using the 207
152 Doppler waveforms were recorded using fraction of the total with percentages. RStudio statistical software (version 208
153 color Doppler, and the pulsatility index Distribution assumptions were tested 1.0.136; RStudio, Inc, Boston, MA). 209
154 (PI) was calculated according to a stan- with a Shapiro-Wilk test and QQ plots. 210
155 dard protocol.11,12 In brief, the MCA PI Group comparison of variables was Results 211
156 values were obtained in the space at made with a Student t test, Mann- We identified 1061 singleton pregnan- 212
157 which the artery passes by the sphenoid Whitney U test, or c2 test where appro- cies that were eligible for inclusion in the 213
158 wing close to the Circle of Willis, and UA priate. Missing variables were imputed study. In total, 927 women were included 214
159 PI values were obtained in free loops of with a linear regression model using the in the analysis after excluding major fetal 215
160 umbilical cord. missing parameter as the outcome vari- abnormalities, aneuploidy, genetic syn- 216
161 The Doppler measurements were able and the missing parameter’s most dromes, missing delivery records, elec- 217
162 performed within 4 weeks of delivery. likely clinical covariate as the regressor. tive cesarean delivery, and operative 218
163 The measurements were obtained in the Parameters in the models were deter- delivery for causes other than fetal 219
164 absence of fetal movement and keeping mined by a forward selection, backward compromise (n ¼ 134). 220
165 the insonation angle with the examined elimination approach. First, a uni- The proportion of missing variables in 221
166 vessels less than 30 .13 The CPR was variable analysis was used to determine the data set was less than 1%. The inci- 222
calculated as the simple ratio between the significant associations between the dence of operative delivery for presumed

1.e2 American Journal of Obstetrics & Gynecology MONTH 2017

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:38 pm
ce
 ajog.org OBSTETRICS Original Research
223 279
224 TABLE 1 280
225 Characteristics of the study cohort according to the need for operative delivery for presumed fetal compromise 281
226 282
227 Operative delivery for presumed No operative delivery for presumed 283
Pregnancy variables fetal compromise (n ¼ 173) fetal compromise (n ¼ 754) P value
228 284
229 Antenatal variables 285
230 Maternal age, y, median (IQR) 29.00 (25.50e33.00) 29.00 (25.00e34.00) .992 286
231 Body mass index, kg/m2, median (IQR) 23.20 (21.00e25.00) 22.30 (20.40e25.43) .901 287
232 288
Multiparous, n, % 30 (17.3) 329 (43.6) < .001
233 289
234 Ethnicity .934 290
235 White, n, % 69 (39.9) 275 (36.5) 291
236 African, n, % 38 (22.0) 177 (23.5) 292
237 293
Asian, n, % 63 (36.4) 270 (35.8)
238 294
239 Mixed, n, % 2 (1.1) 24 (3.2) 295
240 Other, n, % 1 (0.6) 8 (1.0) 296
241 Smoker, n, % 17 (9.8) 105 (13.9) .170 297
242 298
Alcohol use, n, % 10 (5.7) 35 (4.6) .555
243 299
244 Drug use, n (%) 4 (2.3) 18 (2.3) .999 300
245 Ultrasound and Doppler variables 301
246 Gestational age at ultrasound, wks, median (IQR) 37.71 (36.50e39.43) 37.43 (36.43e38.57) .013
302
247 303
Interval between scan and delivery, d, median (IQR) 10.00 (4.00e19.75) 10.00 (4.00e19.00) .774
248 304
249 Abdominal circumference, mm, median (IQR) 296.9 (284.3e308.0) 296.9 (287.5e304.7) .848 305
250 Abdominal circumference centile, median (IQR) 4.68 (2.07e8.40) 5.49 (2.53e10.54) .027 306
251 Estimated fetal weight, g, median (IQR) 2426 (2232e2712) 2393 (2216e2616) .110 307
252 308
Estimated fetal weight centile, median (IQR) 5.51 (2.67e8.44) 6.13 (2.95e9.89) .160
253 309
254 Umbilical artery pulsatility index, median (IQR) 0.96 (0.82e1.10) 0.94 (0.83e1.07) .517 310
255 Umbilical artery pulsatility index MoM, median (IQR) 1.09 (0.95e1.26) 1.07 (0.95e1.22) .269 311
256 Middle cerebral artery pulsatility index, median (IQR) 1.40 (1.20e1.61) 1.48 (1.27e1.71) .009 312
257 313
Middle cerebral artery pulsatility index MoM, 1.17 (1.03e1.36) 1.20 (1.03e1.37) .779
258 314
median (IQR)
259 315
260 Cerebroplacental ratio, median (IQR) 1.49 (1.20e1.88) 1.59 (1.30e1.90) .017 316
261 Cerebroplacental ratio MoM, median (IQR) 0.82 (0.64e0.99) 0.87 (0.71e1.04) .003 317
262 Intrapartum variables 318
263 319
Induction of labor, n, % 113 (65.3) 340 (45.0) < .001
264 320
265 Meconium-stained liquor (grade 2 or 3), n, % 12 (6.9) 36 (4.9) .256 321
266 Oxytocin use for slow progress in labor, n, % 68 (39.3) 139 (18.4) < .001 322
267 Intrapartum hemorrhage, n, % 3 (1.7) 1 (0.1) .022 323
268 324
Intrapartum pyrexia, n, % 9 (5.2) 6 (0.8) < .001
269 325
270 Epidural use, n (%) 96 (55.4) 184 (24.4) < .001 326
271 Variables at birth 327
272 Gestational age at delivery, median (IQR) 40.0 (38.43e41.14) 39.43 (38.29e40.29) .358 328
273 329
Birthweight, g, median (IQR) 2600 (2315e2920) 2614 (2344e2891) .746
274 330
275 Birthweight centile, median (IQR) 4.87 (1.43e11.13) 3.59 (2.54e9.43) .825 331
276 Small for gestational age, n, % 162 (93.6) 598 (79.3) < .001 332
277 IQR, interquartile range; MoM, multiples of median. 333
278 Kalafat et al. Predictive model of operative delivery in SGA. Am J Obstet Gynecol 2017. 334

MONTH 2017 American Journal of Obstetrics & Gynecology 1.e3

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:38 pm
ce
 Original Research OBSTETRICS ajog.org

335 391
336 AC centile (P ¼ .017), and CPR MoM 392
TABLE 2
337 (P ¼ .002) were all significantly associ- 393
Results of the univariable logistic regression analysis of factors associated
338 ated with the risk of operative delivery 394
with the need for operative delivery for presumed fetal compromise
for fetal compromise (Table 2). ½T2
339 395
340 Odds 95% confidence The intrapartum variables that were 396
Risk factors ratio interval P value significantly associated with the risk of
341 397
342 Maternal age, y 1.00 0.98e1.03 .589 operative delivery for fetal compromise 398
343 Body mass index, kg/m2 0.86 0.57e1.27 .471
included induction of labor (P < .001), 399
344 oxytocin use for augmentation of labor 400
Multiparous 0.27 0.17e0.40 < .001 (P < .001), epidural use (P < .001),
345 401
346
Ethnicity 0.99 0.82e1.20 .954 intrapartum hemorrhage (P ¼.025), and 402
347 Smoking 0.67 0.37e1.12 .153 intrapartum pyrexia (P <.001) (Table 2). 403
348 Operative delivery for fetal compromise 404
Drug abuse 0.96 0.27e2.63 .953
349 was significantly associated with the 405
Alcohol use 1.26 0.58e2.50 .531 presence of an SGA neonate (odds
350 406
351 Abdominal circumference centile 0.96 0.93e0.99 .017 ratio [OR], 1.88; 95% CI, 1.18e3.15, 407
352 Estimated fetal weight centiles 0.97 0.96e1.01 .161 P ¼.010). However, because this variable 408
353 cannot be obtained prior to delivery, it 409
Umbilical artery pulsatility index MoM 1.42 0.70e2.80 .311
354 was not included in the multivariable 410
Middle cerebral artery pulsatility index MoM 1.00 0.51e1.95 .977 logistic regression model.
355 411
356 Cerebroplacental ratio MoM 0.35 0.18e0.68 .002 The multivariable model was built in 2 412
357 Gestational age at delivery, wks 1.47 1.04e2.12 .030
steps. The first model (Table 3) included ½T3 413
358 only the antenatal variables, principally 414
True small for gestational age 1.88 1.18e3.15 .010 the AC centile (OR, 0.96; P ¼ .027),
359 (birthweight <10th centile) 415
360 CPR MoM (OR, 0.33; P ¼ .002), GA 416
Ultrasound scan-delivery interval, wks 0.95 0.84e1.07 .439 at delivery beyond 39 weeks’ gestation
361 417
362 Intrapartum factors (OR, 1.88; P< .001), and multiparity 418
363 Induction of labor 2.29 1.63e3.25 < .001 (OR, 0.27; P < .001). The addition 419
364 of intrapartum variables significantly 420
Epidural analgesia 3.86 2.74e5.45 < .001
365 improved the predictive accuracy of 421
Intrapartum pyrexia 6.84 2.43e20.65 < .001 the antenatal parameters for the risk
366 422
367 Intrapartum hemorrhage 13.28 1.68e269.53 .025 of operative delivery (De Long’s test, 423
368 Oxytocin used for augmentation of labor 2.86 2.00e4.08 < .001 P < .001, Table 3). 424
369 Induction of labor without augmen- 425
Meconium grade 2/3 1.48 0.77e2.84 .250
370 tation (OR, 2.26, P < .001), oxytocin 426
MoM, multiples of median.
371 augmentation of labor without induc- 427
Kalafat et al. Predictive model of operative delivery in SGA. Am J Obstet Gynecol 2017.
372 tion (OR, 3.09; P < .001), and epidural 428
373 use (OR, 2.73; P < .001) were the intra- 429
374 partum variables significantly associated 430
375 fetal distress in the study cohort was presumed fetal compromise (Table 1) with the risk of operative delivery for 431
376 18.7% (173 of 927); cesarean delivery had significantly lower abdominal fetal compromise (Table 3). 432
377 was 9.5% (88 of 927) and instrumental circumference (AC) centile (P ¼ .027), The parameters that were significantly 433
378 delivery 9.2% (85 of 927). The positive CPR MoM (P ¼.003), significantly more associated with the risk of operative de- 434
379 predictive value of an antenatal SGA epidural analgesia (P < .001), intra- livery in the final model were the CPR 435
380 diagnosis for confirmed SGA at birth was partum pyrexia (P < .001), induction of MoM, the GA at delivery, parity, induc- 436
381 83.5% (95% confidence interval [CI], labor (P < .001), augmentation of labor tion of labor, oxytocin use for augmen- 437
382 77.1e87.7%). SGA using birthweight (P <.001), and intrapartum hemorrhage tation of labor, and epidural use 438
383 less than 10th centile was more prevalent (P ¼ .022). There were no significant (Table 3). The Hosmer-Lemeshow test 439
384 in the operative delivery for fetal differences between the 2 study groups showed that the model has a good fit 440
385 compromise group compared with those regarding birthweight (P ¼ .746) or GA (P ¼ .591). 441
386 that did not require operative delivery at delivery (P ¼ .358) (Table 1). The validation of the model was per- 442
387½T1 (93.6% vs 79.3%, P < .001) (Table 1). When the antenatal variables were formed using 10,000 bootstrap replicates 443
388 Compared with those that did not analyzed, the univariable model showed of the original cohort. Multiple data sets 444
389 require operative delivery, pregnancies that parity (P < .001), GA at delivery were constructed using bootstrapping 445
390 that required operative delivery for beyond 39 weeks’ gestation (P ¼ .030), with replacement. The AUC curves 446

1.e4 American Journal of Obstetrics & Gynecology MONTH 2017

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:39 pm
ce
 ajog.org OBSTETRICS Original Research
447 503
448 accuracy of ultrasound EFW at term in 504
TABLE 3
449 determining which fetus is at risk. 505
Results of the multivariable logistic regression analysis of the risk factors
450 A number of studies have reported on 506
associated with the need for operative delivery for presumed fetal
451 the association between low CPR values 507
compromise
452 and adverse events during labor or 508
453 Odds 95% confidence with adverse neonatal outcomes.3-6 509
Risk factor ratio interval P value Furthermore, the CPR could have po-
454 510
455 Antenatal risk factors only tential value in assessing the risk not only 511
456 Multiparous 0.27 0.17e0.41 < .001 in SGA fetuses but also in appropriate- 512
457 for-gestational-age fetuses.5,6 It is true, 513
Abdominal circumference centile 0.96 0.93e0.99 .027
458 however, that the CPR on its own is 514
Cerebroplacental ratio MoM 0.33 0.16e0.66 .002 unlikely to be a strong predictor of the
459 515
460 Gestational age at delivery, wks 1.97 1.36e2.90 < .001 risk of operative delivery for fetal 516
461 Combined model (both antenatal and compromise and that the addition of 517
462 intrapartum risk factors) other variables, in particular the intra- 518
partum risk factors, would be essential to
463 Multiparous 0.36 0.23e0.57 < .001 519
464 improve the overall predictive accuracy. 520
Gestational age at delivery, wks 1.65 1.12e2.46 .011 In 2012, a predictive model for labor
465 521
466
Cerebroplacental ratio MoM 0.35 0.13e0.55 .005 outcomes including cesarean delivery 522
467 Induction of labor without augmentation 2.26 1.49e3.76 < .001 and instrumental delivery because of 523
presumed fetal distress was published.7
468 Epidural analgesia 2.73 1.75e3.78 < .001 524
469 In this model, nulliparity (OR, 2.08), 525
Oxytocin used for augmentation of labor 3.02 1.76e5.18 < .001 induction of labor (OR, 1.35), augmen-
470 526
Oxytocin used for augmentation of labor 3.09 1.46e5.59 < .001 tation of labor (OR, 1.26), intrapartum
471 without labor induction 527
472 pyrexia (OR, 1.47), and epidural use 528
MoM, multiples of median. (OR, 2.46) were significantly associated
473 Kalafat et al. Predictive model of operative delivery in SGA. Am J Obstet Gynecol 2017.
529
474 with cesarean delivery for fetal compro- 530
475 mise.7 This model combining antenatal 531
476 and intrapartum variables had an AUC 532
477 of 0.73. 533
478 showed that the antenatal model had an and that it has potential clinical Our model suggests a similar distri- 534
479 AUC of 0.69 (95% CI, 0.65e0.73), applications. bution of risk with slightly higher ORs 535
480 sensitivity of 23.7%, specificity of 90.0%, The CPR remains an independent for individual parameters and an AUC of 536
481 positive likelihood ratio (LR) of 2.37, predictor of the risk of operative delivery 0.76. This difference can probably be 537
482 and negative LR of 0.84. The combined for fetal compromise in SGA fetuses, explained by the high-risk nature of our 538
483 model had an AUC of 0.76 (95% CI even after adjusting for both the other study population. Furthermore, the 539
484 0.72e0.80), sensitivity of 70.5%, speci- antenatal and intrapartum risk factors. study by Schuit et al7 did not include 540
485 ficity of 70.0%, positive LR of 3.95, and Interestingly, the EFW centile was not fetal Doppler assessment. 541
486½F1 negative LR of 0.42 (Figure). The com- significantly associated with the risk of 542
487 bined model had significantly higher operative delivery for fetal compromise. Clinical and research implications 543
488 predictive accuracy than the antenatal Importantly, the positive predictive Most studies reporting on the value of 544
489 model (P < .001). To provide a practical value of the antenatal diagnosis of SGA CPR used cutoffs for categorizing 545
490 interpretation of the model, several hy- was 83%. patients and assessing pregnancy out- 546
491 pothetical clinical scenarios were tested comes; the most commonly used cutoff 547
492½T4 (Table 4). Interpretation of the findings and values were based on CPR centiles 548
493 comparison with existing literature according to gestational age.3,21,22 In 549
494 Comment The results of this study are in accor- contrast to previous studies, this model 550
495 Main findings dance with the published literature on investigated the value of CPR as a 551
496 The findings of this study demonstrate the role of CPR in identifying the fetuses continuous variable, and the results 552
497 that the combination of the antenatal at risk at term. Interestingly, the associ- indicate that the CPR values have a linear 553
498 and the intrapartum risk factors can ation of the intrapartum risk factors with association with the risk of operative 554
499 identify the majority of SGA fetuses that the probability of operative delivery for delivery. 555
500 develop intrapartum compromise fetal compromise in a cohort of preg- This finding is important because, 556
501 requiring operative delivery. The inter- nancies with SGA fetuses has been according to our model, an early-term 557
502 nal validation of our model suggests that poorly investigated so far. Moreover, the fetus suspected to be SGA with a CPR 558
the calibration of the model is very good results reiterate the limited predictive MoM of 1.6 and no additional risk

MONTH 2017 American Journal of Obstetrics & Gynecology 1.e5

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:39 pm
ce
 Original Research OBSTETRICS ajog.org

559 615
fetal compromise. Furthermore, the
560 FIGURE 616
561Q3 CPR values were not calculated before 617
Receiver-operating characteristic of antenatal and combined models
562 the analysis for this study, so the health 618
563 care professionals providing the intra- 619
564 partum care were effectively blinded to 620
565 this value. 621
566 The detection rate of true SGA fetuses 622
567 was similar to previously reported diag- 623
568 nostic parameters for ultrasound scans.1 624
569 These factors are likely to indicate that 625
570 our study population reflects the real-life 626
571 clinical setting. Importantly, clinicians 627
572 were not blinded to the results of the 628
573 ultrasound and Doppler assessment, 629
574 giving rise to the possibility of subse- 630
575 quent clinical intervention and a treat- 631
576 ment effect. However, during the study 632
577 period, intervention in the form of in- 633
578 duction of labor was undertaken only for 634
579 EFW less than the fifth centile or UA PI 635
580 above the 95th centile, as per our local 636
581 protocol. Hence, the relationships 637
582 between fetal Doppler findings, EFW, 638
583 and labor outcomes should be relatively 639
584 uninfluenced by these interventions. 640
585 Moreover, the study cohort will have 641
586 been scanned by a large number of 642
587 different operators, raising the risk of 643
588 interobserver variability in the 644
589 measurements. 645
590 The threshold for the diagnosis of fetal 646
591 Receiver-operating characteristic curve of the antenatal model (dotted line) and the combined compromise is also likely to have been 647
592 (including both the antenatal and intrapartum variables) model (straight line). influenced by changing personnel and 648
593 Kalafat et al. Predictive model of operative delivery in SGA. Am J Obstet Gynecol 2017. attitudes toward intrapartum manage- 649
594 ment over the 16 year period. We have 650
595 also used a robust methodology to 651
596
factors has a predicted probability of of operative delivery is so high that it is ensure goodness of fit and discrimina- 652
operative delivery of 3.6%, whereas a unacceptable to the patient, she might tion power of the results while avoiding
597 653
similar fetus with a CPR PI MoM of 0.8 want to opt for a planned cesarean de- overestimation and overfitting.
598 654
has the predicted probability of 7.9%, livery instead of induction of labor). This Depending on the baseline rate of oper-
599 655
suggesting a doubling in the individual risk assessment could also be used to ative delivery in the population, the
600 656
risk (Table 4). Our results suggest that reassure mothers who want to achieve model could under- or overperform.
601 657
the linear changes in CPR, below or spontaneous vaginal delivery. Finally, cases that underwent opera-
602 658
above the previously suggested thresh- tive delivery were indeed candidates for
603 659
604
olds, still influence the individual risk Study strengths and intrapartum insults, but because an 660
605
and should be taken into consideration. limitations intervention was applied before the 661
The purpose of our prediction model The strengths of our study include using process can take its course, our model
606 662
is to provide a decision-making aid to a large cohort of pregnancies with ante- can never truly predict the actual risk.
607 663
help physicians with the clinical man- natally diagnosed SGA fetuses, the short Our study was not adequately powered
608 664
agement of these pregnancies. The main interval between ultrasound and de- to investigate clinically important, but
609 665
significance of our study would be the livery, ascertainment of the outcome rare, adverse pregnancy outcomes, such
610 666
ability to provide an individualized risk data and adjusting for possible con- as hypoxic ischemic encephalopathy.
611 667
for operative delivery. Our model should founding variables, the standardized However, operative delivery for pre-
612 668
help physicians to decide whether to intrapartum care in a single institution, sumed fetal compromise is known to be
613 669
induce/augment labor (eg, if the chance and applying strict criteria for defining associated with a number of clinically
614 670

1.e6 American Journal of Obstetrics & Gynecology MONTH 2017

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:39 pm
ce
 ajog.org OBSTETRICS Original Research
671 727
672 TABLE 4 728
673 Clinical scenarios and their predicted probabilities according to the final multivariable prediction model combining the 729
674 intrapartum and antepartum variables 730
675 731
676 Late term Oxytocin 732
(above 39 weeks’ CPR Induction used for Epidural Predicted
677 Patient gestation) MoM Multiparity of labor augmentation use probability (95% CI) Description 733
678 734
1 No 1.6 No No No No 3.6% (1.0e6.5%) Nulliparous, normal CPR, no
679 intrapartum risk, early term
735
680 736
681 2 No 0.8 No No No No 7.9% (5.4e10.3%) Nulliparous, low CPR, no 737
intrapartum risk, early term
682 738
683 3 No 0.5 No No No No 10.5% (7.9e13.0) Nulliparous, abnormal CPR, no 739
intrapartum risk, early term
684 740
685 4 Yes 0.5 No No No No 16.2% (13.7e18.8) Nulliparous, abnormal CPR, no 741
686 intrapartum risk, late term 742
687 5 No 0.5 No Yes Yes No 26.1% (23.6e28.7) Nulliparous, abnormal CPR, 743
688 moderate intrapartum risk, 744
689 early term 745
690 6 No 0.5 No Yes Yes Yes 49.1% (46.6e51.7) Nulliparous, abnormal CPR, high 746
691 intrapartum risk, early term 747
692 7 Yes 0.5 No Yes Yes Yes 61.6% (59.0e64.1) Nulliparous, abnormal CPR, high 748
693 intrapartum risk, late term 749
694 8 No 0.5 Yes Yes Yes Yes 26.1% (23.4e28.8) Multiparous, abnormal CPR, 750
695 high intrapartum risk, early term 751
696 9 No 1.2 Yes Yes Yes Yes 14.4% (11.8e17.3) Multiparous, normal CPR, high 752
697 intrapartum risk, early term 753
698 CI, confidence interval; CPR, cerebroplacental ratio; MoM, multiples of median. 754
699 Kalafat et al. Predictive model of operative delivery in SGA. Am J Obstet Gynecol 2017. 755
700 756
701 757
702 758
703 important adverse neonatal out- 2. Gardosi J, Francis A. Adverse pregnancy 8. Robinson HP, Fleming JE. A critical evaluation 759
704 comes.23-25 In fact, the mode of delivery outcome and association with small for gesta- of sonar “crown-rump length” measurements. 760
705 tional age birthweight by customized and Br J Obstet Gynaecol 1975;82:702-10. 761
has an independent impact on neonatal population-based percentiles. Am J Obstet 9. Hadlock FP, Deter RL, Harrist RB, Park SK.
706 condition at birth.26 Gynecol 2009;201:28.e1-8. Estimating fetal age: computer-assisted analysis
762
707 3. Prior T, Paramasivam G, Bennett P, Kumar S. of multiple fetal growth parameters. Radiology 763
708 Conclusion Are fetuses that fail to achieve their growth po- 1984;152:497-501. 764
709 In summary, our prediction model tential at increased risk of intrapartum compro- 10. Hadlock FP, Harrist RB, Sharman RS, 765
710 mise? Ultrasound Obstet Gynecol 2015;46: Deter RL, Park SK. Estimation of fetal weight 766
demonstrates 6 important risk factors 460-4. with the use of head, body, and femur
711 that are associated with the risk of 4. Prior T, Mullins E, Bennett P, Kumar S. Pre- measurementsea prospective study. Am J
767
712 operative delivery for fetal compromise diction of intrapartum fetal compromise using Obstet Gynecol 1985;151:333-7. 768
713 in SGA fetuses. The combination of the cerebroumbilical ratio: a prospective obser- 11. Acharya G, Wilsgaard T, Berntsen GK, 769
714 these variables can identify the majority vational study. Am J Obstet Gynecol 2013;208: Maltau JM, Kiserud T. Reference ranges for 770
715 124.e1-6. serial measurements of umbilical artery Doppler 771
of fetuses at risk of intrapartum fetal indices in the second half of pregnancy. Am J
5. Khalil A, Morales-Rosello J, Khan N, et al. Is
716 compromise requiring operative de- Obstet Gynecol 2005;192:937-44.
772
cerebroplacental ratio a marker of impaired fetal
717 livery. Further prospective studies are growth velocity and adverse pregnancy 12. Bahlmann F, Reinhard I, Krummenauer F, 773
718 required for external validation of this outcome? Am J Obstet Gynecol 2017;216:606. Neubert S, Macchiella D, Wellek S. Blood flow 774
719 model. n e1-10. velocity waveforms of the fetal middle cerebral 775
720 6. Khalil A, Morales-Rosello J, Morlando M, et al. artery in a normal population: reference values 776
Is fetal cerebroplacental ratio an independent from 18 weeks to 42 weeks of gestation.
721 References 777
predictor of intrapartum fetal compromise and J Perinat Med 2002;30:490-501.
722 neonatal unit admission? Am J Obstet Gynecol 13. Clinical Standards Committee. ISUOG
778
1. Sovio U, White IR, Dacey A, Pasupathy D,
723 Smith GC. Screening for fetal growth restriction 2015;213:54.e1-10. Practice Guidelines: use of Doppler ultrasonog- 779
724 with universal third trimester ultrasonography in 7. Schuit E, Kwee A, Westerhuis ME, et al. raphy in obstetrics. Ultrasound Obstet Gynecol 780
725 nulliparous women in the Pregnancy Outcome A clinical prediction model to assess the risk 2013;41:233-9. 781
726 Prediction (POP) study: a prospective cohort of operative delivery. BJOG 2012;119: 14. Arbeille P, Roncin A, Berson M, Patat F, 782
study. Lancet 2015;386:2089-97. 915-23. Pourcelot L. Exploration of the fetal cerebral

MONTH 2017 American Journal of Obstetrics & Gynecology 1.e7

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:39 pm
ce
 Original Research OBSTETRICS ajog.org

783 839
blood flow by duplex Dopplerelinear array sys- 20. Amer-Wahlin I, Arulkumaran S, Hagberg H, 25. Ekéus C, Högberg U, Norman M. Vacuum
784 tem in normal and pathological pregnancies. Marsál K, Visser GH. Fetal electrocardiogram: assisted birth and risk for cerebral complications in 840
785 Ultrasound Med Biol 1987;13:329-37. ST waveform analysis in intrapartum surveil- term newborn infants: a population-based cohort 841
786 15. Bahado-Singh RO, Kovanci E, Jeffres A, lance. BJOG 2007;114:1191-3. study. BMC Pregnancy Childbirth 2014;20:14. 36. 842
787 et al. The Doppler cerebroplacental ratio and 21. Cruz-Martínez R, Figueras F, Hernandez- 26. Prior T, Kumar S. Mode of delivery has an 843
perinatal outcome in intrauterine growth restric- Andrade E, Oros D, Gratacos E. Fetal independent impact on neonatal condition at
788 844
tion. Am J Obstet Gynecol 1999;180:750-6. brain Doppler to predict cesarean delivery birth. Eur J Obstet Gynecol Reprod Biol
789 16. Baschat AA, Gembruch U. The cere- for non-reassuring fetal status in term small-for- 2014;181:135-9. 845
790 broplacental Doppler ratio revisited. Ultrasound gestational-age fetuses. Obstet Gynecol 846
791 Obstet Gynecol 2003;21:124-7. 2011;117:618-26. 847
792 17. Morales-Roselló J, Khalil A, Morlando M, 22. Nassr AA, Abdelmagied AM, Shazly SA. Author and article information 848
Papageorghiou A, Bhide A, Thilaganathan B. Fetal cerebro-placental ratio and adverse From the Fetal Medicine Unit, St George’s Hospital, St Q1
793 849
Changes in fetal Doppler indices as a marker of perinatal outcome: systematic review and George’s University of London, Cranmer Terrace, London,
794 failure to reach growth potential at term. Ultra- meta-analysis of the association and diag- United Kingdom (Drs Kalafat, Morales-Rosello, 850
795 sound Obstet Gynecol 2014;43:303-10. nostic performance. J Perinat Med 2016;44: Thilaganathan, and Tahera and Ms Khalil); Department of 851
796 18. Poon LC, Tan MY, Yerlikaya G, Syngelaki A, 249-56. Obstetrics and Gynecology, Ankara University Faculty of 852
797 Nicolaides KH. Birth weight in live births and 23. Towner D, Castro MA, Eby-Wilkens E, Medicine, Ankara, Turkey (Dr Kalafat); and Department 853
stillbirths. Ultrasound Obstet Gynecol 2016;48: Gilbert WM. Effect of mode of delivery in nullip- of Statistics, Middle East Technical University, Ankara,
798 854
602-6. arous women on neonatal intracranial injury. Turkey (Dr Kalafat).
799 19. National Institute for Health and Clinical N Engl J Med 1999;341:1709-14. Received May 19, 2017; revised Oct. 17, 2017; 855
800 Excellence. Intrapartum care. National Institute 24. Werner EF, Janevic TM, Illuzzi J, Funai EF, accepted Oct. 19, 2017. 856
801 for Health and Care Excellence clinical guideline Savitz DA, Lipkind HS. Mode of delivery in The authors report no conflict of interest. 857
802 55. London: National Institute for Health and nulliparous women and neonatal intracranial Corresponding author: Asma Khalil, MD, MRCOG. 858
Clinical Excellence; 2007. injury. Obstet Gynecol 2011;118:1239-46. akhalil@sgul.ac.uk or asmakhalil79@googlemail.com
803 859
804 860
805 861
806 862
807 863
808 864
809 865
810 866
811 867
812 868
813 869
814 870
815 871
816 872
817 873
818 874
819 875
820 876
821 877
822 878
823 879
824 880
825 881
826 882
827 883
828 884
829 885
830 886
831 887
832 888
833 889
834 890
835 891
836 892
837 893
838 894

1.e8 American Journal of Obstetrics & Gynecology MONTH 2017

FLA 5.5.0 DTD
YMOB11880_proof
8 November 2017
12:39 pm
ce