Baseline Patient Characteristics
The participant selection flowchart for our case-control study is illustrated in Figure 1. Of 998 patients meeting inclusion criteria, 491 were excluded due to factors such as multiple pregnancies, chromosomal abnormalities, and major malformations. Ultimately, a cohort of 507 participants was included in this study, with 354 (70%) experiencing SGA complications enrolled as the derivation cohort and 153 (30%) included in the validation set. A total of 134 (26.43%) women in the cohort underwent cesarean delivery following labor induction.
As depicted in Table 1,a comparison was made between the maternal, neonatal, and obstetric characteristics of women who achieved vaginal delivery and those who required cesarean section. Women who underwent cesarean section demonstrated higher maternal age (29.68 ± 4.58 vs. 28.19 ± 4.22, P < 0.001), greater admission weight (62.81 ± 8.39 vs. 60.65 ± 9.57, P = 0.015), and elevated admission BMI (24.84 ± 2.93 vs. 24.00 ± 3.66, P = 0.008) in comparison to vaginal delivery patients. Cesarean deliveries were also associated with lower Bishop scores following cervical ripening (6.17 ± 1.41 vs. 7.05 ± 1.60, P < 0.001), An increased incidence of oligohydramnios (16.42% vs. 9.92%, P = 0.044), lower rates of early rupture of membranes during induction (5.97% vs. 12.87%, P = 0.029), and increased postpartum blood loss (397.5 ml vs. 185 ml, P < 0.001). Notably, the utilization of dinoprostone for induction resulted in a statistically significant cesarean delivery rate of 47.01% among SGA patients when compared to the Cook double balloon method. Table 2 provides the characteristics of pregnancies in both the derivation and validation cohorts, which shared similar maternal and fetal features, indicating the robustness of random group assignment.
Table 1 Univariate analysis of demographic and clinical characteristics of pregnancies with SGA stratified by mode of delivery (N=507)
Variables
|
Total
(N= 507)
|
vaginal delivery
(N= 373)
|
cesarean delivery
(N = 134)
|
P
|
|
|
Antenatal variables
|
|
|
|
|
|
Maternal age(y)
|
28.59 ± 4.36
|
28.19 ± 4.22
|
29.68 ± 4.58
|
<.001*
|
|
Maternal age, n(%)
|
|
|
|
0.006*
|
|
<35
|
458 (90.34)
|
345 (92.49)
|
113 (84.33)
|
|
|
≥35
|
49 (9.66)
|
28 (7.51)
|
21 (15.67)
|
|
|
Gravidity
|
1.62 ± 0.95
|
1.62 ± 0.92
|
1.63 ± 1.02
|
0.915
|
|
Gestational age at adimission
|
38.99 ± 1.19
|
39.00 ± 1.19
|
38.96 ± 1.20
|
0.707
|
|
Height(cm)
|
158.95 ± 5.16
|
158.95 ± 5.05
|
158.95 ± 5.47
|
0.993
|
|
Prepregnancy weight(kg)
|
50.44 ± 7.34
|
50.17 ± 7.32
|
51.17 ± 7.35
|
0.178
|
|
Prepregnancy BMI(kg/m2)
|
19.95 ± 2.68
|
19.85 ± 2.76
|
20.22 ± 2.42
|
0.174
|
|
Prepregnancy BMI,n(%)
|
|
|
|
0.314
|
|
Underweight(BMI < 18.5)
|
154 (30.37)
|
121 (32.44)
|
33 (24.63)
|
|
|
Normal(18.5 ≤ BMI < 25.0)
|
312 (61.54)
|
222 (59.52)
|
90 (67.16)
|
|
|
Overweigt(25.0 ≤BMI < 30.0)
|
38 (7.50)
|
28 (7.51)
|
10 (7.46)
|
|
|
Obese (BMI ≥ 30.0)
|
3 (0.59)
|
2 (0.54)
|
1 (0.75)
|
|
|
Admission weight(kg)
|
61.22 ± 9.31
|
60.65 ± 9.57
|
62.81 ± 8.39
|
0.015*
|
|
Admission BMI(kg/m2)
|
24.23 ± 3.50
|
24.00 ± 3.66
|
24.84 ± 2.93
|
0.008*
|
|
Admission BMI (%)
|
|
|
|
0.003*
|
|
Underweight(BMI < 18.5)
|
27 (5.33)
|
25 (6.70)
|
2 (1.49)
|
|
|
Normal(18.5 ≤ BMI < 25.0)
|
210 (41.42)
|
165 (44.24)
|
45 (33.58)
|
|
|
Overweight(25.0 ≤ BMI < 30)
|
205 (40.43)
|
135 (36.19)
|
70 (52.24)
|
|
|
Obese (BMI ≥ 30.0)
|
65 (12.82)
|
48 (12.87)
|
17 (12.69)
|
|
|
Parity , n(%)
|
|
|
|
0.263
|
|
0
|
383 (75.54)
|
277 (74.26)
|
106 (79.10)
|
|
|
≥1
|
124 (24.46)
|
96 (25.74)
|
28 (20.90)
|
|
|
GDM, n(%)
|
74 (14.60)
|
58 (15.55)
|
16 (11.94)
|
0.31
|
|
Gestational hypertension, n(%)
|
34 (6.71)
|
27 (7.24)
|
7 (5.22)
|
0.424
|
|
Oligoamnios, n(%)
|
59 (11.64)
|
37 (9.92)
|
22 (16.42)
|
0.044*
|
|
Bishop score before labor induction
|
2.85 ± 0.86
|
2.88 ± 0.86
|
2.76 ± 0.83
|
0.181
|
|
Estimated fetal weight(kg)
|
2650.20 ± 266.55
|
2655.31 ± 249.88
|
2635.96± 308.83
|
0.515
|
|
Intrapartum variables
|
|
|
|
|
|
Method of inducing labor, n(%)
|
|
|
|
0.007*
|
|
cook double balloon
|
318 (62.72)
|
247 (66.22)
|
71 (52.99)
|
|
|
dinoprostone
|
189 (37.28)
|
126 (33.78)
|
63 (47.01)
|
|
|
Bishop score after medication
|
6.82 ± 1.60
|
7.05 ± 1.60
|
6.17 ± 1.41
|
<.001*
|
|
Oxytocin agument, n(%)
|
230 (45.36)
|
171 (45.84)
|
59 (44.03)
|
0.717
|
|
Early Rupture of Membranes, n(%)
|
56 (11.05)
|
48 (12.87)
|
8 (5.97)
|
0.029*
|
|
Intrapartum fever, n(%)
|
12 (2.37)
|
9 (2.41)
|
3 (2.24)
|
1
|
|
Postpartum blood loss,
M (Q₁, Q₃)
|
205(155.0,361.5)
|
185(150,278)
|
397.5(196.25,420)
|
<.001*
|
|
Variables at birth
|
|
|
|
|
|
Gestational age at delivery(wk)
|
39.43 ± 1.16
|
39.44 ± 1.15
|
39.40 ± 1.18
|
0.714
|
|
Neonatal weight(kg)
|
2595.37 ± 206.66
|
2605.42 ± 195.89
|
2567.40 ± 232.55
|
0.068
|
|
1min Apgar score
|
9.89 ± 0.65
|
9.90 ± 0.62
|
9.85 ± 0.71
|
0.419
|
|
5min Apgar<7, n(%)
|
3 (0.59)
|
2 (0.54)
|
1 (0.75)
|
1
|
|
NICU, n(%)
|
187 (36.88)
|
138 (37.00)
|
49 (36.57)
|
0.929
|
|
Small for gestational age, n(%)
|
|
|
|
0.147
|
|
mild
|
418 (82.45)
|
313 (83.91)
|
105 (78.36)
|
|
|
severe
|
89 (17.55)
|
60 (16.09)
|
29 (21.64)
|
|
|
*P≤0.05. GDM:gestational diabetes mellitus; BMI:body mass index;NICU: Neonatal Intensive Care Units.
Table 2 Demographic and Clinical Characteristics of Patients With SGA Undergoing Induction of Labor
Variables
|
Training set
(N = 354)
|
Validation set
(N= 153)
|
P
|
|
|
Maternal age(y)
|
28.38 ± 4.39
|
29.07 ± 4.28
|
0.104
|
|
Maternal age, n(%)
|
|
|
0.293
|
|
<35
|
323 (91.24)
|
135 (88.24)
|
|
|
≥35
|
31 (8.76)
|
18 (11.76)
|
|
|
Prepregnancy weight(kg)
|
50.91 ± 7.63
|
49.33 ± 6.48
|
0.026*
|
|
Prepregnancy BMI(kg/m2)
|
20.02 ± 2.80
|
19.79 ± 2.36
|
0.386
|
|
Prepregnancy BMI,n(%)
|
|
|
0.79
|
|
Underweight(BMI < 18.5)
|
110 (31.07)
|
44 (28.76)
|
|
|
Normal(18.5 ≤ BMI < 25)
|
215 (60.73)
|
97 (63.40)
|
|
|
Overweight(25.0 ≤ BMI < 30)
|
26 (7.34)
|
12 (7.84)
|
|
|
Obese (BMI ≥ 30.0)
|
3 (0.85)
|
0 (0.00)
|
|
|
Height(cm)
|
157.82 ± 5.09
|
159.44 ± 5.12
|
0.001*
|
|
Parity , n(%)
|
|
|
0.896
|
|
0
|
268 (75.71)
|
115 (75.16)
|
|
|
≥1
|
86 (24.29)
|
38 (24.84)
|
|
|
Gravidity
|
1.69 ± 0.88
|
1.59 ± 0.98
|
0.297
|
|
Admission weight(kg)
|
60.53 ± 9.01
|
61.52 ± 9.44
|
0.269
|
|
Admission BMI(kg/m2)
|
24.28 ± 3.38
|
24.20 ± 3.55
|
0.812
|
|
Admission BMI (%)
|
|
|
0.586
|
|
Underweight(BMI < 18.5)
|
22 (6.21)
|
5 (3.27)
|
|
|
Normal(18.5 ≤ BMI < 25)
|
145 (40.96)
|
65 (42.48)
|
|
|
Overweight(25.0 ≤ BMI < 30)
|
141 (39.83)
|
64 (41.83)
|
|
|
Obese (BMI ≥ 30.0)
|
46 (12.99)
|
19 (12.42)
|
|
|
Estimated fetal weight(kg)
|
2662.67±264.79
|
2621.35±269.23
|
0.109
|
|
Bishop score before labor induction
|
2.85 ± 0.86
|
2.84 ± 0.86
|
0.952
|
|
Oxytocin agument, n(%)
|
157 (44.35)
|
73 (47.71)
|
0.49
|
|
Gestational age at adimission(wk)
|
38.96 ± 1.23
|
39.01 ± 1.18
|
0.709
|
|
Gestational age at delivery(wk)
|
39.42 ± 1.19
|
39.44 ± 1.15
|
0.885
|
|
GDM, n(%)
|
54 (15.25)
|
20 (13.07)
|
0.523
|
|
Gestational hypertension, n(%)
|
23 (6.50)
|
11 (7.19)
|
0.775
|
|
Oligoamnios, n(%)
|
43 (12.15)
|
16 (10.46)
|
0.586
|
|
Method of inducing labor, n(%)
|
|
|
0.836
|
|
cook double balloon
|
221 (62.43)
|
97 (63.40)
|
|
|
dinoprostone
|
133 (37.57)
|
56 (36.60)
|
|
|
Bishop score after medication
|
6.84 ± 1.58
|
6.81 ± 1.61
|
0.834
|
|
Oxytocin agument, n(%)
|
157 (44.35)
|
73 (47.71)
|
0.485
|
|
Delivery model, n(%)
|
|
|
0.451
|
|
vaginal delivery
|
257 (72.60)
|
116 (75.82)
|
|
|
cesarean section
|
97 (27.40)
|
37 (24.18)
|
|
|
Early Rupture of Membranes, n(%)
|
44 (12.43)
|
12 (7.84)
|
0.13
|
|
Intrapartum fever, n(%)
|
9 (2.54)
|
3 (1.96)
|
0.938
|
|
Postpartum blood loss
M (Q₁, Q₃)
|
205.0(150.0,325.0)
|
202.5(158.5,373.7)
|
0.516
|
|
Neonatal weight(kg)
|
2590.86±227.16
|
2597.32±197.45
|
0.747
|
|
1min Apgar score
|
9.89 ± 0.71
|
9.87 ± 0.56
|
0.854
|
|
5min Apgar score<7, n(%)
|
2 (0.56)
|
1 (0.65)
|
1
|
|
NICU, n(%)
|
141 (39.83)
|
46 (30.07)
|
0.036*
|
|
*P≤0.05. GDM:gestational diabetes mellitus; BMI:body mass index;NICU: Neonatal Intensive Care Units.
Predictive Variable Screening
Following univariate logistic regression analysis in the training set, several factors emerged with a p-value < 0.05 .There were then integrated into the logistic regression model,including maternal age, prepregnancy weight, weight at admission,body mass index (BMI) at admission, utilization of dinoprostone for induction, and Bishop score after cervical ripening. The independent risk factors linked to cesarean delivery following labor induction were uncovered through subsequent backward stepwise multivariate logistic regression analysis,as depicted in Table3. Four significant risk factors were identified: maternal age, admission weight, the use of dinoprostone for labor induction, and the Bishop score after medication. Notably, for each 1-point increase in Bishop score after medication, the odds of cesarean delivery decreased by 35% (aOR 0.65, 95% CI 0.54 - 0.80). Increments of 1 kg in admission weight and 1 year in maternal age corresponded to a 4.0% (aOR 1.04, 95% CI 1.01-1.07) and an 8.0% (aOR 1.08, 95% CI 1.01-1.15) rise in the likelihood of cesarean delivery, respectively. Furthermore, when comparing dinoprostone with the Cook double balloon method for labor induction, dinoprostone emerged as a notable risk factor, showing more than double the risk of cesarean delivery (aOR 2.08, 95% CI 1.13-3.81).
Table 3 Independent risk factors for cesarean delivery among patients with SGA undergoing induction of labor at term in training set.
Risk factors
|
Unadjusted
OR(95%CI)
|
P
|
Adjusted OR(95%CI)
|
P
|
|
|
Antenatal factors
|
|
|
|
|
|
Maternal age
|
1.11 (1.04 ~ 1.18)
|
0.001*
|
1.08 (1.01 ~ 1.15)
|
0.024*
|
|
Parity
|
|
|
|
|
|
0
|
1.00 (Reference)
|
|
|
|
|
≥1
|
1.09 (0.58 ~ 2.05)
|
0.79
|
|
|
|
Gravidity
|
1.12 (0.89 ~ 1.42)
|
0.328
|
|
|
|
Gestational age at adimission
|
0.96 (0.77 ~ 1.19)
|
0.721
|
|
|
|
Height
|
1.04 (0.98 ~ 1.09)
|
0.192
|
|
|
|
Prepregnancy weight
|
1.04 (1.01 ~ 1.08)
|
0.041*
|
1.02 (0.99 ~ 1.05)
|
0.246
|
|
Prepregnancy BMI
|
1.08 (0.98 ~ 1.20)
|
0.137
|
|
|
|
Prepregnancy BMI
|
|
|
|
|
|
Underweight(BMI < 18.5)
|
0.58 (0.32 ~ 1.06)
|
0.078
|
|
|
|
Normal(18.5 ≤ BMI < 25.0)
|
1.00 (Reference)
|
|
|
|
|
Overweight(25.0 ≤ BMI < 30.0)
|
0.94 (0.32 ~ 2.76)
|
0.903
|
|
|
|
Obese (BMI ≥ 30.0)
|
0.00 (0.00 ~ Inf)
|
0.988
|
|
|
|
Admission weight
|
1.05 (1.02 ~ 1.08)
|
<.001*
|
1.04 (1.01 ~ 1.07)
|
0.027*
|
|
Admission BMI
|
1.12 (1.04 ~ 1.21)
|
0.003*
|
|
|
|
Admission BMI
|
|
|
|
|
|
Underweight(BMI < 18.5)
|
0.59 (0.13 ~ 2.73)
|
0.498
|
0.53 (0.15 ~ 1.93)
|
0.338
|
|
Normal(18.5 ≤ BMI < 25.0)
|
1.00 (Reference)
|
|
1.00 (Reference)
|
|
|
Overweight(25.0 ≤ BMI < 30.0)
|
2.87 (1.58 ~ 5.20)
|
<.001*
|
2.58 (1.32 ~ 5.06)
|
0.078
|
|
Obese (BMI ≥ 30.0)
|
1.96 (0.83 ~ 4.63)
|
0.124
|
0.94 (0.32 ~ 2.76)
|
0.903
|
|
Gestational diabetes mellitus
|
0.53 (0.24 ~ 1.19)
|
0.122
|
|
|
|
Gestational hypertension
|
0.90 (0.32 ~ 2.54)
|
0.838
|
|
|
|
Oligoamnios
|
1.36 (0.61 ~ 3.03)
|
0.446
|
|
|
|
Bishop score before labor induction
|
0.93 (0.69 ~ 1.26)
|
0.646
|
|
|
|
Estimated fetal weight
|
1.00 (1.00 ~ 1.00)
|
0.716
|
|
|
|
Intrapartum factors
|
|
|
|
|
|
Gestational age at delivery
|
0.99 (0.79 ~ 1.23)
|
0.902
|
|
|
|
Method of inducing labor
|
|
|
|
|
|
cook double balloon
|
1.00 (Reference)
|
|
|
|
|
dinoprostone
|
2.00 (1.17 ~ 3.43)
|
0.011*
|
2.08 (1.13 ~ 3.81)
|
0.018*
|
|
Bishop score after medication
|
0.69 (0.57 ~ 0.83)
|
<.001*
|
0.65 (0.54 ~ 0.80)
|
<.001*
|
|
Oxytocin agument
|
1.23 (0.73 ~ 2.08)
|
0.445
|
|
|
|
Early rupture of membranes, n(%)
|
0.48 (0.16 ~ 1.42)
|
0.183
|
|
|
|
Intrapartum fever, n(%)
|
0.96 (0.19 ~ 4.88)
|
0.964
|
|
|
|
*P≤0.05,OR: Odds Ratio, CI: Confidence Interval
|
|
|
|
|
Development of the Nomogram
Utilizing the findings from the logistic regression analysis, we constructed a predictive nomogram for SGA patients undergoing labor induction. This model incorporates four predictive variables: maternal age, admission weight, the use of dinoprostone as an induction method, and the Bishop score after cervical ripening, along with the outcome variable of cesarean delivery (Figure 2).The individual scores for each predictive factors were added together, suggesting that a high overall score implies a greater likelihood of cesarean birth.
Model Performance in the Training Set
The prognostic efficacy of our model was performed within the training cohort. Discrimination is defined as the model's capacity to distinguish between events and non-events, measured by the area under the curve (AUC). As illustrated in Figure 3A, the nomogram achieved an AUC of 0.78 (95% CI: 0.73–0.84) in the training cohort, reflecting robust discriminative ability. With an optimal cutoff value of 36%, the nomogram demonstrated an accuracy of 0.79 (95% CI: 0.75–0.83), sensitivity of 0.86 (95% CI: 0.82–0.91), specificity of 0.61 (95% CI: 0.51–0.71), a positive predictive value (PPV) of 0.85 (95% CI: 0.81–0.90), and a negative predictive value (NPV) of 0.63 (95% CI: 0.53–0.73).
Calibration refers to the degree of alignment between predicted probabilities and actual outcomes. The calibration plots in the training cohort reveal that the predictions of the nomogram are strongly correlated with the data observed(P=0.397)(Figure 3B). We further applied decision curve analysis (DCA) to assess alternative prognostic strategies, illustrating that the model conferred higher net benefits across probability thresholds from 5% to 80%, highlighting its practical applicability in clinical decision-making within this probability range.
Performance of the model in the validation set
The prediction model was evaluated with our validation cohort. To assess discriminative ability, calibration accuracy, and clinical effectiveness of the nomogram, we employed a receiver operating characteristic (ROC) curve, calibration curve, and decision curve analysis (DCA). As demonstrated in Figure 4A, the AUC for the validation cohort stood at 0.77 (95% CI: 0.68–0.86), suggesting strong discriminative power in predicting outcomes. Nomogram accuracy was 0.75, sensitivity 0,84, and specificity 0.46.The calibration curve in the validation cohort closely mirrored the diagonal, evidencing robust calibration ability(P=0.812) (Figure 4B). Furthermore, the DCA curve remained consistently elevated across a broad spectrum of threshold probabilities (Figure 4C), indicating that our model provides noteworthy net benefits in forecasting the likelihood of cesarean delivery for SGA patients following labor induction.