- Email: [email protected]
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study
OBSTETRICS
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study Elizabeth Jeffers, MSc;1 Linda Dodds, PhD;1,2,3 Victoria Allen, MD;1,2 Christy Woolcott, PhD1,2,3 1
Department of Community Health and Epidemiology, Dalhousie University, Halifax, NS
2
Department of Obstetrics and Gynaecology, Dalhousie University, Halifax, NS
3
Department of Pediatrics, Dalhousie University, Halifax, NS
Abstract Objective: To identify factors known in mid-pregnancy to be associated with risk of macrosomia (4000 g) among non-diabetic women and to develop a risk score to allow early identification of women at high risk. Methods: Data were obtained from a population-based perinatal database and a hospital laboratory database in Nova Scotia, Canada. The study included singleton live births born to non-diabetic women between 1998 and 2005. Logistic regression was used to identify risk factors significantly associated with macrosomia. Risk scoring systems were developed for nulliparous and parous women separately and validated using the C-statistic. Results: Of the 23 857 mother-infant pairs included in the study, 16.7% of the infants were macrosomic. In nulliparous women, seven risk factors were identified, of which pre-pregnancy weight 90 kg with an OR of 4.8 (95% CI: 3.9 to 6.0) contributed a greater number of points to the risk score. The resulting risk score corresponded to a range of estimated risk of 0.2% to 47.0% and had a C-statistic of 0.70. In parous women, the most points were assigned to women with a previous large birth (OR: 3.7; 95% CI: 3.2e4.0) and a pre-pregnancy weight 90 kg (OR: 3.8; 95% CI: 3.1e4.7). The resulting risk score corresponded to a range of estimated risk of 0.4% to 88.0% and had a C-statistic of 0.75. Conclusions: Macrosomia risk can be estimated by a simple calculation based on a woman’s risk factor profile at mid-pregnancy.
Résumé Objectif : Mettre en évidence des facteurs qui, à la moitié de la grossesse, sont associés à un risque de macrosomie (4 000 g) chez les femmes non diabétiques, et mettre au point un score de risque permettant la détection précoce des femmes à risque élevé.
Key Words: Macrosomia, non-diabetics, pregnancy, risk score, prediction model, retrospective cohort Corresponding Author: Dr Linda Dodds, Perinatal Epidemiology Research Unit, Dalhousie University, Halifax, NS. [email protected] Competing interests: None declared. Received on March 3, 2017 Accepted on May 29, 2017
Méthodologie : Les données ont été tirées d’une base de données périnatales représentative de la population et d’une base de données de laboratoire en milieu hospitalier de la Nouvelle-Écosse. L’étude s’est penchée sur les naissances vivantes uniques survenues entre 1998 et 2005 pour lesquelles la mère ne souffrait pas de diabète. Les facteurs de risque présentant une association significative avec la macrosomie ont été mis en évidence par régression logistique, et des systèmes de scores de risque distincts ont été mis au point pour les femmes nullipares et pares, puis validés au moyen de la statistique C. Résultats : Parmi les 23 857 paires mères-enfants retenues dans l’étude, 16,7 % des nouveau-nés présentaient une macrosomie. Chez les mères nullipares, sept facteurs de risque ont été mis en évidence; celui qui valait le plus de points dans le score de risque était un poids avant grossesse de 90 kg ou plus, associé à un rapport de cotes (RC) de 4,8 (IC à 95 % : 3,9e6,0). Le score de risque résultant correspondait à une étendue de risque estimé de 0,2 % à 47,0 %, et la statistique C était de 0,70. Chez les femmes pares, les facteurs qui comptaient le plus dans le score étaient la naissance précédente d’un bébé présentant une macrosomie (RC : 3,7; IC à 95 % : 3,2 à 4,0), ainsi qu’un poids avant grossesse de 90 kg ou plus (RC : 3,8; IC à 95 % : 3,1e4,7). Le score de risque résultant correspondait à une étendue de risque estimé de 0,4 % à 88,0 %, et la statistique C était de 0,75. Conclusions : Le risque de macrosomie peut être estimé par un simple calcul fondé sur le profil de risque de la femme à la moitié de la grossesse. Copyright ª 2017 The Society of Obstetricians and Gynaecologists of Canada/La Société des obstétriciens et gynécologues du Canada. Published by Elsevier Inc. All rights reserved.
J Obstet Gynaecol Can 2017;-(-):1e8 https://doi.org/10.1016/j.jogc.2017.05.032
INTRODUCTION
M
acrosomia is a term used to describe an infant born with an excessively high birth weight. Macrosomia is commonly defined as a birth weight greater than 4000 g (approximately the 90th percentile at 40 weeks’ gestation). Fetal overgrowth can lead to significant complications during and after delivery for both the mother and the -
JOGC
-
2017
l
1
OBSTETRICS
infant. Mothers of macrosomic infants are more likely to have prolonged labours and require operative vaginal deliveries and CSs.1,2 Complications to macrosomic infants include traumatic injuries such as shoulder dystocia and long-term conditions such as obesity and diabetes.1,3 Both pre-existing and gestational diabetes are wellestablished risk factors for macrosomia, and the study of these relationships predominates existing research.2,4 This results in a lack of information about the non-diabetic population despite giving rise to approximately 90% of macrosomic births.4,5 Unlike the diabetic population, there is no common agreement on the prediction or management of non-diabetic women at risk of giving birth to macrosomic infants. Macrosomia causes considerable challenges for physicians. The birth weight of an infant is usually not known until very close to or until after delivery.1 Estimating fetal weight using clinical and ultrasound information is challenging, particularly with LGA fetuses.6,7 Predictive models focused on factors in mid-pregnancy would have the potential to assist health care providers in identifying possible cases of macrosomia at a stage in pregnancy when interventions that may help prevent or mitigate the effects of the condition are possible. Possible interventions for the nondiabetic population could be assumed to be similar to the effective interventions for diabetic women and include increased surveillance of high-risk pregnancies (including scheduled growth ultrasounds), dietary and lifestyle counseling, and early induction of labour if appropriate.8,9 For these reasons, the current study was designed to develop a risk score based on factors known at mid-pregnancy to estimate the risk of delivering a macrosomic infant in nondiabetic women. METHODS
This retrospective cohort study included all singleton live births to residents of the Canadian province of Nova Scotia who received prenatal glucose screening at the IWK Health Centre between 1998 and 2005. Births were excluded from
ABBREVIATIONS GCT
glucose challenge test
GDM
gestational diabetes
LBW
low birth weight
MSAFP
maternal serum alpha-fetoprotein levels
NSAPD
Nova Scotia Atlee Perinatal Data
ROC
receiver operator characteristic
2
l
-
JOGC
-
2017
the study if the mother had a pre-existing diagnosis of type 1 or type 2 diabetes or the mother had a diagnosis of GDM for the current pregnancy. Births with one or more major anomalies were also excluded. The Nova Scotia Atlee Perinatal Data base considers a major anomaly to be a “congenital abnormality that requires medical or surgical treatment, has a serious adverse effect on health and development, or has significant cosmetic impact.”10 Clinical data were obtained from the NSAPD, a high quality electronic database administered by the Reproductive Care Program of Nova Scotia. It compiles information from hospital charts using standardized collection forms and trained health record personnel. The database contains records of all hospital births of infants 500 g or greater or of a GA of 20 weeks or greater in Nova Scotia since 1988. The database undergoes data quality assurance checks, including reabstraction and validation studies to ensure its reliability.11 Results of glucose challenge tests for GDM screening were obtained from the laboratory database of the IWK Health Centre. In Canada, the GCT is administered to all women at 24 to 28 weeks gestation as a screening method for GDM. Macrosomia was defined as birth weight greater than or equal to 4000 g. Information on sociodemographic and behavioural factors, pregnancy history, and pre-existing medical conditions were obtained from the NSAPD as potential predictors of macrosomia based on associations found in past research. The GCT result at subclinical diabetic levels (<10.7 mmol/L), divided into quartiles, was also considered as a potential predictor. Height was only introduced as a variable captured by the NSAPD in 2003, resulting in a significant amount of missing data for our study population. Therefore, prepregnancy weight was used instead of BMI. The rate of weight gain at 26 weeks was predicted using a previously established formula that considers overall gestational weight gain and GA of the fetus.12 The scoring system for predicting macrosomia in nondiabetic women was developed using previously established methods.13 Two separate scoring systems were created separating nulliparous women and parous women to utilize information from previous births in the latter group. Significant risk factors were identified using backward stepwise logistic regression, and the regression coefficients and ORs with 95% CI for each significant risk factor were estimated. For continuous variables, categories were created for the ease of the risk scoring system. To determine the number of risk score points to be assigned in
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study
Table 1. Maternal characteristics of study cohort comparing infants with and without macrosomia (‡4000 g); frequencies and unadjusted ORs (and 95% CI) Variable
Macrosomia (n ¼ 3987) Na (%)
No macrosomia (n ¼ 19 870) Na (%)
Unadjusted OR (95% CI)
Maternal age <20 years
71 (1.7)
880 (4.4)
1 (Ref)
20e24 years
458 (11.5)
3068 (15.4)
1.85 (1.42e2.40)
25e29 years
1192 (29.9)
5746 (28.9)
2.57 (2.00e3.30)
30e34 years
1532 (38.4)
6670 (33.6)
2.85 (2.22e3.65)
35e39 years
628 (15.8)
2974 (15.0)
2.62 (2.02e3.38)
40 years
106 (2.6)
532 (2.7)
2.47 (1.79e3.40)
Parity Nulliparous
1588 (39.8)
9864 (49.6)
1 (Ref)
Multiparous
2399 (60.2)
10 006 (50.4)
1.49 (1.39e1.60)
Pre-pregnancy weight <60 kg
369 (9.2)
6235 (31.4)
1 (Ref)
60-69 kg
1052 (26.4)
5037 (25.3)
2.04 (1.83e2.26)
70-79 kg
770 (19.3)
2814 (14.2)
2.67 (2.38e2.99)
80-89 kg
431 (10.8)
1443 (7.3)
2.91 (2.55e3.34)
90 kg
584 (14.6)
1511 (7.6)
3.77 (3.33e4.28)
Marital status Single/widowed/divorced
542 (13.6)
4253 (21.4)
1 (Ref)
Married/common-law
3420 (85.8)
15 482 (77.9)
1.73 (1.57e1.91)
Neighbourhood income quintile Lowest
621 (15.6)
3942 (19.8)
1 (Ref)
Lower-middle
587 (14.7)
2996 (15.1)
1.24 (1.10e1.40)
Middle
765 (19.2)
3586 (18.0)
1.35 (1.21e1.52)
Middle-upper
853 (21.4)
3914 (19.7)
1.38 (1.24e1.55)
Highest
705 (17.7)
3151 (15.8)
1.42 (1.26e1.56)
Female
1562 (39.2)
10 127 (51.0)
1 (Ref)
Male
2425 (60.8)
9743 (49.0)
1.61 (1.50e1.73)
No
3571 (89.6)
15 782 (79.4)
1 (Ref)
Yes
383 (9.6)
3866 (19.4)
0.44 (0.39e0.49)
No
3979 (99.8)
19 781 (99.6)
1 (Ref)
Yes
8 (0.2)
89 (0.4)
0.45 (0.22e0.92)
No
3750 (94.0)
18 214 (91.7)
1 (Ref)
Yes
62 (0.2)
787 (4.0)
0.38 (0.29e0.50)
No
3083 (77.3)
18 175 (91.5)
1 (Ref)
Yes
731 (18.3)
832 (4.2)
5.18 (4.66e5.77)
Sex of infant
Smoked during pregnancy
Asthma
Previous birth <2500 g
Previous birth >9 lbs (4080 g)
Rate of weight gain at 26 weeks
a
<0.32 kg/week
471 (11.8)
3195 (16.1)
1 (Ref)
0.32-<0.45 kg/week
597 (15.0)
3576 (18.0)
1.13 (0.99e1.29)
0.45-<0.59 kg/week
758 (19.0)
3693 (18.6)
1.39 (1.22e1.58)
0.59 kg/week
1242 (31.2)
4494 (22.6)
1.87 (1.67e2.10)
Frequency may not total sample size due to missing values.
-
JOGC
-
2017
l
3
OBSTETRICS
Table 2. Adjusted regression coefficients, ORs, 95% CIs and points associated with each category of risk factor for macrosomia (‡4000 g) among nulliparous women (n [ 8629)a Risk factor
Regression coefficient (bi)
OR (95% CI)
Reference value (Wij)
Pointsb
0 ¼ W1REF
0 2
Infant sex Female Male
0.4151
1.51 (1.33e1.72)
1 0 ¼ W2REF
0
0.2914
1.34 (1.13e1.58)
1
1
0.0286
1.03 (1.02e1.03)
Marital status Not married Married Pre-pregnancy weightc <60 kg
52 ¼ W3REF
0
60-69 kg
65
2
70-79 kg
75
3
80-89 kg
85
5
90 kg
103
7
0 ¼ W4REF
0
Smoked during pregnancy 0.5109
No
0.60 (0.48e0.74)
Yes
1
3
Asthma 0 ¼ W5REF
No 1.4150
Yes
0.24 (0.06e1.03)
1
0 7
Psychiatric illness No
0¼ W6REF
Yes GCT resultsc
0.4062
0.67 (0.45e0.98)
0.1043
1.11 (1.05e1.17)
1
0 2
<5.1 mmol/L
3.7 ¼ W7REF
0
5.1-<5.9 mmol/L
5.5
1
5.9-<6.8 mmol/L
6.4
1
8.6
3
6.8-10.3 mmol/L Rate of weight gain at 26 weeksc
1.7871
5.97 (4.62e7.72)
<0.32 kg/week
0.17 ¼ W8REF
0
0.32-<0.45 kg/week
0.38
2
0.45-<0.59 kg/week
0.52
3
0.59 kg/week
0.88
6
a
Does not total sample size due to missing values.
b
Points ¼ bi(Wij e WiREF)/B, where B is a constant set at 0.2 (which is the regression unit corresponding to 1 point).
c
Regression coefficient and OR (95% CI) based on continuous representation of risk factor.
each category, the midpoint of each category was determined (e.g., the midpoint of the pre-pregnancy weight category of 60e69 kg was 65 kg). To minimize the influence of extreme values for the first and last category of continuous variables, midpoint values were determined using the 1st percentile and the 99th percentile, respectively. The category of each factor corresponding to the lowest risk was chosen as the base category and assigned 0 points in the scoring system. For each risk factor, the distance from base category to each additional category was computed in terms of regression units, and the points associated with each category of each risk factor were calculated and rounded to the nearest integer. Using the final regression model coefficients, the risk of macrosomia
4
l
-
JOGC
-
2017
associated with each possible score was estimated. The performance of the risk prediction model was assessed according to its ability to discriminate between those who did and did not develop macrosomia.14 The model was internally validated by calculating the C-statistic (area under the receiver operator characteristic curve) to measure the probability that the model assigned a higher risk to those who delivered a macrosomic infant than those who did not.14 The Youden Index (the point for which sensitivity and specificity are maximized) on the ROC curve was used to categorize women into high- and low-risk groups. All analyses were completed using STATA 13.1 software (Stata Corp, College Station, TX, U.S.). Data access was
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study
Figure. Range of estimated risk for each possible point total for models predicting macrosomia among nulliparous and multiparous women.
approved by the Joint Data Access Committee of the Reproductive Care Program and Research Ethics Board approval was received from the IWK Health Centre. RESULTS
This cohort comprised 23 857 deliveries to non-diabetic women, of which 16.7% were macrosomic infants. Characteristics of the cohort are displayed in Table 1 with unadjusted ORs for their association with macrosomia also shown.
Table 3 displays the estimates of the regression coefficients, ORs, and points associated with each risk factor for the parous model. Among the 8098 parous women, factors positively associated with macrosomia included male fetus, the previous birth of a large infant, GCT result, prepregnancy weight, and rate of weight gain as risk factors, and factors negatively associated with macrosomia included smoking during pregnancy, psychiatric illness, and the previous birth of a small infant. Smoking during pregnancy and the previous birth of a small infant both contributed the largest number of negative points (e5) and a pre-pregnancy weight 90 kg and the previous birth of a large infant both contributed the largest number of positive points (þ7). The range of total points was e12 to 25, corresponding to a range of estimated risk from 0.5% to 88.0% (Figure). The C-statistic for the parous model was 0.75. The cut-point that best discriminated parous women delivering a macrosomic infant from those who did not was a score of 13, which corresponds to an estimated risk of macrosomia of 36%. The sensitivity and specificity when using this cut-point were estimated to be 0.73 and 0.63 respectively, resulting in a positive likelihood ratio of 2.0 and a negative likelihood ratio of 0.4. DISCUSSION
Consistent with the ORs derived for this risk factor, having a previous large birth was one of the largest contributors of points in the risk prediction models for parous women. In the model predicting macrosomia among parous women, a previous large birth contributed the largest number of points (þ7) equal to the number of points given for prepregnancy weight 90 kg. Male infant sex was identified as a significant risk factor for macrosomia in both models. Male infant sex contributed 2 points in the model predicting macrosomia among nulliparous women, and 3 points among parous women.
Table 2 displays the estimates of the regression coefficients, ORs, and the points assigned in the risk score for the multivariable model of macrosomia among nulliparous women. Among the 8629 nulliparous women, being married or common-law, having a male fetus, GCT result, prepregnancy weight, and rate of weight gain were positively associated with macrosomia. Smoking during pregnancy and the presence of asthma and psychiatric illness were negatively associated with macrosomia. Asthma contributed the largest number of negative points (e7) and having a pre-pregnancy weight 90 kg contributed the largest number of positive points (þ7). The range of total points was e12 to 19, corresponding to a range of estimated risk from 0.2% to 46.9% (Figure). The C-statistic for the nulliparous model was 0.7.
Women with higher pre-pregnancy weight were also significantly more likely to deliver a macrosomic infant. The lowest category of risk contributed 0 points to the model and the highest category of risk contributed up to 7 points.
The cut-point that best discriminated nulliparous women delivering a macrosomic infant from those who did not was a score of 16, which corresponds to an estimated risk of macrosomia of 29%. The sensitivity and specificity when using this cut-point were estimated to be 0.66 and 0.63, respectively, resulting in a positive likelihood ratio of 1.8 and a negative likelihood ratio of 0.5.
The rate of weight gain during pregnancy (calculated at 26 weeks) was also a significant contributor to all models. The risk prediction models displayed a gradient of risk, contributing 1e3 points for each increasing category of rate of weight gain (resulting in 0 points for lowest category of risk and up to 6 points for the highest category of risk).
-
JOGC
-
2017
l
5
OBSTETRICS
Table 3. Regression coefficients, ORs, 95% CIs and points associated with each category of risk factor for macrosomia (‡4000 g) among parous women (n [ 8098)a Risk factor
Regression coefficient (bi)
OR (95% CI)
Reference value (Wij)
Pointsb
0 ¼ W1REF
0 3
Infant sex Female Male
0.6060
1.83 (1.62e2.07)
1
1.0622
0.34 (0.25e0.48)
1 0 ¼ W3REF
0
1.3294
3.78 (3.26e4.37)
1
7
0.02622
1.03 (1.02e1.03)
Previous birth <2500 g 0 ¼ W2REF
No Yes
0 5
Previous birth >9 lbs (4080 g) No Yes Pre-pregnancy weightc <60 kg
52 ¼ W4REF
0
60-69 kg
65
2
70-79 kg
75
3
80-89 kg
85
4
90 kg
104
7
Smoked during pregnancy 0.9180
No
0.40 (0.33e0.49)
Yes
0 ¼ W5REF 1
0 5
Psychiatric illness 0 ¼ W6REF
No Yes GCT resultsc
0.4123
0.66 (0.47e0.93)
0.1098
1.12 (1.06e1.17)
1
0 2
<5.1 mmol/L
3.8 ¼ W7REF
0
5.1-<5.9 mmol/L
5.5
1
5.9-<6.8 mmol/L
6.4
1
8.6
3
6.8-10.3 mmol/L Rate of weight gain at 26 weeksc
1.3320
3.79 (2.65e5.41)
<0.32 kg/week
0.12 ¼ W8REF
0
0.32-<0.45 kg/week
0.38
2
0.45-<0.59 kg/week
0.52
3
0.59 kg/week
0.84
5
a
Does not total sample size due to missing values.
b
Points ¼ bi(Wij e WiREF)/B, where B is a constant set at 0.2 (which is the regression unit corresponding to 1 point).
c
Regression coefficient and OR (95% CI) based on continuous representation of risk factor.
A number of factors were identified to be inversely associated with macrosomia in this study. Smoking during pregnancy and asthma had strong negative associations with macrosomia (contributing up to e7 points). Complications of psychiatric illnesses and previous birth of a low birth weight infant in pregnancy also contributed negative points in the nulliparous and parous models. This is the first study to build a risk prediction model among non-diabetic women to estimate the outcome macrosomia. Non-diabetics have been previously understudied despite giving rise to the majority of macrosomic infants. Other strengths of this study include the use of the NSAPD as the primary data source. The NSAPD is a large
6
l
-
JOGC
-
2017
population-based dataset that contains numerous maternal variables that could be explored as potential risk factors of macrosomia. As well, the inclusion of GCT results makes this study unique. In 2013 the Canadian guidelines for diagnosing GDM changed.15 Although the individuals in the cohort were diagnosed using the pre-existing criteria, the changing standards were taken into account by completing the analysis using the new criteria. In both the nulliparous and parous models, the same risk factors remained significant and only a one-point change was noted in the highest level of GCT result. By adapting to the change in diagnostic criteria, the model remains practical for future use.
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study
Recognizing the study’s limitations is also important. Factors not captured by the NSAPD, or the laboratory database could be associated with macrosomia and may need to be incorporated later to improve the prediction of macrosomia risk. One factor unavailable for inclusion was maternal serum alpha-fetoprotein levels. Baschat et al. demonstrated that women with very low MSAFP delivered infants with a mean birth weight 250 g higher than mothers without low MSAFP (P < 0.05).16 Incorporating data such as MSAFP may have the potential to improve the predictive probabilities of the risk prediction model. Maternal height has been recorded in the NSAPD only since 2003, such that we could investigate maternal weight rather than BMI. To support our use of maternal weight as a proxy for BMI, a similar analysis was completed comparing both variables using NSAPD data that were available for years after 2003. Both the analyses using weight and BMI resulted in very similar ORs. Due to the missing height data, it was also not possible to categorize the rate of weight gain according to the Institute of Medicine’s recommendations of pregnancy weight gain, which are based on prepregnancy BMI. Ethnicity, a potential predictor, was not available in the database to include in the study. However, according to the 2001 Census, the population of Halifax during the timeframe of this study was primarily Caucasian, with visible minorities accounting for only approximately 7% of the population.17 Finally, the prediction model requires further validation. External validation could be done in another provincial perinatal program or in Nova Scotia during a different time period. An estimated fetal weight of >5000 g is often used to recommend CSs. Individuals at high risk of delivering infants >5000 g, would therefore be an important population to target approaches for reducing fetal weight.18 The analysis was repeated using a cutoff of 5000 g to define macrosomia. Using this cutoff for macrosomia, several risk factors were eliminated from the model, including maternal age, psychiatric illness, and asthma. It is likely that as the cutoff for defining macrosomia increases, low-prevalence risk factors would become not significant because of low power. Therefore, in order to establish a model to identify risk factors for macrosomia 5000 g, a larger cohort of women should be used. Psychiatric illnesses such as depression have been linked to adverse behaviours such as smoking (a risk factor for LBW).19 In addition to behaviours associated with
depression, anti-depressive medications may affect neonatal outcomes.20 Similar to depression, the relationship between asthma and macrosomia may have multiple mechanisms. The adverse association between asthma and macrosomia has been observed in studies on LBW and is likely related to both the maternal disorder and the management, which includes inhaled steroids and bronchodilators.21,22 The risk prediction models were internally validated, with C-statistics of 0.70 and 0.75 for nulliparous and parous women respectively. Thus, the models could be considered reasonable at predicting macrosomia.23 It was observed that the parous model was slightly better at predicting risk than the nulliparous model, possibly because of the additional information of previous infant’s birth weight. The cut-points that best discriminated between women who would deliver a macrosomic infant from those who would not were determined. For nulliparous women, an estimated risk of 29% (score of 16) indicated a woman was high-risk for delivering a macrosomic infant. For parous women, the high-risk cut-off was at 36%. Although the sensitivity and specificity for these high-risk cut-off points were reasonable, the likelihood ratio for both parous and nulliparous women (1.8 and 2.0, respectively) suggests that the cut-off provides only a modest increase from the pretest to the post-test probability of macrosomia. While interventions exist for the treatment of GDM, no such evidence-based interventions aimed at decreasing the risk of macrosomia have been confirmed within the nondiabetic population. However, possible interventions for the non-diabetic population could be assumed to be similar to the effective interventions for diabetic women and include increased surveillance of high-risk pregnancies (including dietary and lifestyle counseling), scheduled growth ultrasounds, and early induction of labour if appropriate.8,9 CONCLUSION
A clinically relevant model to predict macrosomia in midpregnancy was developed to estimate the risk of macrosomia with an individual’s risk factor profile. Risk factors that were included in the model were those identified to be significantly associated with macrosomia and readily available in clinical practice at mid-pregnancy. Women at high risk of macrosomia could be targeted for interventions or additional follow-up.
-
JOGC
-
2017
l
7
OBSTETRICS
ACKNOWLEDGEMENTS
The authors thank the Reproductive Care Program of Nova Scotia for access to the data.
REFERENCES
10. DeSilva M, Munoz FM, Mcmillan M, et al. Congenital anomalies: Case definition and guidelines for data collection, analysis, and presentation of immunization safety data. Vaccine 2016;34:6015e26. 11. Joseph KS, Fahey J, Canadian Perinatal Surveillance System. Validation of perinatal data in the Discharge Abstract Database of the Canadian Institute for Health Information. Chronic Dis Can 2009;29:96e100. 12. Ashley-Martin J, Woolcott C. Gestational weight gain and postpartum weight retention in a cohort of Nova Scotian women. Matern Child Health J 2014;18:1927e35.
1. Jolly MC, Sebire NJ, Harris JP, et al. Risk factors for macrosomia and its clinical consequences: a study of 350,311 pregnancies. Eur J Obstet Gynecol Reprod Biol 2003;111:9e14.
13. Sullivan LM, Massaro JM, D’Agostino RB. Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med 2004;23:1631e60.
2. Mocanu EV, Greene RA, Byrne BM, et al. Obstetric and neonatal outcome of babies weighing more than 4.5 kg: an analysis by parity. Eur J Obstet Gynecol Reprod Biol 2000;92:229e33.
14. D’Agostino RB, Grundy S, Sullivan LM, et al. Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. JAMA 2001;286:180e7.
3. Gaudet L, Ferraro ZM, Wen SW, et al. Maternal obesity and occurrence of fetal macrosomia: a systematic review and meta-analysis. BioMed Res Int 2014;2014:640291.
15. Thompson D, Berger H, Feig D, et al. Diabetes and pregnancy. Can J Diabetes 2013;37(1 Suppl):S168e83.
4. Legardeur H, Girard G, Journy N, et al. Factors predictive of macrosomia in pregnancies with a positive oral glucose challenge test: importance of fasting plasma glucose. Diabetes Metab 2014;40:43e8. 5. Perinatal Epidemiology Research Unit, Dalhousie University. Nova Scotia Atlee Perinatal Database Report of Indicators: 2002-2011 [Internet]. 2012 Dec [cited]. Available from, http://rcp.nshealth.ca/sites/default/files/ publications/nsapd_indicator_report_2002_2011.pdf. Accessed July 21, 2017. 6. Goetzinger KR, Odibo AO, Shanks AL, et al. Clinical accuracy of estimated fetal weight in term pregnancies in a teaching hospital. J Matern Fetal Neonatal Med 2014;27:89e93. 7. Scioscia M, Vimercati A, Ceci O, et al. Estimation of birth weight by twodimensional ultrasonography: a critical appraisal of its accuracy. Obstet Gynecol 2008;111:57e65. 8. Hartling L, Dryden DM, Guthrie A, et al. Benefits and harms of treating gestational diabetes mellitus: a systematic review and meta- analysis for the U.S. Preventive Services Task Force and the National Institutes of Health Office of Medical Applications of Research. Ann Intern Med 2013;159:123e9. 9. Boulvain M, Senat M, Perrotin F, et al. Induction of labour versus expectant management for large-for-date fetuses: a randomised controlled trial. Lancet 2015;385:2600e5.
8
l
-
JOGC
-
2017
16. Baschat AA, Harman CR, Farid G, et al. Very low second-trimester maternal serum alpha-fetoprotein: Association with high birth weight. Obstet Gynecol 2002;99:531e6. 17. Statistics Canada. 2001 Census: Proportion of visible minorities, census metropolitan areas, 2001, 1996 and 1991 [Internet]. [cited]. Available from: http://www12.statcan.ca/english/census01/products/analytic/ companion/etoimm/subprovs.cfm. Accessed July 21, 2017. 18. Practice Bulletin No. 173: Fetal Macrosomia. The American College of Obstetricians and Gynecologists. Obstet Gynecol 2016;128:e195e209. 19. Grote NK, Bridge JA, Gavin AR, et al. A meta-analysis of depression during pregnancy and the risk of preterm birth, low birth weight, and intrauterine growth restriction. Arch Gen Psychiatry 2010;67:1012e24. 20. Banderali G, Martelli A, Landi M, et al. Short and long term health effects of parental tobacco smoking during pregnancy and lactation: a descriptive review. J Transl Med 2015;13:327. 21. Clark JM, Hulme E, Devendrakumar V, et al. Effect of maternal asthma on birthweight and neonatal outcome in a British inner-city population. Paediatr Perinat Epidemiol 2007;21:154e62. 22. Motta M, Rodriguez-Perez C, Tincani A, et al. Neonates born from mothers with autoimmune disorders. Early Hum Dev 2009;85:S67e70. 23. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley; 2000.
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study Elizabeth Jeffers, MSc;1 Linda Dodds, PhD;1,2,3 Victoria Allen, MD;1,2 Christy Woolcott, PhD1,2,3 1
Department of Community Health and Epidemiology, Dalhousie University, Halifax, NS
2
Department of Obstetrics and Gynaecology, Dalhousie University, Halifax, NS
3
Department of Pediatrics, Dalhousie University, Halifax, NS
Abstract Objective: To identify factors known in mid-pregnancy to be associated with risk of macrosomia (4000 g) among non-diabetic women and to develop a risk score to allow early identification of women at high risk. Methods: Data were obtained from a population-based perinatal database and a hospital laboratory database in Nova Scotia, Canada. The study included singleton live births born to non-diabetic women between 1998 and 2005. Logistic regression was used to identify risk factors significantly associated with macrosomia. Risk scoring systems were developed for nulliparous and parous women separately and validated using the C-statistic. Results: Of the 23 857 mother-infant pairs included in the study, 16.7% of the infants were macrosomic. In nulliparous women, seven risk factors were identified, of which pre-pregnancy weight 90 kg with an OR of 4.8 (95% CI: 3.9 to 6.0) contributed a greater number of points to the risk score. The resulting risk score corresponded to a range of estimated risk of 0.2% to 47.0% and had a C-statistic of 0.70. In parous women, the most points were assigned to women with a previous large birth (OR: 3.7; 95% CI: 3.2e4.0) and a pre-pregnancy weight 90 kg (OR: 3.8; 95% CI: 3.1e4.7). The resulting risk score corresponded to a range of estimated risk of 0.4% to 88.0% and had a C-statistic of 0.75. Conclusions: Macrosomia risk can be estimated by a simple calculation based on a woman’s risk factor profile at mid-pregnancy.
Résumé Objectif : Mettre en évidence des facteurs qui, à la moitié de la grossesse, sont associés à un risque de macrosomie (4 000 g) chez les femmes non diabétiques, et mettre au point un score de risque permettant la détection précoce des femmes à risque élevé.
Key Words: Macrosomia, non-diabetics, pregnancy, risk score, prediction model, retrospective cohort Corresponding Author: Dr Linda Dodds, Perinatal Epidemiology Research Unit, Dalhousie University, Halifax, NS. [email protected] Competing interests: None declared. Received on March 3, 2017 Accepted on May 29, 2017
Méthodologie : Les données ont été tirées d’une base de données périnatales représentative de la population et d’une base de données de laboratoire en milieu hospitalier de la Nouvelle-Écosse. L’étude s’est penchée sur les naissances vivantes uniques survenues entre 1998 et 2005 pour lesquelles la mère ne souffrait pas de diabète. Les facteurs de risque présentant une association significative avec la macrosomie ont été mis en évidence par régression logistique, et des systèmes de scores de risque distincts ont été mis au point pour les femmes nullipares et pares, puis validés au moyen de la statistique C. Résultats : Parmi les 23 857 paires mères-enfants retenues dans l’étude, 16,7 % des nouveau-nés présentaient une macrosomie. Chez les mères nullipares, sept facteurs de risque ont été mis en évidence; celui qui valait le plus de points dans le score de risque était un poids avant grossesse de 90 kg ou plus, associé à un rapport de cotes (RC) de 4,8 (IC à 95 % : 3,9e6,0). Le score de risque résultant correspondait à une étendue de risque estimé de 0,2 % à 47,0 %, et la statistique C était de 0,70. Chez les femmes pares, les facteurs qui comptaient le plus dans le score étaient la naissance précédente d’un bébé présentant une macrosomie (RC : 3,7; IC à 95 % : 3,2 à 4,0), ainsi qu’un poids avant grossesse de 90 kg ou plus (RC : 3,8; IC à 95 % : 3,1e4,7). Le score de risque résultant correspondait à une étendue de risque estimé de 0,4 % à 88,0 %, et la statistique C était de 0,75. Conclusions : Le risque de macrosomie peut être estimé par un simple calcul fondé sur le profil de risque de la femme à la moitié de la grossesse. Copyright ª 2017 The Society of Obstetricians and Gynaecologists of Canada/La Société des obstétriciens et gynécologues du Canada. Published by Elsevier Inc. All rights reserved.
J Obstet Gynaecol Can 2017;-(-):1e8 https://doi.org/10.1016/j.jogc.2017.05.032
INTRODUCTION
M
acrosomia is a term used to describe an infant born with an excessively high birth weight. Macrosomia is commonly defined as a birth weight greater than 4000 g (approximately the 90th percentile at 40 weeks’ gestation). Fetal overgrowth can lead to significant complications during and after delivery for both the mother and the -
JOGC
-
2017
l
1
OBSTETRICS
infant. Mothers of macrosomic infants are more likely to have prolonged labours and require operative vaginal deliveries and CSs.1,2 Complications to macrosomic infants include traumatic injuries such as shoulder dystocia and long-term conditions such as obesity and diabetes.1,3 Both pre-existing and gestational diabetes are wellestablished risk factors for macrosomia, and the study of these relationships predominates existing research.2,4 This results in a lack of information about the non-diabetic population despite giving rise to approximately 90% of macrosomic births.4,5 Unlike the diabetic population, there is no common agreement on the prediction or management of non-diabetic women at risk of giving birth to macrosomic infants. Macrosomia causes considerable challenges for physicians. The birth weight of an infant is usually not known until very close to or until after delivery.1 Estimating fetal weight using clinical and ultrasound information is challenging, particularly with LGA fetuses.6,7 Predictive models focused on factors in mid-pregnancy would have the potential to assist health care providers in identifying possible cases of macrosomia at a stage in pregnancy when interventions that may help prevent or mitigate the effects of the condition are possible. Possible interventions for the nondiabetic population could be assumed to be similar to the effective interventions for diabetic women and include increased surveillance of high-risk pregnancies (including scheduled growth ultrasounds), dietary and lifestyle counseling, and early induction of labour if appropriate.8,9 For these reasons, the current study was designed to develop a risk score based on factors known at mid-pregnancy to estimate the risk of delivering a macrosomic infant in nondiabetic women. METHODS
This retrospective cohort study included all singleton live births to residents of the Canadian province of Nova Scotia who received prenatal glucose screening at the IWK Health Centre between 1998 and 2005. Births were excluded from
ABBREVIATIONS GCT
glucose challenge test
GDM
gestational diabetes
LBW
low birth weight
MSAFP
maternal serum alpha-fetoprotein levels
NSAPD
Nova Scotia Atlee Perinatal Data
ROC
receiver operator characteristic
2
l
-
JOGC
-
2017
the study if the mother had a pre-existing diagnosis of type 1 or type 2 diabetes or the mother had a diagnosis of GDM for the current pregnancy. Births with one or more major anomalies were also excluded. The Nova Scotia Atlee Perinatal Data base considers a major anomaly to be a “congenital abnormality that requires medical or surgical treatment, has a serious adverse effect on health and development, or has significant cosmetic impact.”10 Clinical data were obtained from the NSAPD, a high quality electronic database administered by the Reproductive Care Program of Nova Scotia. It compiles information from hospital charts using standardized collection forms and trained health record personnel. The database contains records of all hospital births of infants 500 g or greater or of a GA of 20 weeks or greater in Nova Scotia since 1988. The database undergoes data quality assurance checks, including reabstraction and validation studies to ensure its reliability.11 Results of glucose challenge tests for GDM screening were obtained from the laboratory database of the IWK Health Centre. In Canada, the GCT is administered to all women at 24 to 28 weeks gestation as a screening method for GDM. Macrosomia was defined as birth weight greater than or equal to 4000 g. Information on sociodemographic and behavioural factors, pregnancy history, and pre-existing medical conditions were obtained from the NSAPD as potential predictors of macrosomia based on associations found in past research. The GCT result at subclinical diabetic levels (<10.7 mmol/L), divided into quartiles, was also considered as a potential predictor. Height was only introduced as a variable captured by the NSAPD in 2003, resulting in a significant amount of missing data for our study population. Therefore, prepregnancy weight was used instead of BMI. The rate of weight gain at 26 weeks was predicted using a previously established formula that considers overall gestational weight gain and GA of the fetus.12 The scoring system for predicting macrosomia in nondiabetic women was developed using previously established methods.13 Two separate scoring systems were created separating nulliparous women and parous women to utilize information from previous births in the latter group. Significant risk factors were identified using backward stepwise logistic regression, and the regression coefficients and ORs with 95% CI for each significant risk factor were estimated. For continuous variables, categories were created for the ease of the risk scoring system. To determine the number of risk score points to be assigned in
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study
Table 1. Maternal characteristics of study cohort comparing infants with and without macrosomia (‡4000 g); frequencies and unadjusted ORs (and 95% CI) Variable
Macrosomia (n ¼ 3987) Na (%)
No macrosomia (n ¼ 19 870) Na (%)
Unadjusted OR (95% CI)
Maternal age <20 years
71 (1.7)
880 (4.4)
1 (Ref)
20e24 years
458 (11.5)
3068 (15.4)
1.85 (1.42e2.40)
25e29 years
1192 (29.9)
5746 (28.9)
2.57 (2.00e3.30)
30e34 years
1532 (38.4)
6670 (33.6)
2.85 (2.22e3.65)
35e39 years
628 (15.8)
2974 (15.0)
2.62 (2.02e3.38)
40 years
106 (2.6)
532 (2.7)
2.47 (1.79e3.40)
Parity Nulliparous
1588 (39.8)
9864 (49.6)
1 (Ref)
Multiparous
2399 (60.2)
10 006 (50.4)
1.49 (1.39e1.60)
Pre-pregnancy weight <60 kg
369 (9.2)
6235 (31.4)
1 (Ref)
60-69 kg
1052 (26.4)
5037 (25.3)
2.04 (1.83e2.26)
70-79 kg
770 (19.3)
2814 (14.2)
2.67 (2.38e2.99)
80-89 kg
431 (10.8)
1443 (7.3)
2.91 (2.55e3.34)
90 kg
584 (14.6)
1511 (7.6)
3.77 (3.33e4.28)
Marital status Single/widowed/divorced
542 (13.6)
4253 (21.4)
1 (Ref)
Married/common-law
3420 (85.8)
15 482 (77.9)
1.73 (1.57e1.91)
Neighbourhood income quintile Lowest
621 (15.6)
3942 (19.8)
1 (Ref)
Lower-middle
587 (14.7)
2996 (15.1)
1.24 (1.10e1.40)
Middle
765 (19.2)
3586 (18.0)
1.35 (1.21e1.52)
Middle-upper
853 (21.4)
3914 (19.7)
1.38 (1.24e1.55)
Highest
705 (17.7)
3151 (15.8)
1.42 (1.26e1.56)
Female
1562 (39.2)
10 127 (51.0)
1 (Ref)
Male
2425 (60.8)
9743 (49.0)
1.61 (1.50e1.73)
No
3571 (89.6)
15 782 (79.4)
1 (Ref)
Yes
383 (9.6)
3866 (19.4)
0.44 (0.39e0.49)
No
3979 (99.8)
19 781 (99.6)
1 (Ref)
Yes
8 (0.2)
89 (0.4)
0.45 (0.22e0.92)
No
3750 (94.0)
18 214 (91.7)
1 (Ref)
Yes
62 (0.2)
787 (4.0)
0.38 (0.29e0.50)
No
3083 (77.3)
18 175 (91.5)
1 (Ref)
Yes
731 (18.3)
832 (4.2)
5.18 (4.66e5.77)
Sex of infant
Smoked during pregnancy
Asthma
Previous birth <2500 g
Previous birth >9 lbs (4080 g)
Rate of weight gain at 26 weeks
a
<0.32 kg/week
471 (11.8)
3195 (16.1)
1 (Ref)
0.32-<0.45 kg/week
597 (15.0)
3576 (18.0)
1.13 (0.99e1.29)
0.45-<0.59 kg/week
758 (19.0)
3693 (18.6)
1.39 (1.22e1.58)
0.59 kg/week
1242 (31.2)
4494 (22.6)
1.87 (1.67e2.10)
Frequency may not total sample size due to missing values.
-
JOGC
-
2017
l
3
OBSTETRICS
Table 2. Adjusted regression coefficients, ORs, 95% CIs and points associated with each category of risk factor for macrosomia (‡4000 g) among nulliparous women (n [ 8629)a Risk factor
Regression coefficient (bi)
OR (95% CI)
Reference value (Wij)
Pointsb
0 ¼ W1REF
0 2
Infant sex Female Male
0.4151
1.51 (1.33e1.72)
1 0 ¼ W2REF
0
0.2914
1.34 (1.13e1.58)
1
1
0.0286
1.03 (1.02e1.03)
Marital status Not married Married Pre-pregnancy weightc <60 kg
52 ¼ W3REF
0
60-69 kg
65
2
70-79 kg
75
3
80-89 kg
85
5
90 kg
103
7
0 ¼ W4REF
0
Smoked during pregnancy 0.5109
No
0.60 (0.48e0.74)
Yes
1
3
Asthma 0 ¼ W5REF
No 1.4150
Yes
0.24 (0.06e1.03)
1
0 7
Psychiatric illness No
0¼ W6REF
Yes GCT resultsc
0.4062
0.67 (0.45e0.98)
0.1043
1.11 (1.05e1.17)
1
0 2
<5.1 mmol/L
3.7 ¼ W7REF
0
5.1-<5.9 mmol/L
5.5
1
5.9-<6.8 mmol/L
6.4
1
8.6
3
6.8-10.3 mmol/L Rate of weight gain at 26 weeksc
1.7871
5.97 (4.62e7.72)
<0.32 kg/week
0.17 ¼ W8REF
0
0.32-<0.45 kg/week
0.38
2
0.45-<0.59 kg/week
0.52
3
0.59 kg/week
0.88
6
a
Does not total sample size due to missing values.
b
Points ¼ bi(Wij e WiREF)/B, where B is a constant set at 0.2 (which is the regression unit corresponding to 1 point).
c
Regression coefficient and OR (95% CI) based on continuous representation of risk factor.
each category, the midpoint of each category was determined (e.g., the midpoint of the pre-pregnancy weight category of 60e69 kg was 65 kg). To minimize the influence of extreme values for the first and last category of continuous variables, midpoint values were determined using the 1st percentile and the 99th percentile, respectively. The category of each factor corresponding to the lowest risk was chosen as the base category and assigned 0 points in the scoring system. For each risk factor, the distance from base category to each additional category was computed in terms of regression units, and the points associated with each category of each risk factor were calculated and rounded to the nearest integer. Using the final regression model coefficients, the risk of macrosomia
4
l
-
JOGC
-
2017
associated with each possible score was estimated. The performance of the risk prediction model was assessed according to its ability to discriminate between those who did and did not develop macrosomia.14 The model was internally validated by calculating the C-statistic (area under the receiver operator characteristic curve) to measure the probability that the model assigned a higher risk to those who delivered a macrosomic infant than those who did not.14 The Youden Index (the point for which sensitivity and specificity are maximized) on the ROC curve was used to categorize women into high- and low-risk groups. All analyses were completed using STATA 13.1 software (Stata Corp, College Station, TX, U.S.). Data access was
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study
Figure. Range of estimated risk for each possible point total for models predicting macrosomia among nulliparous and multiparous women.
approved by the Joint Data Access Committee of the Reproductive Care Program and Research Ethics Board approval was received from the IWK Health Centre. RESULTS
This cohort comprised 23 857 deliveries to non-diabetic women, of which 16.7% were macrosomic infants. Characteristics of the cohort are displayed in Table 1 with unadjusted ORs for their association with macrosomia also shown.
Table 3 displays the estimates of the regression coefficients, ORs, and points associated with each risk factor for the parous model. Among the 8098 parous women, factors positively associated with macrosomia included male fetus, the previous birth of a large infant, GCT result, prepregnancy weight, and rate of weight gain as risk factors, and factors negatively associated with macrosomia included smoking during pregnancy, psychiatric illness, and the previous birth of a small infant. Smoking during pregnancy and the previous birth of a small infant both contributed the largest number of negative points (e5) and a pre-pregnancy weight 90 kg and the previous birth of a large infant both contributed the largest number of positive points (þ7). The range of total points was e12 to 25, corresponding to a range of estimated risk from 0.5% to 88.0% (Figure). The C-statistic for the parous model was 0.75. The cut-point that best discriminated parous women delivering a macrosomic infant from those who did not was a score of 13, which corresponds to an estimated risk of macrosomia of 36%. The sensitivity and specificity when using this cut-point were estimated to be 0.73 and 0.63 respectively, resulting in a positive likelihood ratio of 2.0 and a negative likelihood ratio of 0.4. DISCUSSION
Consistent with the ORs derived for this risk factor, having a previous large birth was one of the largest contributors of points in the risk prediction models for parous women. In the model predicting macrosomia among parous women, a previous large birth contributed the largest number of points (þ7) equal to the number of points given for prepregnancy weight 90 kg. Male infant sex was identified as a significant risk factor for macrosomia in both models. Male infant sex contributed 2 points in the model predicting macrosomia among nulliparous women, and 3 points among parous women.
Table 2 displays the estimates of the regression coefficients, ORs, and the points assigned in the risk score for the multivariable model of macrosomia among nulliparous women. Among the 8629 nulliparous women, being married or common-law, having a male fetus, GCT result, prepregnancy weight, and rate of weight gain were positively associated with macrosomia. Smoking during pregnancy and the presence of asthma and psychiatric illness were negatively associated with macrosomia. Asthma contributed the largest number of negative points (e7) and having a pre-pregnancy weight 90 kg contributed the largest number of positive points (þ7). The range of total points was e12 to 19, corresponding to a range of estimated risk from 0.2% to 46.9% (Figure). The C-statistic for the nulliparous model was 0.7.
Women with higher pre-pregnancy weight were also significantly more likely to deliver a macrosomic infant. The lowest category of risk contributed 0 points to the model and the highest category of risk contributed up to 7 points.
The cut-point that best discriminated nulliparous women delivering a macrosomic infant from those who did not was a score of 16, which corresponds to an estimated risk of macrosomia of 29%. The sensitivity and specificity when using this cut-point were estimated to be 0.66 and 0.63, respectively, resulting in a positive likelihood ratio of 1.8 and a negative likelihood ratio of 0.5.
The rate of weight gain during pregnancy (calculated at 26 weeks) was also a significant contributor to all models. The risk prediction models displayed a gradient of risk, contributing 1e3 points for each increasing category of rate of weight gain (resulting in 0 points for lowest category of risk and up to 6 points for the highest category of risk).
-
JOGC
-
2017
l
5
OBSTETRICS
Table 3. Regression coefficients, ORs, 95% CIs and points associated with each category of risk factor for macrosomia (‡4000 g) among parous women (n [ 8098)a Risk factor
Regression coefficient (bi)
OR (95% CI)
Reference value (Wij)
Pointsb
0 ¼ W1REF
0 3
Infant sex Female Male
0.6060
1.83 (1.62e2.07)
1
1.0622
0.34 (0.25e0.48)
1 0 ¼ W3REF
0
1.3294
3.78 (3.26e4.37)
1
7
0.02622
1.03 (1.02e1.03)
Previous birth <2500 g 0 ¼ W2REF
No Yes
0 5
Previous birth >9 lbs (4080 g) No Yes Pre-pregnancy weightc <60 kg
52 ¼ W4REF
0
60-69 kg
65
2
70-79 kg
75
3
80-89 kg
85
4
90 kg
104
7
Smoked during pregnancy 0.9180
No
0.40 (0.33e0.49)
Yes
0 ¼ W5REF 1
0 5
Psychiatric illness 0 ¼ W6REF
No Yes GCT resultsc
0.4123
0.66 (0.47e0.93)
0.1098
1.12 (1.06e1.17)
1
0 2
<5.1 mmol/L
3.8 ¼ W7REF
0
5.1-<5.9 mmol/L
5.5
1
5.9-<6.8 mmol/L
6.4
1
8.6
3
6.8-10.3 mmol/L Rate of weight gain at 26 weeksc
1.3320
3.79 (2.65e5.41)
<0.32 kg/week
0.12 ¼ W8REF
0
0.32-<0.45 kg/week
0.38
2
0.45-<0.59 kg/week
0.52
3
0.59 kg/week
0.84
5
a
Does not total sample size due to missing values.
b
Points ¼ bi(Wij e WiREF)/B, where B is a constant set at 0.2 (which is the regression unit corresponding to 1 point).
c
Regression coefficient and OR (95% CI) based on continuous representation of risk factor.
A number of factors were identified to be inversely associated with macrosomia in this study. Smoking during pregnancy and asthma had strong negative associations with macrosomia (contributing up to e7 points). Complications of psychiatric illnesses and previous birth of a low birth weight infant in pregnancy also contributed negative points in the nulliparous and parous models. This is the first study to build a risk prediction model among non-diabetic women to estimate the outcome macrosomia. Non-diabetics have been previously understudied despite giving rise to the majority of macrosomic infants. Other strengths of this study include the use of the NSAPD as the primary data source. The NSAPD is a large
6
l
-
JOGC
-
2017
population-based dataset that contains numerous maternal variables that could be explored as potential risk factors of macrosomia. As well, the inclusion of GCT results makes this study unique. In 2013 the Canadian guidelines for diagnosing GDM changed.15 Although the individuals in the cohort were diagnosed using the pre-existing criteria, the changing standards were taken into account by completing the analysis using the new criteria. In both the nulliparous and parous models, the same risk factors remained significant and only a one-point change was noted in the highest level of GCT result. By adapting to the change in diagnostic criteria, the model remains practical for future use.
Predicting the Risk of Macrosomia at Mid-Pregnancy Among Non-Diabetics: A Retrospective Cohort Study
Recognizing the study’s limitations is also important. Factors not captured by the NSAPD, or the laboratory database could be associated with macrosomia and may need to be incorporated later to improve the prediction of macrosomia risk. One factor unavailable for inclusion was maternal serum alpha-fetoprotein levels. Baschat et al. demonstrated that women with very low MSAFP delivered infants with a mean birth weight 250 g higher than mothers without low MSAFP (P < 0.05).16 Incorporating data such as MSAFP may have the potential to improve the predictive probabilities of the risk prediction model. Maternal height has been recorded in the NSAPD only since 2003, such that we could investigate maternal weight rather than BMI. To support our use of maternal weight as a proxy for BMI, a similar analysis was completed comparing both variables using NSAPD data that were available for years after 2003. Both the analyses using weight and BMI resulted in very similar ORs. Due to the missing height data, it was also not possible to categorize the rate of weight gain according to the Institute of Medicine’s recommendations of pregnancy weight gain, which are based on prepregnancy BMI. Ethnicity, a potential predictor, was not available in the database to include in the study. However, according to the 2001 Census, the population of Halifax during the timeframe of this study was primarily Caucasian, with visible minorities accounting for only approximately 7% of the population.17 Finally, the prediction model requires further validation. External validation could be done in another provincial perinatal program or in Nova Scotia during a different time period. An estimated fetal weight of >5000 g is often used to recommend CSs. Individuals at high risk of delivering infants >5000 g, would therefore be an important population to target approaches for reducing fetal weight.18 The analysis was repeated using a cutoff of 5000 g to define macrosomia. Using this cutoff for macrosomia, several risk factors were eliminated from the model, including maternal age, psychiatric illness, and asthma. It is likely that as the cutoff for defining macrosomia increases, low-prevalence risk factors would become not significant because of low power. Therefore, in order to establish a model to identify risk factors for macrosomia 5000 g, a larger cohort of women should be used. Psychiatric illnesses such as depression have been linked to adverse behaviours such as smoking (a risk factor for LBW).19 In addition to behaviours associated with
depression, anti-depressive medications may affect neonatal outcomes.20 Similar to depression, the relationship between asthma and macrosomia may have multiple mechanisms. The adverse association between asthma and macrosomia has been observed in studies on LBW and is likely related to both the maternal disorder and the management, which includes inhaled steroids and bronchodilators.21,22 The risk prediction models were internally validated, with C-statistics of 0.70 and 0.75 for nulliparous and parous women respectively. Thus, the models could be considered reasonable at predicting macrosomia.23 It was observed that the parous model was slightly better at predicting risk than the nulliparous model, possibly because of the additional information of previous infant’s birth weight. The cut-points that best discriminated between women who would deliver a macrosomic infant from those who would not were determined. For nulliparous women, an estimated risk of 29% (score of 16) indicated a woman was high-risk for delivering a macrosomic infant. For parous women, the high-risk cut-off was at 36%. Although the sensitivity and specificity for these high-risk cut-off points were reasonable, the likelihood ratio for both parous and nulliparous women (1.8 and 2.0, respectively) suggests that the cut-off provides only a modest increase from the pretest to the post-test probability of macrosomia. While interventions exist for the treatment of GDM, no such evidence-based interventions aimed at decreasing the risk of macrosomia have been confirmed within the nondiabetic population. However, possible interventions for the non-diabetic population could be assumed to be similar to the effective interventions for diabetic women and include increased surveillance of high-risk pregnancies (including dietary and lifestyle counseling), scheduled growth ultrasounds, and early induction of labour if appropriate.8,9 CONCLUSION
A clinically relevant model to predict macrosomia in midpregnancy was developed to estimate the risk of macrosomia with an individual’s risk factor profile. Risk factors that were included in the model were those identified to be significantly associated with macrosomia and readily available in clinical practice at mid-pregnancy. Women at high risk of macrosomia could be targeted for interventions or additional follow-up.
-
JOGC
-
2017
l
7
OBSTETRICS
ACKNOWLEDGEMENTS
The authors thank the Reproductive Care Program of Nova Scotia for access to the data.
REFERENCES
10. DeSilva M, Munoz FM, Mcmillan M, et al. Congenital anomalies: Case definition and guidelines for data collection, analysis, and presentation of immunization safety data. Vaccine 2016;34:6015e26. 11. Joseph KS, Fahey J, Canadian Perinatal Surveillance System. Validation of perinatal data in the Discharge Abstract Database of the Canadian Institute for Health Information. Chronic Dis Can 2009;29:96e100. 12. Ashley-Martin J, Woolcott C. Gestational weight gain and postpartum weight retention in a cohort of Nova Scotian women. Matern Child Health J 2014;18:1927e35.
1. Jolly MC, Sebire NJ, Harris JP, et al. Risk factors for macrosomia and its clinical consequences: a study of 350,311 pregnancies. Eur J Obstet Gynecol Reprod Biol 2003;111:9e14.
13. Sullivan LM, Massaro JM, D’Agostino RB. Presentation of multivariate data for clinical use: The Framingham Study risk score functions. Stat Med 2004;23:1631e60.
2. Mocanu EV, Greene RA, Byrne BM, et al. Obstetric and neonatal outcome of babies weighing more than 4.5 kg: an analysis by parity. Eur J Obstet Gynecol Reprod Biol 2000;92:229e33.
14. D’Agostino RB, Grundy S, Sullivan LM, et al. Validation of the Framingham coronary heart disease prediction scores: results of a multiple ethnic groups investigation. JAMA 2001;286:180e7.
3. Gaudet L, Ferraro ZM, Wen SW, et al. Maternal obesity and occurrence of fetal macrosomia: a systematic review and meta-analysis. BioMed Res Int 2014;2014:640291.
15. Thompson D, Berger H, Feig D, et al. Diabetes and pregnancy. Can J Diabetes 2013;37(1 Suppl):S168e83.
4. Legardeur H, Girard G, Journy N, et al. Factors predictive of macrosomia in pregnancies with a positive oral glucose challenge test: importance of fasting plasma glucose. Diabetes Metab 2014;40:43e8. 5. Perinatal Epidemiology Research Unit, Dalhousie University. Nova Scotia Atlee Perinatal Database Report of Indicators: 2002-2011 [Internet]. 2012 Dec [cited]. Available from, http://rcp.nshealth.ca/sites/default/files/ publications/nsapd_indicator_report_2002_2011.pdf. Accessed July 21, 2017. 6. Goetzinger KR, Odibo AO, Shanks AL, et al. Clinical accuracy of estimated fetal weight in term pregnancies in a teaching hospital. J Matern Fetal Neonatal Med 2014;27:89e93. 7. Scioscia M, Vimercati A, Ceci O, et al. Estimation of birth weight by twodimensional ultrasonography: a critical appraisal of its accuracy. Obstet Gynecol 2008;111:57e65. 8. Hartling L, Dryden DM, Guthrie A, et al. Benefits and harms of treating gestational diabetes mellitus: a systematic review and meta- analysis for the U.S. Preventive Services Task Force and the National Institutes of Health Office of Medical Applications of Research. Ann Intern Med 2013;159:123e9. 9. Boulvain M, Senat M, Perrotin F, et al. Induction of labour versus expectant management for large-for-date fetuses: a randomised controlled trial. Lancet 2015;385:2600e5.
8
l
-
JOGC
-
2017
16. Baschat AA, Harman CR, Farid G, et al. Very low second-trimester maternal serum alpha-fetoprotein: Association with high birth weight. Obstet Gynecol 2002;99:531e6. 17. Statistics Canada. 2001 Census: Proportion of visible minorities, census metropolitan areas, 2001, 1996 and 1991 [Internet]. [cited]. Available from: http://www12.statcan.ca/english/census01/products/analytic/ companion/etoimm/subprovs.cfm. Accessed July 21, 2017. 18. Practice Bulletin No. 173: Fetal Macrosomia. The American College of Obstetricians and Gynecologists. Obstet Gynecol 2016;128:e195e209. 19. Grote NK, Bridge JA, Gavin AR, et al. A meta-analysis of depression during pregnancy and the risk of preterm birth, low birth weight, and intrauterine growth restriction. Arch Gen Psychiatry 2010;67:1012e24. 20. Banderali G, Martelli A, Landi M, et al. Short and long term health effects of parental tobacco smoking during pregnancy and lactation: a descriptive review. J Transl Med 2015;13:327. 21. Clark JM, Hulme E, Devendrakumar V, et al. Effect of maternal asthma on birthweight and neonatal outcome in a British inner-city population. Paediatr Perinat Epidemiol 2007;21:154e62. 22. Motta M, Rodriguez-Perez C, Tincani A, et al. Neonates born from mothers with autoimmune disorders. Early Hum Dev 2009;85:S67e70. 23. Hosmer DW, Lemeshow S. Applied logistic regression. 2nd ed. New York: Wiley; 2000.