Objective
Many obstetric interventions are performed to improve long-term neonatal outcome. However, long-term neonatal outcome is usually not a primary outcome because it is time-consuming and expensive. The aim of this project was to identify different perinatal risk indicators and to develop prediction models for neurologic morbidity at 2 and 5 years of age.
Study design
Data from a Dutch cohort study of preterm and small-for-gestational-age infants was used. Neonates who were born in The Netherlands in 1983 with a gestational age of <34 weeks and without congenital abnormalities were included (n = 753). Infants were divided in 3 groups: no handicap, minor handicap, and major handicap.
Results
Common risk indicators for major handicaps at 2 and 5 years of age were male sex (odds ratio, 2.7 and 3.0, respectively), seizures after ≥2 days of life (odds ratio, 5.8 and 5.8, respectively), and intracranial hemorrhage (odds ratio, 3.8 and 2.6, respectively).
Conclusion
In this cohort, male sex, intracranial hemorrhage, and seizures seem to be important risk indicators for long-term neurologic morbidity.
Many obstetric interventions are performed to improve both short- and long-term outcome. Evaluation of the long-term effect of a perinatal intervention is necessary because serious sequelae from perinatal complications frequently manifest themselves only after several years. Nevertheless, long-term follow-up evaluation is time-consuming, expensive, beyond obstetricians’ awareness, and falls outside the funding period of most obstetric studies. Consequently, obstetric interventions usually are not evaluated for their long-term outcomes, and short-term outcomes are selected as the primary endpoint of an obstetric study.
One way to overcome this problem would be to model long-term consequences on the basis of short-term neonatal outcomes. This could be realized by the development of prediction models in which the association between short-term and long-term outcomes is determined statistically and adjusted for relevant covariates.
Subsequently, these prediction models for long-term neurologic morbidity could be used to extrapolate short-term outcomes on the neurologic status of neonates or to indicate for which neonates neurologic long-term follow-up evaluation is required, as their outcomes (either absence or presence of sequelae) cannot be predicted from short-term outcomes and clinical background characteristics. The development of such models requires a longitudinal approach in which data surrounding pregnancy, delivery, and short-term outcomes and follow-up data are available on various health-related outcomes.
The Dutch project on preterm and small-for-gestational-age infants (POPS) cohort is one of the few birth cohorts with a systematic assessment of these data. Data of all Dutch infants who were born alive in 1983 with a gestational age of <32 completed weeks and/or with a birthweight of <1500 g were collected prospectively. This birth cohort could provide insight in the long-term consequences of perinatal outcomes.
In the literature, many risk indicators for neurologic morbidity are mentioned. Birth catastrophes such as placental abruption, cord prolapse, and uterine rupture sharply increase the risk for neurologic morbidity, but these conditions fortunately are uncommon and even sometimes not survived; individually and collectively, these indicators account for only a small portion of neurologic morbidity. Although any other indicator, if severe, may be sufficient to cause neurologic morbidity, more often it is the presence of multiple risk indicators that causes neurologic morbidity later in life. Development of multivariable prediction models for neurologic morbidity can increase our understanding of predictors for neurologic morbidity and can help us to develop interventions to prevent these complications in the future.
In this study, we aimed to identify different perinatal risk indicators for long-term neurologic morbidity and to use these perinatal risk indicators to develop prediction models for long-term neurologic morbidity at 2 and 5 years of age.
Materials and Methods
Study design
For the development of prediction models for long-term neurologic morbidity, we used data that were available from a Dutch cohort study of preterm and/or small-for-gestational-age infants (POPS study). In this cohort, all of the live born infants who were included were delivered in The Netherlands between January and December 1983, either at <32 completed weeks of gestation and/or with a birthweight of <1500 g. The study ultimately consisted of 1338 infants, which was 94% of the eligible infants who had been born in 1983 in The Netherlands. Because of the “mixed metaphor” of combining gestational age and low birthweight in this cohort, only infants with gestational age of <34 weeks were included in our analysis. Infants with congenital abnormalities were excluded.
Outcomes
Endpoints that were used for this prediction model were neurologic morbidity at 2 and 5 years of age. The follow-up evaluation until the age of 2 years was carried out by local pediatricians all over The Netherlands. An overall developmental level was done with the Gesell test that had been adapted for Dutch children and also neurologic, visual, and hearing examinations had been performed.
According to the outcome, the data were divided into 3 groups: no handicap, minor handicap, and major handicap. The infant was considered to have no handicap when developmental delay was absent (developmental quotient >90) and there were no motor, visual, or hearing disabilities. A minor handicap was diagnosed when some delay was present (3-4 months retarded or developmental quotient between 80 and 90) and/or at least one of the following handicaps: a mild cerebral paresis (such a slight hemiparesis or quadriparesis), mild visual or hearing defects, or moderate psychosocial problems. Such disabilities were unlikely to prevent the child from going to a normal school or to interfere seriously with normal life. A major handicap was diagnosed when severe retardation was present (≥5 months delay or developmental quotient <80) and/or at least one of the following handicaps: a severe cerebral paresis, severe visual or hearing defects, or serious psychologic problems. Such disabilities probably would stop the child from going to a normal school or cause serious interference with normal functioning in society.
At 5 years chronologic age, a follow-up program was carried out by 3 specially trained pediatricians during a visit to the home. Eight areas of development were assessed: neuromotor function (Touwen ); mental development (Denver developmental screening test) ; hearing function (audiometry/otoscopy); visual function; language and speech development (Standardized Dutch Test; Gerritsen ); musculoskeletal system (physical examination) and respiratory morbidity (parents’ questionnaire). In each area, an infant was categorized as impaired, disabled, or handicapped, according to World Health Organization definitions. An infant was regarded as handicapped at 5 years of age if he or she had a handicap in an area of examination. Infants who needed special education as a result of ≥1 impairments or disabilities were considered to be at least minor handicapped. A handicap was considered minor if it did not interfere seriously with everyday life and did not require extensive caretaking and major when it did interfere with everyday life and when it led to a life of dependency or institutionalization.
Candidate predictors
Candidate predictors for neurodevelopment handicaps were determined on the basis of existing literature of perinatal predictors for long-term neurologic morbidity, combined with consulting experts in the field. The following candidate predictors were included in the analysis: social class, ethnicity, education level of the mother (low, moderate, high), maternal smoking, hypertension before pregnancy, pregnancy-induced hypertension (diastolic pressure >90 mm Hg), preeclampsia/eclampsia, maternal epilepsy, diabetes mellitus, gestational diabetes mellitus, multiple pregnancy, vertex or other presentation, prolonged rupture of membranes, meconium-stained fluid, glucocorticosteroids, small for gestational age (<10th percentile), gestational age, sex, neonatal asphyxia, respiratory distress syndrome, bronchopulmonary dysplasia, seizures, intracranial hemorrhage, necrotizing enterocolitis, hyperbilirubinemia, sepsis (blood culture proven), and duration of mechanical ventilation (continuous or intermittent).
Neonatal asphyxia was defined as low 5-minute Apgar score (<7) and/or umbilical cord acidosis (pH <7.05). Bronchopulmonary dysplasia was defined as clinical signs of respiratory distress, with an abnormal chest X-ray and an oxygen requirement after 28 days of age (criteria of Bancalari et al ). Intracranial hemorrhage was defined as a clinical diagnosis (based on rapid or salutatory deterioration, fall in hematocrit level) and/or ultrasound scans or computed tomography. All seizures (clinical definition: including subtle seizures, generalized tonic, multifocal clonic, focal clonic, and myoclonic seizures) were recorded as either absent or as present on day 1 of life or day 2 of life or later.
Statistical analysis
We developed 4 multivariable logistic regression models in which we analyzed the association between the candidate predictors and infants with minor or major handicap vs infants with no handicap and infants with major handicap vs infants with no or minor handicap at 2 and 5 years of age. Multiple imputations were used to adjust for missing values. We created 5 imputed datasets that were based on the candidate predictors mentioned earlier and all available outcome-specific data at 2 and 5 years of age. Imputed values were limited to the lowest and highest values that were observed for the measured outcome variable. Uncertainty about imputed values is reflected in differences between different imputed datasets and incorporated in the estimated standard errors and associated probability values for the pooled model. We used SPSS software (version 17.0; SPSS Inc, Chicago, IL) for the imputation. The imputation method in SPSS software is based largely on the chained equations approach in multivariate imputation by chained equations (MICE).
After imputation, the prevalence of the candidate predictors was first analyzed. Thereafter, a univariable and multivariable regression analysis was performed to estimate odds ratios (ORs), 95% confidence interval [CI], and corresponding probability values for dichotomous and continuous variables. Because the use of too stringent probability values for variable selection is more deleterious for a model than including too many factors, all variables that showed a significance level of < .50 in univariable analyses were entered in the multivariable logistic regression model. Furthermore, we used a stepwise backward selection procedure with a predefined significance level of < .20 for removing variables from the models. Variables that remained in the last step of the backward selection procedure in at least 4 of the 5 imputed datasets were included in the final logistic regression analysis. Discriminative capacity of the models was evaluated by calculation of the area under the curve. Calibration of the models was assessed by comparison of the calculated probabilities with the observed proportion of neurologic morbidity. The goodness-of-fit was tested formally with the Hosmer and Lemeshow test statistic. Data were analyzed with the SPSS software.
Results
Sample and respiratory morbidity incidence
Of the original cohort of 1338 infants, 1026 infants survived the neonatal period (>28 days); 969 infants were alive at 2 years of age; 966 infants were alive at 5 years of age, and 959 infants were alive at 19 years of age. The risk of death in the first 28 days of life was equal for boys and girls. Because of the “mixed metaphor” of the combination of gestational age and low birthweight in this cohort, infants with a gestational age of ≥34 weeks were excluded (n = 136). Because congenital malformations were considered to influence neurologic function, all infants with congenital abnormalities were also excluded (n = 70), which left 753 infants for the final analysis. At 2 years of age, information on neurologic morbidity was missing for 23 infants (follow-up rate, 97%). At 5 years of age, information on neurologic morbidity was missing for 33 infants (follow-up rate, 96%). At 2 years of age, the rate of infants with no handicap, minor handicap, or major handicap was 83.2% (n = 607 infants), 11.5% (n = 84 infants), and 5.3% (n = 39 infants), respectively, before imputation and 81.5% (n = 614 infants), 11.7% (n = 88 infants), and 6.8% (n = 51 infants), respectively, after imputation. At 5 years of age, the rate of infants with no handicap, a minor handicap, or a major handicap was 86.0% (n = 619 infants), 8.3% (n = 60 infants), and 5.7% (n = 41 infants), respectively, before imputation and 84.5% (n = 636 infants), 9.4% (n = 71 infants), and 6.1% (n = 46 infants), respectively, after imputation.
Univariable and multivariable models
Neurologic morbidity at 2 years of age
Tables 1 and 2 show the results of the univariable and multivariable regression analysis for neurologic morbidity at 2 years of age. Male sex (adjusted OR [aOR], 1.6; 95% CI, 1.1–2.4) and intracranial hemorrhage that was diagnosed with ultrasound scanning or computed tomography (aOR, 2.3; 95% CI, 1.2–4.3) were significant risk indicators for minor/major handicaps at 2 years of age ( Table 1 ). Risk indicators for major handicaps only were male sex (aOR, 2.7; 95% CI, 1.2–5.8), seizures at ≥2 days of life (aOR, 5.8; 95% CI, 1.9–17.8), intracranial hemorrhage that was diagnosed with ultrasound scanning or computed tomography (aOR, 3.8; 95% CI, 1.6–9.1) and hyperbilirubinemia (aOR, 2.6; 95% CI, 1.2–5.3). Surprisingly, maternal smoking (1-10 cig/d) seemed to decrease the risk for major handicaps (aOR, 0.32; 95% CI, 0.12–0.88) ( Table 2 ).
| Candidate predictors | No. of children (pooled) | Univariable analysis (pooled) | Multivariable analysis (pooled) | ||
|---|---|---|---|---|---|
| Crude odds ratio (95% CI) | P value | Adjusted odds ratio (95% CI) | P value | ||
| Environmental factors | |||||
| Ethnicity | |||||
| Both parents white | 637 (85%) | 1.2 (0.68–2.2) | .489 | ||
| One/both parents Mediterranean | 41 (5%) | 1.1 (0.41–3.0) | .824 | ||
| One/both parents black | 31 (4%) | 0.54 (0.16–1.8) | .322 | ||
| One/both parents Asian | 39 (5%) | 0.66 (0.24–1.8) | .421 | ||
| Other | 9 (1%) | ||||
| Social class | |||||
| Low | 295 (39%) | 1.0 | |||
| Moderate | 275 (37%) | 0.76 (0.46–1.2) | .274 | ||
| High | 183 (24%) | 0.71 (0.40–1.3) | .256 | ||
| Education mother | |||||
| Low | 423 (56%) | 0.84 (0.48–1.5) | .531 | ||
| Moderate | 131 (17%) | 0.88 (0.54–1.4) | .606 | ||
| High | 199 (26%) | ||||
| Maternal smoking during pregnancy per day | |||||
| No | 493 (65%) | 1.0 | |||
| 1-10 | 147 (20%) | 1.05 (0.64–1.7) | .860 | ||
| ≥10 | 112 (15%) | 1.4 (0.82–2.4) | .220 | ||
| Hypertension before pregnancy | 34 (5%) | 0.74 (0.28–2.0) | .548 | ||
| Epilepsy | 4 (1%) | 1.8 (0.15–20.7) | .645 | ||
| Obstetric | |||||
| Multiple pregnancy | 172 (23%) | 1.0 (0.64–1.6) | .995 | ||
| Corticosteroids | 131 (17%) | 1.2 (0.72–1.9) | .543 | ||
| Gestational diabetes mellitus | |||||
| No | 715 (95%) | 1.0 | |||
| With diet | 22 (3%) | 0.43 (0.1–1.9) | .258 | ||
| With insulin | 16 (2%) | 0.97 (0.27–3.5) | .964 | ||
| Hypertension during pregnancy | |||||
| No | 583 (77%) | 1.0 | |||
| ≥90 mm Hg | 110 (15%) | 0.73 (0.41–1.3) | .281 | ||
| Preeclampsia/eclampsia | 60 (8%) | 0.68 (0.32–1.5) | |||
| Prolonged rupture of membranes | |||||
| No | 440 (58%) | 1.0 | |||
| <1-11 h | 127 (17%) | 1.2 (0.69–1.9) | .611 | ||
| 12-24 h | 28 (4%) | 1.4 (0.50–3.8) | .544 | ||
| 1-7 D | 106 (14%) | 1.05 (0.60–1.9) | .855 | ||
| >7 D | 53 (7%) | 1.5 (0.74–2.9) | .271 | ||
| Meconium stained fluid | 41 (5%) | 1.3 (0.57–3.0) | .524 | ||
| Presentation: other than vertex | 231 (31%) | 0.88 (0.58–1.3) | .550 | ||
| Neonatal | |||||
| Gestational age, wk | 0.93 (0.83–1.03) | .168 | |||
| 25-28 | 108 (14%) | ||||
| 28-30 | 219 (29%) | ||||
| 30-32 | 320 (43%) | ||||
| 32-34 | 106 (14%) | ||||
| Low birthweight (<10th percentile) | 197 (26%) | 0.95 (0.62–1.5) | .829 | ||
| Male sex | 396 (53%) | 1.7 (1.2–2.5) | .006 | 1.6 (1.1–2.4) | .014 |
| Asphyxia | 71 (9%) | 2.3 (1.2–4.3) | .016 | 1.8 (0.92–3.6) | .094 |
| Bronchopulmonary dysplasia | 112 (15%) | 1.9 (1.1–3.2) | .020 | ||
| Respiratory distress syndrome | |||||
| No | 421 (56%) | 1.0 | |||
| Clinical | 111 (15%) | 1.02 (0.58–1.8) | .935 | ||
| Radiographic | 221 (29%) | 1.3 (0.87–2.0) | .192 | ||
| Pneumothorax | 50 (7%) | 1.1 (0.53–2.4) | .736 | ||
| Seizures | |||||
| No | 724 (96%) | 1.0 | 1.0 | ||
| First d | 4 (1%) | 2.3 (0.19–27.5) | .520 | 2.6 (0.20–34.4) | .469 |
| ≥2 d | 26 (3%) | 2.8 (1.2–6.4) | .018 | 2.1 (0.83–5.3) | .120 |
| Intracranial hemorrhages | |||||
| No | 653 (87%) | 1.0 | 1.0 | ||
| Suspect | 41 (5%) | 1.5 (0.71–3.3) | .279 | 0.94 (0.41–2.1) | .874 |
| Proven | 58 (8%) | 2.9 (1.6–5.3) | .000 | 2.3 (1.2–4.3) | .009 |
| Necrotizing enterocolitis | 42 (6%) | 2.1 (1.02–4.4) | .046 | 2.1 (0.95–4.5) | .069 |
| Hyperbilirubinemia ≥200 μmol/L | 212 (28%) | 1.5 (1.01–2.4) | .047 | 1.5 (0.97–2.4) | .071 |
| Sepsis (culture proven) | 78 (10%) | 1.5 (0.83–2.6) | .192 | 1.5 (0.80–2.7) | .221 |
| Continuous positive airway pressure, d | mean = 2 days | 1.03 (0.99–1.1) | .114 | ||
| Artificial ventilation, d | mean = 3 days | 1.04 (1.01–1.1) | .004 | 1.02 (0.99–1.05) | .125 |
Stay updated, free articles. Join our Telegram channel
Full access? Get Clinical Tree
