Article type
Year
Abstract
Introduction:
Objectives: There is a potential for selection bias in preterm infant cohort studies that only report survival in infants admitted to neonatal units and in those that include livebirths but exclude stillbirth data. The objective of this systematic review was to determine if this potential selection bias leads to an overestimate of survival.
Methods: A systematic review was performed, aiming to identify all published studies reporting gestational age specific survival of infants <28 weeks gestation born 1978-98 in developed countries. Each was graded according to the definition of the cohort, namely: A - stillbirths & livebirths, B - livebirths, C - neonatal unit admissions. Proportions of infants surviving the neonatal period were calculated for each week of gestation, overall estimates of survival were calculated by combining proportions from different cohorts in a weighted average. A Chi-squared analysis was performed in order to detect a gradient in qualitatively ordered proportions (Bartholomew 1959).
Results: 191 reports and abstracts were reviewed. 31 (16%) were not appraised (non-English language), 56 (29%) reported data relating to birthweight only, 37 (19%) did not report data in one week gestation age bands. The remaining 67 studies reported data regarding 57 cohorts. The following table gives the estimates of survival grouped according to grade of study. There was no evidence that other factors (birth time period and usage of antenatal steroid and postnatal surfactant therapy) were confounding the observed gradient in survival.
Survival estimates grouped by study grade / %(95%CI) Analysis
Gestation
/weeks A (16 cohorts) B (23 cohorts) C (18 cohorts)
Stillbirths + livebirths Livebirths NICU admissions X ^3 p
23 11 (8-14) 18 (14-21) 23 (17-2B) 6.3 <0.01
24 30 (26-33) 33 (30-36) 50 (45-54) 58.9 <0.01
25 49 (45-52) 50 (48-53) 60 (56-64) 22.0 <0.01
26 65 (60-69) 63 (61-65) 70 (67-74) 11.8 <0.01
27 74 (/1-/8) 72 (69-74) 75 (72-78) 2.40 >0.05
Discussion: Professionals responsible for provision or perinatal services need to be aware of the effect of selection bias has in overestimating survival of infants 23-26 weeks gestation when appraising cohort studies.
Objectives: There is a potential for selection bias in preterm infant cohort studies that only report survival in infants admitted to neonatal units and in those that include livebirths but exclude stillbirth data. The objective of this systematic review was to determine if this potential selection bias leads to an overestimate of survival.
Methods: A systematic review was performed, aiming to identify all published studies reporting gestational age specific survival of infants <28 weeks gestation born 1978-98 in developed countries. Each was graded according to the definition of the cohort, namely: A - stillbirths & livebirths, B - livebirths, C - neonatal unit admissions. Proportions of infants surviving the neonatal period were calculated for each week of gestation, overall estimates of survival were calculated by combining proportions from different cohorts in a weighted average. A Chi-squared analysis was performed in order to detect a gradient in qualitatively ordered proportions (Bartholomew 1959).
Results: 191 reports and abstracts were reviewed. 31 (16%) were not appraised (non-English language), 56 (29%) reported data relating to birthweight only, 37 (19%) did not report data in one week gestation age bands. The remaining 67 studies reported data regarding 57 cohorts. The following table gives the estimates of survival grouped according to grade of study. There was no evidence that other factors (birth time period and usage of antenatal steroid and postnatal surfactant therapy) were confounding the observed gradient in survival.
Survival estimates grouped by study grade / %(95%CI) Analysis
Gestation
/weeks A (16 cohorts) B (23 cohorts) C (18 cohorts)
Stillbirths + livebirths Livebirths NICU admissions X ^3 p
23 11 (8-14) 18 (14-21) 23 (17-2B) 6.3 <0.01
24 30 (26-33) 33 (30-36) 50 (45-54) 58.9 <0.01
25 49 (45-52) 50 (48-53) 60 (56-64) 22.0 <0.01
26 65 (60-69) 63 (61-65) 70 (67-74) 11.8 <0.01
27 74 (/1-/8) 72 (69-74) 75 (72-78) 2.40 >0.05
Discussion: Professionals responsible for provision or perinatal services need to be aware of the effect of selection bias has in overestimating survival of infants 23-26 weeks gestation when appraising cohort studies.