Preterm Birth and the Development of Visual Attention During the First 2 Years of Life

This systematic review and meta-analysis investigates visual attention functioning during the first 2 years of life in infants born preterm versus full term.


Study
Included attention measure/s Age in months GAPT (weeks) M (SD; Range)

BWPT (g) M (SD) Cohort Setting Exclusion criteria
Bonin et al, 1 1998 Novelty preference PD Ratio of looking at the novel stimulus to the looking time at the familiar and novel stimuli; stimuli: abstract patterns and naturalistic faces. Habituation PD Total looking time to reach criterion (infant-controlled); criterion: two consecutive fixations lasting 50% or less than the mean duration of the two longest fixations among the first three looks. Butcher et al, 2 2002 Latency to fixate PD Latency to fixate (RT) and correct frequency on refixations on a peripheral stimulus; fixation stimuli: either shifting abstract shapes or a phase-reversing schematic face appearing on either the left or the right side of a monitor; peripheral target was a flashing shape; off-line scoring via video recordings; only noncompetition trials considered. Cherkes-Julkowski, 3 1998

Focused attention
Total amount of engaged time with the object (based on Krakow & Kopp, 4 1983; "Sustained attention: engaged and occupied with toys"); seven min of free-play with age appropriate toys (last 6 min were coded).  6 2012 Latency to fixate Latency to first fixation (RT) on a peripheral stimulus (rectangles on a monitor); computerized eye-tracking; only noncompetition trials considered. 5.02 L (T1: 4.02; T2: 6.01) 30 Ortiz-Mantilla et al, 27 2008

Novelty preference
Ratio of looking at the novel stimulus to the looking time at the familiar and novel stimuli (only visual task considered); stimuli: naturalistic faces; simultaneous presentation of familiar and novel stimuli; gaze direction coded online using a video feed.

Habituation
Trials to reach criterion (infant-controlled); criterion: mean duration of looking for two consecutive trials < 50% than the mean duration for the first two trials.  The following coding guidelines were followed: Selection 1) Representativeness of the preterm cohort Truly or somewhat representative of the average characteristics of preterm-born infants in the community. Selected group of preterm-born infants (e.g., specific hospital/clinic); no description of the derivation of the cohort.
2) Selection of the full-term cohort Drawn from the same community as the preterm cohort. Drawn from a different source; no description of the derivation of the full-term cohort.

1) Comparable age
Study controls for corrected age.
No description of group ages / a statistic test for comparing group ages / age correction for prematurity.

2) Other comparable factors
Study controls for gender, socioeconomic status, ethnicity or parental education. Significant differences between the cohorts in these factors.
Outcome 1) Independent assessment Independent coders with an excellent inter-rater reliability statistic; computerized assessment. No description; non-excellent inter-rater reliability.

2) Blind assessment
Experimenters/coders were blind to the infant's group (preterm/full-term) or to the test condition. No description; experimenters/coders aware of the infant's group.

3) Data loss
Complete data from all subjects accounted for; subjects' data loss unlikely to introduce a biasless than 20% of the data, or if an adequate reason for data loss is provided.
More than 20% of the data lost and no description of those lost; no statement. Moderation analysis in the latency to fixate meta-analysis A moderation analysis was conducted to assess whether differences in latency to fixate between the preterm and term groups were associated with the cohort's birth era. The analysis indicated that the cohort's birth era classification was associated with the differences in latency to fixate between the preterm-term groups (QM = 6.38, df = 2, p = 0.041). This suggests that when compared to studies with infants born prior to the corticosteroids and surfactant era (β = -0.442, z = -2.34, p = 0.019), in studies conducted with cohorts that were born between 1990 and 2000 better outcomes were more likely to be observed (β = 0.598, z = 2.41, p = 0.016). Regarding cohorts from Era III, neither increased risk nor advantage were observed when compared to cohorts born at Era I (β = 0.231, z = 1.10, p = 0.27); however, increased risk for deficits was found compared to cohorts from Era II (β = -0.367, z = -1.98, p = 0.048). A possible explanation for this finding is the inclusion of two studies 17,30 that involved infants born extremely preterm in Era III, while no such preterm GA group was examined in studies from previous eras. Indeed, upon removal of these studies from the analysis no increased risk for deficits was found in preterm infants born in Era III, when compared to Era II (β = -0.282, z = -1.42, p = 0.15).

Study
A further inspection of studies involving cohorts from Era III, suggests that the differences between the preterm-term groups are still statistically significant in this era (d = -0.211; 95% CI, -0.39 to -0.03, z = -2.28, p = 0.023; k = 6), reflecting increased risk for deficits in latency to fixate ability following preterm birth. However, there are two factors that might explain this finding. First, as articulated above, Era III was the only Era including cohorts of infants born extremely preterm-a descriptive examination of the studies in Era III suggests the possibility of larger effect sizes in studies involving such populations (i.e., d = -0.53, 30 -0.32 17 ) when compared to studies involving very-to-late preterm born populations (i.e., d = 0.12, 16 -0.01, 14 , -0.18, 48 -0.42 6 ). Upon removal of these two studies from the analysis no significant differences were observed between the preterm-term groups (d = -0.126; 95% CI, -0.35 to 0.10, z = -1.11, p = 0.27; k = 4). A second factor that could explain the findings is advancements in gaze assessment technologies, involving the implementation of computerized gaze tracking systems that were not utilized in previous eras (half of the studies in Era III utilized such equipment). A descriptive examination of the effect sizes in studies from Era III suggests the possibility of larger effect sizes in studies utilizing computerized gaze tracking systems (i.e., d = -0.42, 6 -0.53, 30 and -0.32 17 ) rather than manual/videotaped assessments (i.e., d = 0.12, 16 -0.01, 14 and -0.18 48 ). Taken together, it could be conjectured that the differences between the preterm-term groups in Era III are at least partially related to the inclusion of extremely preterm cohorts and the implementation of more sensitive methods for gaze assessment. The possibility of assessing the interaction between these two factors and birth era classification was not viable, as no studies from Eras I and II included them.

eAppendix 4. Moderation analysis in the habituation meta-analysis
A moderation analysis was conducted to assess whether differences in habituation between the preterm and term groups were associated with the cohort's birth era.
The analysis indicated that when considering all three birth era groups, only a trend towards significance was found for the differences in effect sizes (QM = 4.98, df = 2, p = 0.083). However, when considering the difference between studies conducted before 2000 to all other studies a significant moderation effect was found (QM = 4.77, df = 1, p = 0.029; k = 13; see eFigure 4), suggesting increased risk for deficits in habituation in preterm infants born before 2000 (β = -0.186, z = -2.3, p = 0.016; d = -0.186, 95% CI, -0.34 to -0.03, p = 0.016; k = 10), and lesser likelihood for such deficits in infants born in Era III (β = 0.318, z = 2.18, p = 0.029; d = 0.132, 95% CI, -0.11 to 0.37, p = 0.29; k = 3). Further, this association remained significant even when the only study 27 involving a group of infants born extremely preterm was removed from the analysis (QM = 3.91, df = 1, p = 0.048; k = 12). eFigure 4. Graphical depiction of the moderation of the differences between the preterm and full-term groups in habituation by birth era eFigure 4. Each circle represents an effect size of a distinct study, conducted either in cohorts born before (k = 10; depicted in blue) or after (k = 3; depicted in green) the year 2000. Diamonds represent the standardized total mean differences between the term-preterm groups at each birth era. The asterisk represents significant difference between the groups in studies involving cohorts born before the year 2000; no significant difference was observed in studies from the current era. Circle sizes are proportional to studies' weight. * p < 0.05 eAppendix 5. Additional analyses in the focused attention meta-analysis Screening for outliers in the focused attention meta-analysis A strong evidence of heterogeneity in effect sizes between studies was found in the focused attention metaanalysis (Q17 = 36.46, p = 0.004; I 2 = 53.4%). Additional screening has pinpointed two potential outliers with studentized residuals of -2.53 50 and -3.45. 55 A possible explanation for the deviant effect size in the study conducted by Holly Ruff 50 (a pioneer in the research of focused attention and exploratory behavior in infants) could be related to a partial report of the outcome measures. In the study by Sun & Buys, 55 a possible explanation for the deviant effect size is that the study included a cohort of infants born extremely preterm (see the subsequent moderation analysis for a further evaluation). Removal of the two potential outliers, slightly attenuated the pooled effect size, but did not annulled it (d = -0.17; 95% CI, -0.28 to -0.05; z = -2.88, p = 0.004). However, these studies were responsible for the heterogeneity, as no heterogeneity was evident following their removal (Q15 = 10.4, p = 0.79; I 2 = 0.0%).
Moderation analysis assessing whether birth era and GA group are associated with focused attention outcomes In the focused attention meta-analysis, sufficient data were available to assess the interaction between cohort's birth era classification (i.e., birth before or after 2000; as no studies with cohorts born before 1990 included infants born extremely preterm, only a comparison between these two eras was applicable) and GA group 62 (i.e., extreme preterm [GA < 28 weeks], very preterm [GA of 28 to 32 weeks] or moderate/late preterm [GA of 32 to 37 weeks]). Classifications of the GA groups were based on the reported means and ranges for GA at birth.
The analysis indicated a significant interaction between cohort's birth era and GA group (QM = 11.6, df = 5, p = 0.04). To test whether birth after the year 2000 was associated with better focused attention performance in infants born extremely preterm, a post hoc analysis using a Dunnett's test (using the Bonferroni correction) with the before 2000-extreme preterm as the control group was conducted. The analysis indicated that infants born extremely preterm after 2000, had a lesser risk to show deficits in focused attention when compared to infants born extremely preterm before 2000 (β = 1.73, z = 2.87, p = 0.021). Further, in cohorts born before 2000, both the very-(β = 1.99, z = 3.52, p = 0.002) and moderate-late-preterm (β = 1.82, z = 3.11, p = 0.009) cohorts also had a lesser risk for deficits when compared to the extreme-preterm cohort that was born before 2000 (see eFigure 5 for a graphical depiction of the interaction).
Taken together the moderation analysis suggests that in cohorts born in Era III following extreme preterm birth, the risk for deficits in focused attention has attenuated-plausibly due to further advancements in neonatal care. eFigure 5. Graphical depiction of the moderation of the differences between the preterm and full-term groups in focused attention by birth era and GA group eFigure 5. Asterisks represent significant differences in effect sizes. ** p < 0.01 * p < 0.05