A flow chart of SVR trial subjects from randomization to neurodevelopmental follow-up is shown in . The follow-up rate for the BSID-II examination among transplant-free survivors was 86%. The mean (±SD) age at follow-up was 14.3±1.1 months (range 12.2–19.5 months). The 314 patients who completed the examination were less likely to be of black or other race (P<.001). However, the groups did not differ in their socioeconomic class or level of highest maternal education. The mean time interval between the Norwood and stage II procedures was 5.2±1.9 months (median 5.0 months). Ten (3%) subjects in the cohort were not discharged between stage I and II surgery.
Flow chart of SVR trial subjects from randomization to neurodevelopmental follow-up.
At 14 months, children had impaired growth: mean weight, height, and head circumference z-scores-for-age were −0.7±1.1, −1.3±1.7, and −0.4±1.4, respectively. Among the 296 subjects for whom a parent history was available, 47% had received developmental support in the first year of life: 39% with physical therapy, 26% with speech/language therapy, 10% with early intervention, and 4% with other forms of support. Genetic syndromes or other anomalies were detected in 25% of the cohort and were absent in 57%; the remaining 18% were not evaluated by a geneticist. Subjects with confirmed genetic syndrome and other anomalies had longer hospitalization (median 28 days [IQR 20–43 days] vs. 22 days [IQR 16 – 35 days] and 24 days [IQR 17–38 days] for no and unknown syndrome status, P=.01). Birth weight tended to be lower in the patients with confirmed genetic syndromes (3.0±0.6 vs. 3.2±0.5 or 3.2±0.5 kg in the no and unknown groups, P=.06).
Distributions of PDI and MDI scores are depicted in . and summarize the significant univariate associations (P≤.05) of PDI or MDI score with patient factors, management practices, and medical course.
Histogram depicting the number of subjects according to scores on the Psychomotor Development Index (PDI, left panel) and Mental Development Index (MDI, right panel) of the Bayley Scales of Infant Development®—Second Edition.
Univariate Regressions using Predictors from Birth to Norwood Preoperative Period. Factors with P<.05 for either the PDI or MDI score shown.
Univariate Regressions Using Predictors from Norwood Operation to 14 months. Factors with P<.05 for either the PDI or MDI score shown.
For the overall cohort, PDI scores were profoundly lower than in the normative population (): scores were < 85 (>1 SD below expected mean) in 65% of subjects, and <70 (>2 SD below expected mean) in 44% of subjects. The subgroup of subjects with no genetic syndrome and birth weight of ≥2.5 kg had the highest PDI and MDI scores (each P<.001) and also had shorter Norwood hospital length of stay (P=.033). Subjects who received the MBT shunt and RV-to-PA shunt had similar PDI scores (75±19 vs. 74±20, respectively, P=0.48; Supplemental Figure
). In multivariable linear regression modeling (), independent predictors of lower PDI scores were clinical center where the Norwood procedure was performed (P=.003), birth weight <2.5 kg (P=.02), longer log-transformed days of Norwood hospitalization (P<.001), and greater number of complications between the time of discharge after the Norwood procedure and age 12 months (P<.001). The percentage of variance explained by the model (adjusted R2
) was 26%. Mean PDI score dropped approximately 13 points for every three extra days of hospital stay during the Norwood hospitalization. Presence of a genetic syndrome or other anomaly did not achieve statistical significance (P=.07).
Multivariable Regression Models for Bayley Summary Scores
MDI scores were also much lower than in the normative population (): scores were < 85 (>1 SD below normal mean) in 36% of subjects, and <70 (>2 SD below normal mean) in 16% of subjects. Within subject, MDI scores were higher than PDI scores (P<.001). Mean MDI scores did not differ significantly between patients who received the MBT shunt vs. the RV-to-PA shunt (89±17 vs. 88±18, respectively, P=.55; eFigure 1
). In multivariable regression analysis (), independent predictors of lower MDI score included the clinical center where the Norwood procedure was performed (P<.001), birth weight <2.5 kg (P=.04), the presence of a genetic syndrome or other anomalies (P=.04), lower maternal education level (P=.04), longer log days on the ventilator after the Norwood procedure (P<.001), and a greater number of complications between the time of hospital discharge after the Norwood procedure and age 12 months (P<.001). The percentage of variance explained by the model (adjusted R2
) was 34%. Birth weight percentile for gestational age was a less significant predictor for PDI and MDI scores than raw birth weight.
Early intervention in the first year of life was not included in our original multivariable models, because it was considered to be a correlate of adverse outcome. However, the percentage of subjects who received early intervention varied significantly by center, ranging from 0% to 23% (p=0.02) for early intervention administered for the indication of a cognitive disorder and from 0% to 78% for receipt of any form of early intervention (e.g., physical therapy, occupational therapy). We therefore explored whether center differences in the percentages of children receiving either early intervention for a cognitive disorder or any form of early intervention in the first year of life could explain the center effect on outcomes. Mean PDI and MDI scores were significantly lower on average by 11 points in those who received early intervention for a cognitive disorder compared to those who did not (PDI, 66.4±18.0 vs. 75.5±19.3, P=.014 ; MDI, 78.7±20.7 vs. 90.1±17.0, P<.001). Similarly, children who received any form of early intervention in the first year of life (e.g., physical therapy, occupational therapy) fared worse than those who received no form of early intervention (PDI, 67.8±18.3 vs. 81.5±17.9, P<.001 ; MDI, 84.1±19.0 vs. 93.9±14.7, P<.001). However, adjustment for either of these intervention terms in the final multivariable models for MDI and PDI did not appreciably alter the effects of center on developmental scores. Similarly, we could find no other center characteristics, such as center volume or surgeon volume, which had a significant effect on PDI or MDI score in multivariable analysis. Finally, we explored whether the center effect could be related to an outlier; when eliminating centers one at a time from the multivariable models; study inferences were similar.
Some variables that had been hypothesized to predict developmental outcome were notable for their lack of association with either PDI or MDI scores. Perfusion type examined both as a categorical variable (DHCA, RCP, or DHCA together with RCP), and as total minutes of DHCA during Norwood surgery was not associated with PDI score in univariate analyses. MDI scores differed (P<.001) in univariate analysis among patients who underwent vital organ support during the Norwood procedure using DHCA, RCP, or a combination of DHCA and RCP (mean scores 85.1, 93.4, and 92.0, respectively, P<.001). As noted above, perfusion type was not an independent predictor of MDI in multivariable analysis. Because the final multivariable model included clinical center, we further explored whether perfusion type would become an independent predictor of MDI if site were not considered. Even then, perfusion strategy did not meet the criteria for inclusion because of low reliability by bootstrapping (40%). In addition, we found no significant relationship of PDI or MDI score to other aspects of vital organ support, such as hematocrit or pH strategy during core cooling or to patient factors such as diameter of the ascending aorta or cardiac anatomy.
Finally, we assessed whether the strength of the candidate predictors varied according to four pre-specified patient factors (see Methods) related to birth weight, preterm status, pre-Norwood head circumference z-score, and presence of a genetic syndrome or other anomalies. The lack of effect of shunt type on PDI and MDI scores was consistent across the predetermined subgroups. The association of other candidate predictors with outcome also did not vary according to these pre-specified patient factors. None of the interaction terms entered the final multivariable models for either MDI or PDI.