Study population

All consecutive patients who were treated for TTTS between December 2007 and May 2010 were considered eligible and were contacted for this study. TTTS was diagnosed at initial assessment at Los Angeles Fetal Therapy (University of Southern California) if the monochorionic-diamniotic multiple gestation had a maximum vertical pocket of fluid ≥8 cm in the recipient’s sac and ≤2 cm in the donor’s sac. Each case was classified prospectively according to the Quintero staging system. All patients were given the options of expectant treatment, pregnancy termination, amnioreduction, laser surgery, or selective reduction (at another center). Patients with stage I TTTS were informed of the controversy of undergoing laser surgery and were offered the option of expectant treatment with laser surgery only for disease progression. Patients were not offered laser surgery if preoperative ultrasound scans revealed gross abnormalities of intracranial anatomy. Cases were treated exclusively by SLPCV with or without sequential technique, as described in detail previously.

All consecutive laser-treated TTTS patients during the study period were contacted before the time their child was to reach 2 years old (± 6 weeks) corrected for gestational age and invited to participate by a study nurse, who was blinded to the predictors. Reasons that patients declined participation were recorded. All subjects were evaluated in the Southern California Clinical Translational Science Institute’s Clinical Trials Unit at Children’s Hospital Los Angeles, which is an institution independent and physically separated from Los Angeles Fetal Therapy. Families were given an incentive per child of $25 for their participation. There was no travel budget.

This study was approved by the institutional review board of the Health Sciences Campus of the University of Southern California and the Committee on Clinical Investigations at Children’s Hospital Los Angeles.

Measures

The Amiel-Tison Neurodevelopmental Examination was used to determine cerebral palsy by a board-certified developmental-behavioral pediatrician. This test is based on the clinical presence of gross and/or fine motor delay together with neurologic signs such as abnormal persistence of primitive reflexes, muscle tone, motor developmental progression, or coordination. Hearing and vision impairment was determined through the same examination or previous clinical report. Additionally, parent and child demographics, which included child height, weight, head circumference, and blood pressure, were collected at the 2-year-old visit.

A single board-certified neuropsychologist who was masked to the subjects’ clinical characteristics evaluated the cognitive and developmental status of participants with the Battelle Developmental Inventory, Second Edition (BDI-2). This tool was selected on the basis of availability of normative data, targeted age range, and the availability of standardized versions in both English and Spanish languages. A Spanish translator assisted the neuropsychologist with the administration of the BDI-2 as needed. Patients and their twin cohort (if applicable) were assessed consecutively on the same day at approximately 2 years of age (±6 weeks). Consistent with standardized administration of the BDI-2, each participant was assessed in the presence of 1 primary caretaker. In a few cases, the participant was unable to tolerate physical separation from their twin, who was then allowed to be present (playing separately) in the examination room.

The BDI-2 involves direct individual assessment and parental interview to measure key developmental skills in children from birth to 95 months (7 years, 11 months). Based on widely accepted developmental milestones for children, the BDI-2 assesses 5 developmental subdomains (personal-social, adaptive, motor, communication, and cognition) comprising overall development. The total BDI-2 developmental quotient score is computed as a sum of the 5 BDI-2 subdomains, and it has a mean of 100 with standard deviation of 15. Individual item scores range from 0-2 points, with scores based on 1 of 3 predetermined criteria that include observation, parent interview, and/or performance on a structured task.

Neurodevelopmental impairment (NDI) was defined as having bilateral blindness (unable to fix on or track an object), bilateral deafness (requiring amplification), cerebral palsy (based on physical examination), and/or a BDI-2 total developmental quotient of <70 (standardized score), in accordance with previous TTTS literature.

Statistical analyses

Descriptive statistics and bivariate comparisons were produced with IBM SPSS Statistics software (version 19; SPSS Inc, Chicago, IL). Bivariate comparisons were also calculated between participating and not participating families and between donors and recipients (eg, t tests for continuous measures, chi-square tests for categoric variables). Multilevel linear regression analysis was conducted in Mplus software (version 6; Muthén & Muthén, Los Angeles, CA), with twins grouped or nested within pregnancy and missing data were subject to listwise deletion in the regression model. Unlike logistic regression, which describes the relationship between the presence or level of a predictor and the odds of obtaining a yes or 1 on the binary outcome variable, linear regression coefficients quantify the relationship between a predictor and the expected level or amount of a continuous outcome. Specifically, a standardized linear regression coefficient denotes the expected change in the (standardized) outcome for 1 standard unit change in the predictor. The multilevel approach extends the general linear regression model by accounting for the genetic and environmental experience that is shared by twins of the same family (the shared or pregnancy level) and the experiences that are unique to each child (the individual or child level). Multilevel regression allows for simultaneous inclusion of the effects of both child-level and pregnancy-level risk factors of cognitive performance; further, it supports the inclusion of families with 2 (twin) children and families with only 1 surviving child.

Because of the small number of children with NDI, the total BDI-2 developmental quotient score was considered the primary outcome for analysis. Child-level risk factors for development that were tested in the regression models included donor status (donor = 1; recipient = 0), male sex (male = 1; female = 0), height (centimeters), weight (kilograms), head circumference (centimeters), systolic and diastolic blood pressure (millimeters of mercury), and heart rate (beats per minute). Pregnancy-level risk factors included gestational age at surgery and birth in weeks, the twins’ current corrected age in years, Quintero stage, presence (1) or absence (0) of fetal growth restriction, fetal death, and complications at delivery, maternal and paternal age in years, maternal and paternal education (some college or more = 1; high school or less = 0), number of adults and children in the home, parents’ marital status (married = 1; other = 0), minority race/ethnicity (minority = 1; white = 0), and whether the twins’ mother or father smokes tobacco (yes = 1; no = 0). Although all potential risk factors for poor cognitive performance initially were incorporated into the adjusted model, it was expected that some demographic covariates would not be related to the outcome. Thus, nonsignificant predictors were removed systematically to arrive at the most parsimonious possible final model. The study was powered on the basis of the assumption of a relatively parsimonious random effects regression model with at most 10 predictors at either level. In such a model with alpha level of .05 and adequate power of 0.80, a priori power analysis determined that a total sample of 100 children was required to detect small-to-moderate regression effects.