Waisbren SE, Hanley W, Levy HL, Shifrin H, Allred E, Azen C, Chang P, Cipcic-Schmidt S, de la Cruz F, Hall R, Matalon R, Nanson J, Rouse B, Trefz F, Koch R. Outcome at Age 4 Years in Offspring of Women With Maternal PhenylketonuriaThe Maternal PKU Collaborative Study. JAMA. 2000;283(6):756-762. doi:10.1001/jama.283.6.756
Author Affiliations: Children's Hospital, Boston, Mass (Drs Waisbren and Levy and Ms Allred); Hospital for Sick Children, Toronto, Ontario (Dr Hanley); National Institute of Child Health and Human Development, Bethesda, Md (Mr Shifrin and Dr de la Cruz); Children's Hospital, Los Angeles, Calif (Drs Azen and Koch and Ms Hall); Children's Hospital, Minneapolis, Minn (Dr Chang); University of Tuebingen, Tuebingen, Germany (Ms Cipcic-Schmidt); University of Texas Medical Branch,Galveston (Drs Matalon and Rouse); Royal University Hospital, Saskatoon, Saskatchewan (Dr Nanson); and Children's Hospital of Reutlingen, Reutlingen, Germany (Dr Trefz).
Context Untreated maternal phenylketonuria (PKU) increases risk for developmental
problems in offspring. The extent to which this risk is reduced by maternal
dietary therapy at various stages of pregnancy is not known.
Objective To determine whether dietary treatment during pregnancy of women with
PKU affects developmental outcomes of offspring.
Design The Maternal PKU Collaborative Study, an ongoing, longitudinal prospective
study begun in 1984.
Setting A total of 78 metabolic clinics and obstetrical offices in the United
States, Canada, and Germany.
Participants A total of 253 children of women with PKU (n=149), with untreated mild
hyperphenylalaninemia (n=33), or without known metabolic problems (comparison
group; n=71) were followed up to age 4 years.
Intervention Women with PKU were offered a low-phenylalanine diet prior to or during
pregnancy with the aim of maintaining metabolic control (plasma phenylalanine ≤10
mg/dL [≤605 µmol/L]). Women with mild hyperphenylalaninemia, who
had plasma phenylalanine levels of no more than 10 mg/dL (605 µmol/L)
on a normal diet, were not treated.
Main Outcome Measures Children's scores on cognitive and behavioral assessments (McCarthy
Scales of Children's Abilities, Test of Language Development, Achenbach Child
Behavior Checklist, Vineland Adaptive Behavior Scales, and Home Observation
for Measurement of the Environment), compared by maternal metabolic status
at 0 to 10 weeks', 10 to 20 weeks', and after 20 weeks' gestation.
Results Scores on the McCarthy General Cognitive Index decreased as weeks to
metabolic control increased (r = −0.58; P<.001). Offspring of women who had metabolic control
prior to pregnancy had a mean (SD) score of 99 (13). Forty-seven percent of
offspring whose mothers did not have metabolic control by 20 weeks' gestation
had a General Cognitive Index score 2 SDs below the norm. Overall, 30% of
children born to mothers with PKU had social and behavioral problems.
Conclusions Our data suggest that delayed development in offspring of women with
PKU is associated with lack of maternal metabolic control prior to or early
in pregnancy. Treatment at any time during pregnancy may reduce the severity
Phenylketonuria (PKU) in women increases the risk to their offspring
for adverse outcomes. In the offspring of untreated women with maternal PKU,
there is a 92% risk for mental retardation, 73% risk for microcephaly, 40%
risk for low birth weight, and 12% risk for congenital heart disease.1 The precise mechanism of fetal damage in maternal
PKU is still unknown, although it is clear that the abnormal intrauterine
environment harms the fetus.2- 4
Treatment with a phenylalanine-restricted diet begun prior to pregnancy reduces
these risks.5 However, the extent to which
risks are reduced in late or inadequately treated pregnancies has not been
The Maternal PKU Collaborative Study, begun in 1984, prospectively evaluates
the effects of dietary treatment during pregnancy.8
The Children's Hospital of Los Angeles serves as the coordinating center for
data entry and analysis. Four regional contributing centers in the United
States, 2 in Canada, and 1 in Germany are responsible for recruitment and
data collection at 78 participating centers.8
The participating centers obtained institutional review board approval either
by submitting a proposal to their research committees or by accepting approval
from the contributing center hospital. The institutional review boards from
each of the 4 contributing centers reviewed the protocols and procedures and
approved separate informed consent forms. Previous articles have focused on
nutritional9 and genetic10
characteristics of the mother and on offspring anomalies,11
birth measurements,12 neonatal neurological
status,13 and preliminary results of developmental
All women with PKU known to metabolic clinics in the United States,
Canada, and Germany were tracked at the start of the study and enrolled at
the time of pregnancy. Women with mild hyperphenylalaninemia (MHP) were also
invited to participate. Women with PKU were offered a low-phenylalanine diet
prior to or during pregnancy with the aim of maintaining metabolic control
(plasma phenylalanine, ≤10 mg/dL [≤605 µmol/L]). The women with
MHP had plasma phenylalanine levels of less than 10 mg/dL (605 µmol/L)
on a normal diet and treatment was not recommended. Almost all pregnant women
with PKU or MHP agreed to participate. Spouses of men with PKU, women who
had a previous child with a metabolic disorder, and women working in the hospitals
where the study took place were then recruited to participate in the nonhyperphenylalaninemic
(non-HPA) comparison group.
The final study includes 572 pregnancies, of which 412 were completed.
There were 75 spontaneous terminations, 79 elective terminations, 3 stillborns,
and 3 ectopic pregnancies. There were 4 sets of twins. Eight children died
of complications associated with maternal PKU. A standard protocol for treatment
of maternal PKU was provided to each participating center. The protocol specified
the monitoring of plasma phenylalanine levels and dietary intake on a weekly
basis and visits to a metabolic center once per trimester. Indices of nutrition
status (including weight gain and amino acid levels) were evaluated during
each trimester. Ultrasound examinations were performed at 20, 28, and 34 gestational
weeks. In addition, an obstetrician followed up each woman for regular prenatal
care. Offspring were evaluated developmentally in the neonatal period and
at 12 and 24 months (Bayley Scales of Infant Development). Thereafter, they
were evaluated every 2 to 3 years, through age 10 years. Clinic attendance,
adherence to medical recommendations, and cooperation with follow-up varied
At the time of this report, 368 children from the United States and
Canada meeting the criteria for inclusion had reached their 4-year birthday.
Two offspring diagnosed as having PKU and 2 offspring diagnosed as having
MHP were excluded. The children from Germany were excluded because a different
test battery was used. Of the 253 offspring receiving the preschool psychological
test battery, 149 were born to mothers with PKU, 33 had mothers with untreated
MHP, and 71 had mothers in the non-HPA comparison group.
The battery of psychological tests administered is presented in Table 1. For the first 6 years (at ages
3 and 5 years), the McCarthy Scales of Children's Abilities15
was the only test administered. The McCarthy Scales include a General Cognitive
Index (GCI) and indices for verbal, perceptual-performance, quantitative,
memory, and motor skills. Although now often replaced by other preschool tests,
the McCarthy Scales of Children's Abilities includes memory and motor development
scales. The McCarthy Scales also have been used for other longitudinal studies
involving preschool children.16,17
By the seventh year of the study, additional tests were administered,
but were scheduled at age 4 years. While 253 children received the McCarthy
Scales of Children's Abilities, 183 received the full battery of tests. The
evaluation obtained closest to age 4 years was used when more than 1 was performed.
The Test of Language Development–Primary provides a composite
spoken language score, which includes scales for listening, speaking, semantics,
syntax, and phonology.18
The Achenbach Child Behavior Checklist includes 112 behavioral items
that are rated by the parent or primary caregiver. T scores for total problems,
internalizing behaviors, and externalizing behaviors are calculated. A cutoff
score of 60 serves as the bottom of the clinical range for categorical discrimination
between deviant and nondeviant groups. T scores for 9 behavioral subscales
are also available. Scores above 67 indicate the clinical range.19
The Vineland Adaptive Behavior Scales are completed through a parent
or caregiver interview. Standard scores are available for an adaptive behavior
composite standard score and 4 domains: communication, daily living skills,
socialization, and motor skills.20
The Home Observation for Measurement of the Environment (HOME) scale
provides a measure of the physical circumstances of the home, variety of experiences
offered to the child, and the richness of language used. It provides a total
environmental stimulation score derived from the sum of items checked.21 When the evaluation was performed outside of the
home, the inventory was administered through an interview format. Scores less
than 85% were considered low.
The Wechsler Adult Intelligence Scale–Revised22
was used as the measure of maternal IQ and was administered to all women who
had not been administered this test within 5 years of the start of the study.
The majority of evaluations were conducted in the homes of the participants.
The psychologists were not informed about maternal metabolic control during
pregnancy or maternal IQ results. Standard test administration procedures
were followed. Informed consent was obtained at the time of enrollment. All
completed test protocols were sent without names to the psychology coordinating
center in Boston, Mass, where results were recorded and checked for accuracy.
They were then submitted to the coordinating center in Los Angeles, Calif,
where the data were entered.
Timing of maternal metabolic control was defined as the number of weeks
of gestation that elapsed before plasma phenylalanine levels remained lower
than 10 mg/dL (605 µmol/L). The maternal PKU offspring were categorized
into 4 treatment groups related to timing of maternal metabolic control: (1)
prior to pregnancy, (2) more than 0 up to 10 weeks; (3) more than 10 up to
20 weeks; and (4) more than 20 weeks or never in control. Results of cognitive
and behavioral assessments in HPA and untreated MHP offspring were compared
with results in the offspring from women in the non-HPA comparison group.
Since timing of maternal metabolic control and other maternal factors were
taken into account in multivariate analyses, each offspring was considered
a case. Thirty-nine mothers had more than 1 child in the study. The distribution
of test scores permitted use of parametric procedures. Statistical power was
calculated at more than 80% for comparisons of each subgroup with the non-HPA
comparison group if differences of 10 points were found on the standardized
psychological tests, but not large enough to detect differences of 5 or fewer
points in the untreated MHP and prior to pregnancy groups. Factor analysis
was performed to confirm domains of strengths and weaknesses in the children
(details available from the authors on request). The relationship between
maternal metabolic control and offspring outcome (risk for a low GCI score)
was determined through the use of logistic regressions. Variables that were
potential confounders of the relationship between in utero exposure to phenylalanine
and offspring GCI scores also were entered in the initial model. These included
maternal IQ, assigned maternal plasma phenylalanine level (plasma phenylalanine
level when on an unrestricted diet), maternal age, socioeconomic status,23 and level of stimulation provided to the child in
the home. The model was reduced, with variables being dropped 1 at a time,
until remaining variables reached P<.30. For treated
HPA pregnancies, the effect of birth order was tested through paired t tests. Also, for the treated HPA pregnancies, results
from the 2-year evaluation were compared with results from the preschool evaluation
at age 4 years using paired t tests. Statistical
programs from Systat24 and Stata25
were used. P<.05 was used as the criterion for
A total of 205 women enrolled in the Maternal PKU Collaborative Study
gave birth to 253 offspring who received evaluations at age 4 years. As noted
in Table 2, the majority of women
with PKU attained metabolic control after 10 gestational weeks. Maternal IQ
was lower in these women and their assigned plasma phenylalanine level was
higher. Their education and level of social position were lower.
Preschool evaluations were not obtained for 115 children, usually because
their mothers could not be located or did not respond to repeated telephone
calls and letters. Missing cases were almost evenly distributed among the
study groups (χ25 = 9.3; P
= .10). Using t tests, no significant differences
were found between mothers whose children received the preschool evaluation
and mothers of children for whom these data were missing in terms of the following
variables: maternal IQ (t303 = 0.91; P = .36), assigned plasma phenylalanine level (t262 = 0.91; P = .36), or socioeconomic
position (P = .80). Children whose mothers were younger
(t370 = 2.4; P
= .01) and whose average plasma phenylalanine levels were higher during pregnancy
(t269 = 2.1; P
= .04) were less likely to have been evaluated.
Results from the McCarthy Scales of Children's Abilities are presented
in Table 3. The GCI and subscale
scores on the McCarthy Scales declined as the number of weeks to metabolic
control increased. A clear linear relationship between number of weeks gestation
until maternal metabolic control and the McCarthy GCI score suggests a dose-response
association (r = −0.58; P<.001). The percentage of children attaining scores 1 and 2 SDs
below the mean increased as metabolic control decreased. Scores on each of
the McCarthy subscales followed a similar pattern.
A group of 10 children had mothers who attained metabolic control after
pregnancy began but within 6 weeks of gestation. They had a mean (SD) GCI
score of 96 (15). Overall, they performed not as well as those treated prior
to conception (mean [SD] GCI score, 99 ). Two performed more than 1 SD
below the normative mean (GCI ≤86). None of these children performed in
the range of mental retardation (GCI ≤72), while 8 of 16 offspring whose
mothers attained metabolic control between 6 and 10 weeks' gestation had a
GCI score of 86 or less, with 3 of them having a score of 72 or less.
In analyses describing the developmental profile in maternal PKU, children
from the untreated MHP and treated pregnancies were grouped since they shared
patterns of high and low scores. Factor analysis of data from the combined
HPA groups confirmed 3 factors: verbal/memory (varimax rotation factor loadings
>0.75), perceptual performance/motor score (factor loadings >0.67), and quantitative
(factor loading, 0.60). In all HPA groups, there was a consistent pattern
of weaknesses in language/memory and quantitative domains on the McCarthy
Scales. The verbal score was significantly lower than the perceptual performance
score (t181 = −4.9; P<.001) and the motor score (t173 = −3.2; P = .002), but was not significantly
different from the quantitative score (t181 = 0.82; P = .41) or the memory score (t181 = 0.42; P = .67).
The results of the Test of Language Development suggest specific effects in
expressive language and memory. Scores were 1 SD or more below the mean for
79% of the children on sentence imitation, 70% on word discrimination, 86%
on grammar use, and 53% on articulation. The children performed somewhat better
on tests measuring receptive language skills, with 40% attaining low scores
on grammatical understanding and picture vocabulary.
Within the behavioral domain, 30% (compared with the expected frequency
of 18%19) of the children received a rating
in the clinical range on the total behavior problems index of the Achenbach
Child Behavior Checklist. Their parents rated them as more external than internal
(t135 = 7.5; P<.001).
They were less likely to be rated as withdrawn than as having problems with
attention (t135 = 2.9; P = .005) or social relationships (t135 = 2.4; P = .02). Few children were rated
as having somatic problems, anxiety, or delinquency.
The children appeared to develop appropriate social skills at a slower
rate. On the Vineland Adaptive Behavior Scales, parents were somewhat more
likely to note slower development in social relationships than in daily living
skills (t143 = 1.6; P = .12).
Maternal IQ was significantly related to the HOME score (r = 0.36; P<.001) and to timing of maternal
metabolic control (r = 0.58; P<.001). Table 4 summarizes
logistic regression analyses used to determine the relative risk for a GCI
score of 86 or less. The variables that best predicted overall risk for a
low GCI score in the offspring were low maternal IQ, an assigned plasma phenylalanine
level greater than 20 mg/dL (1210 µmol/L), delayed maternal metabolic
control during pregnancy, and a low HOME score, reflecting the postnatal environment.
The final model had a sensitivity rating of 82% and a specificity rating of
65%, with 74% of cases correctly classified.
Twenty-seven mothers treated for maternal PKU had more than 1 child
enrolled in the study who received a preschool evaluation at age 4 years.
On average, the children were 2.6 years apart in age. The mean difference
in timing of maternal metabolic control was 2.2 weeks, with subsequent pregnancies
being treated later than first pregnancies (t = −0.96; P = .34). The mean GCI score for the later-born siblings
was 6 points lower (t = −1.6; P = .13).
As noted in Table 5, for
children born to women with PKU or MHP, scores on the McCarthy Scales at age
4 years were significantly lower than scores on the Bayley Scales at age 2
years. For children in the non-HPA comparison group, the differences were
Delayed metabolic control in maternal PKU remains a serious problem.
Treatment at any time during pregnancy appears to reduce the risks of cognitive
impairment and developmental delay associated with untreated maternal PKU.
Children are best protected from elevated maternal phenylalanine levels when
the woman attains metabolic control prior to pregnancy. There is no evidence
for a safe zone once pregnancy begins that marks a period in which the fetus
is fully protected.
While there is no doubt that maternal phenylalanine levels higher than
10 mg/dL (605 µmol/L) during pregnancy affect the fetus, levels lower
than 10 mg/dL (605 µmol/L) are not associated with significant declines
in GCI scores. However, the mean GCI scores of children born to mothers with
untreated MHP were somewhat lower than that of children born to mothers in
the non-HPA comparison group. While this might suggest some effect from MHP,
the women in the non-HPA group included physicians and other health care professionals
working at the medical centers. Their higher IQ and socioeconomic position
might account for the somewhat higher scores among their offspring.
When maternal metabolic control is not attained until after pregnancy
begins, neuropsychological functions are differentially impaired. The developmental
profile of maternal PKU offspring shows a distinct trend toward lower scores
on tests of language, memory, and quantitative abilities, while motor skills
and behavior are relatively less affected. Whether this reflects a fundamental
processing deficit or an underlying language deficit may be clearer in subsequent
testing of the children.
The scores obtained when the children were 4 years old were lower than
the scores on the infant development test. This is not an artifact of the
tests, since the children in the non-HPA comparison group did not display
a significant decline in scores. The decline in scores for maternal PKU children
may presage an inability to meet increasing cognitive and language demands
in the future.
This cohort of women with PKU experienced significant difficulties planning
their pregnancies, attending scheduled medical appointments, complying with
dietary restrictions, and monitoring their metabolic status. Adhering to the
diet did not become easier with successive pregnancies. Pregnancies of later-born
offspring were not within metabolic control sooner than earlier pregnancies
and outcomes were not better. The women were often less able to maintain metabolic
control, apparently not because of biological or genetic factors, but because
of a lack of resources, time, motivation, support, and organization. This
finding underscores the relevance of psychosocial factors in maternal PKU.
Methodologically, the study presented challenges for long-term follow-up
and data interpretation. Tracking, treatment compliance, adherence to the
study protocol, and follow-up rates varied between centers. The mothers of
children for whom the preschool evaluation was missing were younger and had
a higher average plasma phenylalanine level during pregnancy. This suggests
that an even greater proportion of children might have been in the latest
treated groups. Despite the possible impact of maternal IQ on offspring outcome,
this variable could not be completely controlled for in the analyses since
it was so highly correlated with metabolic control during pregnancy and with
other maternal variables. Nonetheless, the results provide compelling evidence
for the importance of strict treatment for maternal PKU.
One of the more interesting aspects of maternal PKU is its dissimilarity
in outcome to PKU and its similarity in outcome to other prenatal conditions.
For example, unlike maternal PKU offspring, children with PKU, whose exposure
to elevated phenylalanine occurs postnatally, have deficits in the areas of
visuomotor skills and fine motor speed.26 Untreated
children with PKU are profoundly affected. They display autistic features
or self-abusive behaviors. Children from late-treated or untreated maternal
PKU pregnancies experience mental retardation, but not to a profound degree.
This observation suggests that the timing of exposure may determine
the specific neuropsychological outcome. The literature on other maternal
conditions also suggests that there may be effects in particular areas of
functioning related primarily to timing and extent of exposure. For example,
children born to women with high alcohol consumption are almost indistinguishable
from children born to women with PKU.4,27
In fetal alcohol syndrome, studies have documented deficits in language, attention,
memory, and information processing.28 As in
maternal PKU, a clear dose-response relationship has been identified.29
Interventions directed at reducing risks at each step in the maternal
PKU cycle may prove to be most effective. In terms of outcome, every week
counts. Previous studies documented the importance of social support in preventing
unplanned (and hence late-treated) pregnancies in women with PKU.30 Camps, retreats, social support networks, newsletters,
and other activities to enhance social support should be offered.31,32 The women may need hospitalization.33 Compliance can sometimes be improved through changes
in formula or through the use of capsules containing the formula.34 In one home visitation program, mothers of children
with PKU (called resource mothers) provided practical assistance and social
support to pregnant women with PKU. A pilot study demonstrated that maternal
metabolic control was achieved 4 weeks sooner and offspring outcome was significantly
better in women who received the services of a resource mother.35
After the infant's birth, the possibility that the home environment
could moderate the adverse effects of maternal PKU should not be ignored.
To date, extraordinary efforts are sometimes made to treat the woman during
pregnancy, but then little assistance is offered following delivery. In this
study, the postnatal environment was one of the stronger predictors of the
GCI score. This finding coincides with the earlier result that maternal PKU
children from families with low HOME scores performed less well at 1 year
than children from families with adequate home stimulation, despite their
initial neonatal neurological rating.13 More
research is needed to determine what interventions are effective in reducing
risks for adverse outcomes before, during, and after pregnancy.