The proposed study comprised 23 autistic children, in-
cluding the 2 siblings previously reported,' from the out-
patient department of a pediatric hospital. The study was
reviewed and approved by the Institutional Review Board.
Each child was examined by a pediatric neurologist,
and then classified as autistic according to the latest
criteria proposed by the American Psychiatric Associa-
tion's Diagnostic Statistical Manual of Mental Dis-
orders.16 After informed consent, a random urine sample
was collected without special preparation for organic
acid analysis by GC/MS for the presence of abnormal
metabolites. A solvent extract of the urine sample was
derivatized by trimethylsilylation and then analyzed with
a GC/MS, as described by Shaw et al.' No analytical
standards were available for the furan compounds, and
quantification of these compounds was performed by
assigning the response of an average size ion chro-
matogram peak as 100 units and then calibrating all other
peaks of the same compounds against these arbitrary cal-
ibrators; each of these compounds was positively identi-
fied by comparison to mass spectra in a GC/MS library.
For quantitative analysis of 3-(3-hydroxyphenyl)-3-
hydroxypropionic acid, 3,4-dihydroxyphenylpropionic
acid was used as a surrogate calibration standard since 3-
(3-hydroxyphenyl)-3-hydroxypropionic acid is not com-
mercially available. The reconstructed ion chromato-
graph signal of the ion at mass to charge ratio 398 was
used for quantitation. Similar use of surrogate standards
is used extensively in the GC/MS analysis of environ-
mental compounds in methods approved by the United
States Environmental Protection Agency.
If the presence of abnormal metabolites was detected
in the baseline urine sample (the sample collected before
antifungal therapy), the child was offered treatment of
suspected yeast infection with nystatin 100000 units 4
times daily orally for 10 days. Then, another random
urine sample was obtained and analyzed for organic
acids. If the presence of abnormal metabolites was detect-
ed in the second urine sample, the child was offered a sec-
ond course of treatment with nystatin for 2 months, and
another urine sample was tested.
The patients' baseline urine values served as con-
trols. Urine samples from normal children of hospital
employees served as additional controls. All of the con-
trol children were in good health and had no significant
diseases or developmental disorders. An assessment of
the severity of autistic behaviors was done by both a staff
psychologist working with parents and by teachers using
the Childhood Autism Rating Scale (CARS) scale.'7 This
assessment was done both at the beginning and at the end
of nystatin therapy. The use of 2 evaluators was included
to increase the reliability of this test. Furthermore, par-
periods of time.
All 23 children were evaluated and their baseline
urine was evaluated. One of the families moved shortly
after the study began and was lost to follow-up. Thus,
only 22 of 23 children completed the first course of nys-
tatin therapy. Seven children did not complete the second
course of nystatin for a variety of reasons. One of the chil-
dren was dropped from the study after the first nystatin
trial because the parent feared that the nystatin might
exacerbate a preexisting colitis condition for which the
child was hospitalized. One of the children refused to take
the drug. In another case, the personnel at a child's school
were uncooperative in administering the medication.
Statistical analysis of the data was done as follows:
The concentrations of each organic acid or other com-
pounds in the groups of male or female children with
autism were compared to the control groups using t tests
that assumed unequal variances. For comparison of pa-
tients following treatment with nystatin, or after comple-
tion of therapy, paired t tests were performed on each
metabolite concentration so that the value of each
metabolite in the treatment groups was paired with the
same patient's baseline value. This statistical technique is
sensitive in detecting small differences due to treatment.
Results
Characteristics of Patient and Control Populations
The patients who were initially recruited included 21
males and 2 females. None of these children had organic
acid abnormalities characteristic of any of the well-
defined inborn errors of metabolism based on baseline
organic acid testing. Patients were admitted into the study
in the order in which the parents or guardians applied. The
sex distribution of the controls included 17 females and
20 males. The mean age of the autistic children at the time
of admission into the study was 7.1 years (SD=3.2), with
a median age of 7.4 years and a range of 2.4 to 12.7 years.
The mean age of the normal children at the time of urine
collection was 7.7 years (SD=3.0), with a median age of
7.7 years and a range of 3.1 to 12.9 years. The mean age
of the male autistic children at the time of admission into
the study was 7.1 years (SD=3.3), with a median age of
7.4 years and a range of 2.4 to 12.7 years. The mean age
of the male normal children at the time of urine collection
was 8.1 years (SD=2.7), with a median age of 8.2 years
and a range of 3.3 to 12.8 years. The mean age of the
female autistic children at the time of admission into the
study was 6.7 years (SD=2.1), with a median age of
6.7 years and a range of 5.2 to 8.2 years. The mean age of
the female normal children at the time of urine collection
was 7.8 years (SD=3.3), with a median age of 7.3 years
and a range of 3.3 to 12.9 years.
