Urinary compounds in autism

Introduction

A Scandinavian laboratory technique claimed to identify autism (Trygstad et al. 1980; Gillberg et al. 1982). A molecular sieving technique provided patterns for the distribution of urinary compounds that were thought to be predominantly peptides and of diagnostic significance. The process was complex, slow, required 24-h samples of urine and, in a blind study of young male adults, did not distinguish those with autism from others of similar age and ability (le Couteur et al. 1988). Subsequently, Reichelt and colleagues in Oslo developed the technique further and, extending the hypothesis that many of the symptoms associated with autism might be explained by early opioid over-activity (Panksepp 1979), suggested that the compounds might derive from dietary opioids (Reichelt et al. 1986). The laboratory test was the basis for a dietary treatment that excluded gluten and casein as possible precursors of the urinary peptides (Reichelt et al. 1991). The Oslo procedure was replicated and further developed, being adapted to make use of high pressure liquid chromatography methods, by Shattock and colleagues in Sunderland (Anderson et al. 2002).

The aim of this study is to test the laboratory's ability to distinguish a group of children who had autism in addition to varying degrees of learning disability, using samples of their urine, from those of two other groups matched respectively for (1) both chronological age and degree of learning disability (LD); and (2) chronological age only, being of normal ability (NA).

This study differs from the earlier one (le Couteur et al. 1988); first, in using prepubertal children and, second, in its exclusion of anyone with a known medical basis for their disability.

There are suggestions that, in autism, those with a mild learning disability may be aetiologically distinct from those with a severe learning disability (DeLong 1999). Accordingly, we analysed these two groups separately as well as together.

We enrolled a group of prepubertal boys, who were living at home, from a data bank of children meeting the ICD-10 criteria for childhood autism (F84.0) and who had been diagnosed by an experienced clinician (T.P.B.) at interview, using the Childhood Autism Rating Scale (Schopler et al. 1988) to combine data from informant accounts and direct observation. Thirty-four of the 38 families who agreed to take part in the study were able to collect the urine sample. We had excluded those with a known medical condition, a specific bowel problem or a highly restricted diet. The children were not taking medication with the exception of anti-epileptics. Functional ability was graded according to ICD-10 (F70-F79) on the basis of the interview as well as any previous assessments, including educational, that were available: chronological age and level of ability are shown in Table 1.

The two comparison groups of children were perceived as warm and sociable by their teachers and were recruited with the help of local schools.

Method

The parents collected the first morning sample of urine in two bottles that were identified only by a number. It was not easy to get the co-operation of their children and three of the samples were obtained in special, absorbent urology pads. A separate series of tests were carried out to ensure that this technique did not interfere with the peptide profile, either by selective absorption of the compounds or by introducing extraneous compounds.

The bottles contained thymol as a preservative and the urine was immediately chilled. The same day it was collected, the urine sample was divided and frozen at −4°C (by A.A. and K.B.). The laboratory then was given a mixed batch of anonymous samples for analysis (by P.S. and D.S.) so that it was unaware of the clinical diagnosis. After cleaning, the urine samples were separated by high-performance liquid chromatography (reversed phase gradient elution) to give a visual profile, time being represented on the x-axis and the concentration of the eluted compound on the y-axis (Fig. 1). A previous study had found these profiles to remain constant over several months (Shattock et al. 1990).

The laboratory workers (P.S., D.S. and M.M.) made an empirical judgement of the significance of the chromatographic profiles categorizing them on visual examination as to whether they identified autism. These laboratory diagnoses were then compared with the clinical diagnoses; first, for the group as a whole (Table 2) and then for two subgroups, those with normal ability/mild learning disability and those with moderate/severe learning disability (Table 3).

An attempt was made to operationalize the criteria on which the laboratory diagnosis was made. The only area of real interest was those peaks which were thought to represent the peptides with known biological activity appearing between 17 and 30 min of elution. The diagnosis of autism takes two main aspects into consideration. First, there had to be evidence of peptiduria as provided by substantial peaks within this area. These had to be greater than those of the smaller, water soluble components which eluted much earlier in order to exclude those subjects whose large peaks simply reflected a high concentrate of urinary solutes as a result of low fluid intake. Second, the main marker peak, occurring at the 20–21.9-min position, had to be the largest, greater than 1.5 cm height, and without a shoulder (which might suggest its height was being bolstered by another substance). Another member (K.B.) then applied these rules to the traces.

The results of the analyses were then collated (by T.P.B.) and related to the clinical diagnosis, using the Statistical Package for the Social Sciences.

Results

The results are summarized in Tables 2 and 3. They demonstrate that the laboratory was able to identify correctly the urine of boys with autism in over half the cases when analysed blind to diagnosis (P = 0.009). The results are markedly improved, however, when the analysis is restricted to those of normal ability or mild learning disability (P = 0.00005).

Our attempt to operationalize the criteria were only partially successful (P = 0.044), identifying a greater proportion of the autistic population but also falsely identifying higher proportions of the non-autistic groups (sensitivity = 63%; specificity = 62%).

The samples had been divided and a series sent simultaneously to a Scandinavian laboratory for analysis. It became clear that this laboratory was working on different evaluative criteria and norms to identify autism and that their laboratory results bore little relationship to the clinical diagnosis. It withdrew from the study.

Conclusions

People with autism have a distinctive biochemical profile that is readily measured, being non-invasive and within laboratory resources. However, there were a substantial number of false negatives within the autistic group (47%) and, in retrospect, two elements contribute to this. Firstly, a later review of the chromatograms of this group, across the whole spectrum of ability, showed that they had smaller but easily seen peaks scattered over the 20–25-min period of elution that, by the criteria, indicated a negative result. Identification of the nature of these compounds may show them to be relevant to autism. Secondly, the urine of those children with a moderate/severe degree of disability frequently showed a large number of peaks across the whole chromatogram. These may mask the peptide peak relevant to autism, making the test unsuitable for this population.

In the normal ability group while 28 out of 34 (83%) were correctly identified as ‘normal’, six (18%) were incorrectly identified as autistic. Here the peptide peak at 20–21 min may represent a marker for a factor predisposing to autism and might therefore be expected to occur in a minority of the normal population.

The methodological problems postulated in previous work were addressed. In particular, we used urine from young prepubertal boys with autism with no other coexisting cause for their degree of learning disability. All analyses were carried on mixed batches of urine and the laboratories remained blind to the diagnosis until after they had forwarded their results. The lack of success by the Scandinavian laboratory may be because of methodological differences in analytical technique. However, discussion of these differences also suggested that there may be regional differences in the urinary profile; that the characteristics of a population in England, both with and without autism, are different from those of in Scandivania, a point not pursued in the earlier study (le Couteur et al. 1988).

This study indicates the potential for a biological marker for autism. It also supports the hypothesis that there may be an essential difference between the autism of children with a severe learning disability compared with those more able (DeLong 1999).

Clearly, the findings require replication and the test has the advantage that it uses a single early morning urine sample and that it can be analysed quickly. However, the results indicate that the abnormal compound is not specific to autism and that the chromatographic anomalies only identify a subset of the population with autism. It is important to emphasize that this is not a diagnostic test nor has it been shown to give any guide to management. Its utility appears limited to prepubescent children who do not have a severe learning disability.

Subsequent work to purify and identify the peptide responsible for the 20-min peak suggested that this might be indolyl-3-acryloylglycine (IAG) (Anderson et al. 2002).