Forty-seven children were recruited. Six recruits did not complete the study: 3 potential cases dropped out prior to colonoscopy; 1 potential case and 2 potential controls completed colonoscopy but had incomplete clinical assessments. No differences were found in age, sex, or case-control status between study completers and non-completers. An additional 2 potential cases were excluded for failure to meet diagnostic inclusion criteria (below cutoffs for autistic disorder [AUT] on ADI-R); and 1 case was excluded because no bowel biopsy material was available. The final study population consisted of 25 cases (AUT/GI group) and 13 controls (GI control group) presenting consecutively for ileocolonoscopy who received at least one dose of MMR and completed all study procedures.
Age at biopsy was similar for cases and controls [median (interquartile range, IQR), cases, 5.5 (3.0) years; controls, 5.1 (3.0) years], as was the distribution of cases and controls across the three age strata. Girls were older at biopsy than boys (P
0.01). There were no significant differences between cases and controls in their distribution by sex within the three age groups ().
The clinical indications for endoscopic/colonoscopic procedures commonly noted in both AUT/GI and GI groups included recurrent abdominal pain (RAP), gastroesophageal reflux, vomiting, and food allergies. Although the more subjective factor of RAP was frequently present in both cases (36%) and controls (38%), it was rarely the sole rationale for GI examination in either group (1 of 25 cases, or 4%; 2 of 13 controls, or 15%; P
Median age at receipt of first MMR was similar for cases [15.3 (1.7) months] and controls [16.0 (4.9) months]. The majority of study subjects were in the 3–5 year age stratum and below the age recommended for second MMR (4–6 years 
); expectedly, 80% of cases and 69% of controls received only one MMR prior to the study (P
0.36). Consistent with the older age of girls in the study, there was a trend toward a higher proportion of girls than boys receiving a second MMR (P
0.13). None of the children received MV-containing vaccines other than MMR.
Clearance of MV depends on development of adaptive immunity. As cell-associated MV RNA may be present transiently after receiving MMR 
, timing of vaccination relative to biopsy was potentially important. Parental reports of timing of MMR receipt 6 months or more prior to biopsy were in accord with pediatric provider immunization charts for the final study population with the exception of one control boy whose immunization record revealed receipt of a second MMR 3.5 months prior to biopsy. This subject was retained in final analyses after determining results to be the same both with and without inclusion of his data. The median MMR-biopsy interval was similar for cases [40.8 (26.7) months] and controls [39.8 (21.1) months], and was not influenced by sex (). Older age at biopsy was associated with a longer MMR-biopsy interval, independent of case status (Spearman rank correlation, Rho
Controls received a greater median number of all types of vaccines than cases [20 (1) vaccines vs. cases, 17 (4); P
0.04, ). Total number of vaccines received was not related to age or sex.
The study sample included two sibling pairs; three of these children were controls (2 males, 1 female) and one was a case (male). Data from sibling pairs were retained after determining that patterns of results were unaltered by sibling pair exclusion.
AUT diagnoses were confirmed for all cases. Absence of AUT, other ASD, or other developmental disturbances was confirmed for controls. For one control, ADI-R was incomplete; this subject was retained after determining that CDI and clinical assessment were consistent with typical development.
Median AUT onset age was 13.5 (7.0) months (). Cases had a high rate of CPEA-defined behavioral regression (loss of language and/or other skills following acquisition), 88%, compared to published rates of 20–40% for the general ASD population 
Onset of GI episodes and autism relative to MMR administration.
Real-time RT-PCR assays
Prior to examination of study samples, performance of the four different primer sets (two for H gene, two for F gene) was evaluated for the 12 cloned target regions using synthetic RNA standards. A lower limit of detection of 50 RNA molecules per reaction was confirmed for each primer set in all laboratories.
All laboratories correctly identified all positive controls using pre-established criteria for positivity (positive results in at least two of three wells with at least one of the primer pairs for F and one of the primer pairs for H). All laboratories correctly identified all negative controls.
Concordance across laboratories was achieved in the initial round of real-time RT-PCR assays for all positive and negative results with the exception of a single study sample, an ileal biopsy from a control. An additional three samples, one ileal sample (from a control) and two cecal samples (one case, one control) yielded signal in at least one assay in one laboratory but did not meet criteria for positivity. All four samples were retested as below to resolve discrepancies.
As detailed above, only one sample met the pre-established definition of discordance; in this instance, an ileal sample from a control was positive with all four MV primer pairs in a single laboratory. Neither of the other two laboratories reported positive wells with any primer/probe combinations for this sample. The amplification product from this reaction was sequenced and determined to contain the engineered restriction site, confirming that it represented the synthetic transcript control. This sample was classified as negative. Aliquots of the three other samples that had yielded signal in one assay in a single laboratory were shipped to all three laboratory sites for retesting under new IDs. Two negative and one positive control were included to ensure blinding and monitor assay performance. Repeat testing of these three discordant samples with the F or H gene sequence primer/probe set responsible for the initial single positive finding failed to reproduce positive results in any of the three laboratories on the second round. In all three instances, results were negative on second round testing, including the one laboratory initially reporting positive results for a single primer pair.
MV RNA in bowel biopsies
Analyses in all three laboratories found two ileal biopsy samples with MV F gene and H gene RNA: one from a boy in the AUT/GI group, the other from a boy in the control group. Real-time RT-PCR indicated a range of 2–7 molecules per PCR reaction, corresponding to approximately 50–500 MV RNA molecules per 100 ng of total RNA extract (). Sequence analysis confirmed that products of these samples were authentic. MV RNA was not detected in cecum of these subjects, or in ileum or cecum of any other subject. The presence of MV sequences was not associated with an AUT diagnosis (cases, 4%, controls, 8%).
MV RNA detected by real-time RT-PCR.
Both subjects with positive samples had reactive lymphoid follicles (RLF). In the AUT/GI subject, RLF were present in both small and large intestine; the control had RLF restricted to colon. Endoscopy revealed inflammation in both subjects: the case had nonspecific gastritis; the control had acute distal esophagitis. Other cases and controls had RLF and/or inflammation in their upper and lower GI tracts, but MV sequences were not detected in their GI samples.
Timing of MMR, GI episodes and AUT
If MMR is causally related to either GI disturbances or AUT it should precede their onset. Similarly, if GI disturbances contribute to AUT they should precede onset of AUT. We approached temporal relationships in the following manner: subjects with MMR administration and GI onset in the same month were considered to have MMR administration before the onset of GI episodes; subjects with GI episode and AUT onset within the same month were considered to have GI onset before AUT onset; and subjects with MMR and AUT onset within the same month were considered to have MMR onset before the onset of AUT.
There were no significant differences in the proportion of cases and controls with MMR before onset of GI episodes: 12 of 25 cases (48%) received MMR before GI episodes began as compared with 3 of 13 controls (23%; P
0.13; ). To examine whether the MMR-GI onset interval differed for cases and controls, survival analysis was pursued, using only those children with onset of GI episodes after MMR administration. Kaplan-Meier analysis showed no differences between cases and controls in latency from MMR to initial GI disturbances (Mantel-Cox logrank test).
To determine whether our data supported the hypothesis that GI pathology contributes to ASD pathogenesis, we examined the temporal relationship between MMR immunization, first GI episode, and AUT onset. If the putative relationship of MMR to GI pathology and AUT is valid, MMR must precede GI dysfunction and AUT, and GI dysfunction must precede AUT. If GI dysfunction contributes to AUT independent of MMR, it is necessary only that GI dysfunction precede development of AUT. X2
analyses indicated no role for MMR in either the pathogenesis of AUT or GI dysfunction (). Only 5 of 25 subjects (20%) had received MMR before the onset of GI complaints and had also had onset of GI episodes before the onset of AUT (P
Number and frequency of AUT/GI subjects receiving MMR before or after GI onset and with index GI episode before or after ASD.
Cases first receiving MMR prior to onset of GI complaints were older at index GI episodes [21.0 (22.0) months] than cases receiving their first MMR after GI episodes already began [1.0 (12.0) months; p<0.0001; ]. Conversely, cases with GI episodes preceding AUT onset had much earlier onset of GI problems than cases with initiation of GI episodes after onset of AUT [2.5 (13.0) vs. 30.0 (23.3) months, respectively; P
Ages of AUT/GI subjects receiving MMR before or after GI onset and with index GI episode before or after ASD.