| Citation | Design | Population | Assessment of Vaccine Exposure | Outcomes | Results | Comment | Contribution to Causality Argument |
|---|---|---|---|---|---|---|---|
| DeStefano et al. (2004) | Case-control (Controlled observational) | 624 cases and 1,824 controls who were ages 3-10 years old in 1996
Cases: Population-based sample of children with autism in Atlanta metro area Controls: Children selected from regular education programs in Atlanta metro area and matched on age (in 1996), gender, and school | Immunization history abstracted from immunization forms required for school entry.
Demographic data collected for all children. | Autistic cases identified through MADDSP, a population-based surveillance program monitoring the occurrence of select disabilities among children in the Atlanta metro area.
Additional disability-related information, including birth defects, epilepsy, IQ level, and other disabilities collected from MADDSP. Additional data on birth and maternal characteristics was collected on a subset of children who had Georgia birth certificates. | Conditional logistic regression model stratified by matched control sets was used to estimate odds ratios.
Associations with autism analyzed using three specific vaccination age cutoffs (1) <18 months; (2) <24 months; (3) <36 months. A subanalysis was conducted of children with Georgia birth certificates to evaluate potential confounders in terms of birth and maternal characteristics. Similar proportions of children received the MMR vaccine by the recommended age of 18 months and <24 months. The rates of vaccination prior to 36 months were higher among cases (93.4%) than among controls (90.6%) (OR=1.49; 95% CI 1.04–2.14 in the total sample); (OR=1.23; 95% CI 0.64–2.36 in birth certificate sample). The association was strongest among the age group of 3- to 5-year-olds. | Similar patterns of age at first MMR vaccination among cases and controls. Similar proportions of cases and controls vaccinated according to ACIP schedule (<18 months) and by typical age of onset of autism (<24 months). Children with autism were more likely to be vaccinated before 26 months of age compared to controls. | The study shows no association between MMR and autism. |
| Taylor et al. (1999) | Self-matched case series (Controlled observational; Case crossover) | 498 children with autism born since 1979 from eight North Thames health districts, UK
261 core autism 166 atypical autism 71 Asperger’s syndrome | MMR information was obtained independently from clinical records from the Regional Interactive Child Health Computing System (RICHS) | Cases with autistic disorders were obtained from special school records and special needs/disability registers at child developmental centers.
A pediatric registrar extracted information from clinical records of <16 years of age. Information included age at diagnosis, child’s age when parents were first concerned, and age when regression was first noticed. Authors verified autism diagnosis by comparing ICD-10 autism definition to information on the child in clinical records. | Trends Core and atypical cases from 1979–1992: Test for zero trend: p < 0.001 Asperger’s syndrome from 1979–1992: p = 0.06 Core and atypical autism No evidence of sudden “step up” in 1987 No exponential trend change after MMR vaccine was introduced in 1987 Trends in autism incidence by birth cohort since 1987 were not temporally associated with changes in vaccine coverage. Difference in age at diagnosis between those vaccinated before or after 18 months of age and those never vaccinated p = 0.41 Interaction between vaccine categories and year of birth p = 0.29 Fold differences in mean ages (95% CI): Vaccinated before 18 mos. over unvaccinated: 0.91 (0.79–1.05) Vaccinated after 18 mos. over unvaccinated: 0.93 (0.81–1.08) MMR vaccine Relative incidence (95% CI) Autism diagnosis <12 months: 0.94 (0.60–1.47) Autism diagnosis < 24 months: 1.09 (0.79–1.52) Parental concern < 6 months: 1.48 (1.04–2.12) Parental concern < 12 months: 0.90 (0.63–1.29) Regression < 2 months: 0.92 (0.38–2.21) < 4 months: 1.00 (0.52–1.95) <6 months: 0.85 (0.45–1.60) | The study shows no association between MMR and autism. | |
| Farrington et al. (2001) Updates/Reanalyzes data from Taylor et al. (2000) | Self-matched case series (also referred to as case crossover) (Controlled observational) | Self-matched case series method (see Taylor et al., 1999, for details), same age groups.
Cases: 357 | Receipt of MMR vaccine or any measles-containing vaccine (MMR, single- antigen measles, measles and rubella). | Core autism and atypical autism. Autism diagnosis, parental concern, regression. | Relative incidence (95% CI) Receipt of MMR vaccine: Autism diagnosis <60: 1.24 (0.67–2.27) Any time after vaccine: 1.06 (0.49–2.30) Parental concern <36: 0.83 (0.50–1.36) Any time after vaccine: 0.76 (0.45–1.27) Regression <25: 0.76 (0.33–1.71) Any time after vaccine: 0.66 (0.26–1.66) Receipt of any measles-containing vaccines: Autism diagnosis <60: 0.96 (0.52–1.77) Any time after vaccine: 2.03 (0.80–5.18) Parental concern <36: 0.92 (0.56–1.49) Any time after vaccine: 0.89 (0.54–1.48) Regression <25: 0.76 (0.33–1.71) Any time after vaccine: 0.66 (0.26–1.66) | The authors conclude that the results did not support the hypothesis that MMR vaccine or measles-containing vaccine cause autism. | The study shows no association between MMR and autism. |
| Taylor et al. (2002) | Self-matched case series (also referred to as case crossover) (Controlled observational) | 278 children with childhood autism and 195 with atypical autism
Children born between 1979 and 1998 in five health districts in east London and identified from computerized disability register | Abstracted information from clinical notes, then linked the information to independent computerized vaccination records in the regional interactive child health system (TotalCare). Researchers also obtained information regarding whether the child received the vaccine before or after parental concern, or whether child had never received the vaccine. | Measured bowel problems lasting 3 months Age of reported regression Looking for evidence of a “new variant” form of autism | Proportion of children with childhood autism and those with atypical autism Bowel symptoms: w/childhood autism: 18% w/atypical autism: 16% p = 0.73 Regression: w/childhood autism: 23% w/atypical autism: 27% p = 0.27 Trends by year of birth (1979–1999) in proportion of children with autism Bowel symptoms OR: 0.98 (95% CI 0.93–1.04, p = 0.50) Regression OR: 0.98 (95% CI 0.93–1.03, p = 0.47) Difference between if child receiving the MMR vaccine before or after parental concern Bowel problems: p = 0.48 Regression: p = 0.83 Regression was reported by parents in 25% of the 469 children cases in which developmental information was recorded. Bowel problems reported in children with regression: 26% Without regression: 14% p = 0.002 | Authors suggest that the findings provide no support for an MMR associated “new variant” form of autism with developmental regression and bowel problems, and do provide further evidence against involvement of MMR vaccine in the initiation of autism. | The study shows no association between MMR and autism. |
| Fombonne and Chakrabarti (2001) | Cross-sectional study (Controlled observational) | Stafford sample: 96 children born between 1992 and 1995 and diagnosed with pervasive developmental disorder.
Autistic disorder = 26 Atypical autism = 56 Asperger syndrome = 13 Childhood disintegrative disorder = 1 Comparison samples: MHC sample: 69 children born between 1987 and 1996 with a confirmed diagnosis of PDD. MFS sample: 99 subjects born between 1954 and 1979 with an ICD-10 diagnosis of autism. | Stafford sample: Community pediatrician who assessed all of the children obtained immunization history from medical data. Immunization dates were verified by comparing information in the Child Health System. 99% of children received the first MMR vaccine; 65.6% received the second MMR MHC sample: Authors note that subjects in sample were likely to have been exposed to MMR vaccine based on birth dates. MFS sample: None were exposed to the MMR vaccine, based on birth year. | Stafford sample: Clinical investigations confirmed PDD diagnosis. Parents were administered the Autism Diagnostic Interview-Revised (ADI-R) by trained registrars to obtain information on symptoms related to regression. MHC Sample: Subjects were from patients seen at a specialist autism team. *Original study has methods for diagnosis. Parents were administered the ADI-R by trained registrars to obtain information on symptoms related to regression. MFS sample: Data were available for 98 subjects from ADI (first version) and an earlier version of the ADI-R. | CDD incidence (95% CI): 0.6/10,000 (0.02–3.6/10,000) Age at first Parental Concern (standard deviation) Stafford sample: 19.3 (8.7) MHC Sample: 19.2 (8.8) MFS Sample: 19.5 (13.6) F2250 = 0.02, not significant Rate of Regressive Autism (standard deviation) Probable: Stafford: 7 (7.3) MFS: 14 (14.3) Definite Stafford: 8 (8.3) MFS 4 (4.1) Any (Probable/Definite) Stafford: 15 (15.6) MFS: 18 (18.4) Statistical Testing: Probable and definite: χ2 = 3.65; df = 2; p > 0.15 Definite only: Fisher’s exact test; p = 0.25 Any (Probable/Definite) Fisher’s exact test; p = 0.70 Mean age of children when parents were first concerned (standard deviation) Stafford sample With regression: 19.8 months (7.4) Without regression: 19.3 months (9.0) t = 0.22; df = 94; not significant Mean interval between immunization date and age of children at first parental concern With regression: 248 days Without regression: 272 days t = 0.32; not significant ADI-R scores were not statistically significant between Stafford children with or without regression. No report of inflammatory bowel disorder OR between GI symptoms and regression (95% CI)/Fisher’s exact test: 0.63 (0.06–3.2)/p = 0.86, not significant | The authors note several limitations, including small sample size, information about children’s GI symptoms were obtained from survey of parents and clinical investigations, and children’s guts were not directly examined. | The study shows no association between MMR and autism. |
| DeWilde et al. (2001) | Case control (Controlled observational) | Cases: 71 children with autism
Controls: 284 (4 per case) matched for age, sex, month of MMR vaccination, and GP practice | MMR vaccination obtained from records in the Doctor’s Independent Network database. | The number of consultations 2 months and 6 months before and after MMR vaccination | Mean difference in number of consultations 2 months before and after MMR vaccination (95% CI)/Wilcoxon rank sum test: −0.05 (−0.54–0.44)/p = 0.45 Mean difference in number of consultations 6 months before and after MMR vaccination (95% CI)/Wilcoxon rank sum test: 0.04 (−0.75–0.83)/p = 0.59 Mean difference in number of consultations before autism diagnosis: 60 days before: 1.00 (0.38–1.61)/p = 0.007 180 days before: 1.90 (0.81–2.99)/p = 0.009 | Authors note that autism diagnosis was not confirmed. Also unclear how receipt of MMR vaccination was validated in subjects. Also unclear when autism was diagnosed. | The study shows no association between MMR and autism. |
| Madsen et al. (2002) | Retrospective cohort (Controlled observational) | Children born in Denmark from January 1991 to December 1998: 537,303
Vaccinated: 440,655 (1,647,504 person-years follow-up) Unvaccinated: 96,648 (482,360 person-years follow-up) | Vaccination information obtained from data reported to the National Board of Health by general practitioners. Information obtained from 1991 to 1999 for MMR vaccination received at 15 months of age. | Danish Psychiatric Central Register provided information on autism. Diagnoses made in psychiatric hospitals, psychiatric departments, and outpatient clinics in Denmark. Outcomes include autistic disorder (ICD-10 F84.0 and DSM-IV 299.00) and another autistic spectrum disorder (ICD-10 F84.1 through F84.9 and DSM-IV 299.10 and 299.80). | Adjusted Relative risk of vaccinated vs. unvaccinated children Autistic disorder: 0.92 (0.68–1.24) Other autistic-spectrum disorders 0.83 (0.65–1.07) Adjusted for age, calendar period, sex, birth weight, gestational age, mother’s education, and socioeconomic status. No association between development of autistic disorder and age at vaccination (p = 0.23) Interval since vaccination (p = 0.42) Calendar period at time of vaccination (p = 0.06) Adjustment for potential confounders yielded similar risk estimates. | Consultant in child psychiatry with experience in autism validated the diagnoses in 40 children with autistic disorder. 92% of the diagnoses fit the definition used by the CDC in an autism- prevalence study. | The study shows no association between MMR and autism. |
| Makela et al. (2002) | Retrospective cohort (Controlled observational) | 535,544 1- to 7-year-old children vaccinated between November 1982 and June 1986 in Finland. | MMR vaccination began in 1982 of children age 14 to 18 months and 6 years of age. Data from a surveillance study by the National Public Health Institute. Information also included age at vaccination, timing (year and month) of the first MMR vaccination. | ICD-8 Autistic disorders: 290–299 Psychoses; 95.8 Infantile autism; 308.99 Gerendum abnorme infantum
ICD-9 Autistic disorders: 299 Psychoses ex origine infantia 2990 Autismus infantilis 2998 Developmental disorder 2999 Developmental disorder From nationwide hospital discharge register Hospitalizations for autism between November 1982 and December 1995 | A total of 352 vaccinees were hospitalized for autism, of which 309 were hospitalized after vaccination. Interval between MMR vaccination and hospitalization was between 3 days to 12 years and 5 months. The authors did not detect clustering in the intervals from vaccination to hospitalization. 43 vaccinated after the first hospitalization; 31 hospitalized but were unvaccinated between November 1982 and June 1986. | The authors note several limitations: autism incidence could not be precise; nonhospitalized autism cases were not included; and some vaccinated children may have been classified as unvaccinated. | The study shows no association between MMR and autism. |
| Takahashi et al. (2003) | Case control (Controlled observational) | Cases: 21 autistic children born between 1988 and 1992. Two cases were diagnosed with Asperger’s syndrome; another case has a regressive clinical course.
Controls: 42 children matched to cases according to sex and birth year. From Tokyo Metropolitan Umegaoka Hospital. | Immunization history, including vaccination date, vaccine type, lot number, and provider’s name were obtained from the Maternal Child Health Handbook. From 1989 to 1993, MMR vaccine/monovalent measles vaccine administered at age 12 months; DTP admistered at 2, 4, and 6 months; OPV administered at 3 and 9 months; BCG administered at 3 months. Monovalent mumps and rubella were given to those who did not receive the MMR vaccine. | Medical records of autistic patients included information on age at initial onset of autistic symptoms, age of diagnosis according to ICD-10 code, and family history of ASD. Diagnosis of autism was made based on DSM-IV definition. | OR (95% CI) of autism to MMR Monovalent measles: 5.33 (1.03–27.7) Monovalent mumps: 3.33 (0.45–24.6) Monovalent rubella: 3.82 (0.59–24.7) Nonmeasles: immeasurable Nonmumps: 8 (1.33–48.2) Nonrubella: 8.57) (1.30–56.4 | The authors interpret the findings to demonstrate that autism risk after receipt of the MMR vaccine is lower than after receipt of monovalent vaccines. However, the authors note that selection bias, vaccination delay in autistic cases, and the small sample size may have affected the results.
There was a significant difference in immunization completeness between cases and controls for mumps, rubella, varicella, and encephalitis vaccination. 87.5% response rate for cases. Controls had 58% response rate. | The study shows no association between MMR and autism. |
| Dales et al. (2001) | Ecological | California children born between 1980 and 1994. | MMR immunization levels between 1980 and 1994 were determined by birth year from the California Department of Health Services Annual surveys of statewide random samples and from public and private school kindergarten immunization records. School immunization records were reviewed to determine age at receipt of immunization. Coverage levels were measured as proportions of children at 17 months in each year’s survey who received the MMR vaccine and those at 24 months who received the vaccine. | Autism cases, as reported to the California Department of Developmental Services. Children were born between 1980 and 1994 and had an ICD-9 code of 299. Other PPD outcomes were excluded. | Relative increase from 1980 to 1994 MMR coverage by age 24 months: 14% (from 72% to 82%) Number of autism cases: 572% (from 176 cases to 1182 cases) Rate of autism cases: 373% 1980: 44/100,000 births 1994: 208/100,000 births | Ecological nature of this study.
The authors note several limitations: Unknown how many autism cases were not accounted for in the system; expansion of the system probably affected the enrollment; the proportion of children born outside of California; diagnostic methods and categorization of people. | The study shows no association between MMR and autism. The study design limits its contribution to the causality argument. |
| Geier and Geier (2004a) | Study design is indeterminate. Classified as ecological because it relies on aggregate data for its analyses. | Children with autism from U.S. Department of Education data files | Estimates of primary pediatric measles- containing vaccine coverage for birth cohorts 1982, 1984, 1991–1996 | Estimated prevalence by creating birth cohorts for 1982, 1984, and 1991–1996 using age at which child was recorded as having autism. | Plotted estimated number of primary pediatric measles-containing vaccines in comparison to prevalence of autism for each birth cohort examined. Reported slope of the line was 48311 and linear regression coefficient was0.912 (regression model not specified). | These studies, which report a link between MMR and autism, have serious methodological flaws. Please see text. | The results are uninterpretable, and therefore noncontributory, with respect to causality. |
| Geier and Geier (2003c) | Study design is indeterminate. Classified as ecological because it relies on aggregate data for its analyses. | VAERS reports of serious neurologic symptoms following primary pediatric MMR immunization from 1994 to 2000 that developed within 30 days.
Comparison group: VAERS reports of vaccine recipients who developed serious neurologic symptoms after DTwP vaccine | Estimates of primary pediatric MMR vaccine coverage and DTwP coverage in birth cohorts from 1994 to 2000 from the CDC’s BSS. | Serious neurologic symptoms, as reported in VAERS: Cerebellar ataxia, autism, mental retardation, and permanent brain damage. | Risk estimates of autism after MMR vaccination compared to after DTwP vaccination Relative risk: 5.2 Attributable risk:3 4.2 “Percent Association”:4 84 Statistical significance: p < 0.001 95% CI: 3.0–9.2 | These studies, which report a link between MMR and autism, have serious methodological flaws. Please see text. | The results are uninterpretable, and therefore noncontributory, with respect to causality. |
| Kaye et al. (2001) | Ecological | Birth cohorts 1988–1993 of children registered in the GPRD. | Proportion of children registered in the GPRD who received the MMR vaccine for each annual birth cohort. Children received the vaccine within 60 days of birth and had at least 2 years of recorded follow-up. | 114 boys in the GPRD born between 1988 and 1993 and first diagnosed with autism at ages 2–5 years. The 4-year cumulative incidence of autism was calculated for each birth cohort. | 4 year risk of autism (95% CI) 1988: 8/10,000 (4–14) 1993: 29/10,000 (20–43) p < 0.0001 by score test for trend in odds MMR prevalence ≈97% for each successive birth annual cohort from 1988 to 1993 | The authors note that based on the findings, there was no correlation between autism incidence in each birth cohort (1988–1993) and MMR prevalence: as autism increased from 1988 to 1993, MMR coverage levels remained constant. Limitations include the ecological nature of the study, and the lack of confirmation of autism diagnosis. | The study shows no association between MMR and autism. The study design limits its contribution to the causality argument. |
| Gillberg and Heijbel (1998) Reanalysis of Gillberg et al. (1991) | Ecological | 55 Swedish children born 1975–1994 and diagnosed with autism
19 Swedish children born 1975–1994 and diagnosed with autistic like condition | MMR exposure was based on birth year. MMR was introduced for all 18-month-old children in Sweden in 1982. | Children were diagnosed with autistic disorder based on DSM-III-R definition. The authors note that autistic-like condition is similar to ICD-10 atypical autism. | Autistic children who did not receive the MMR vaccine (born between 1/1/1975 and 6/30/1980): 34 (62%)
Autistic children who received the MMR vaccine (born between 7/1/1980 and 12/31/1984): 21 (34%) Autistic-like condition in children who did not receive the MMR vaccine (born between 1/1/1975 and 6/30/1980): 13 (68%) Autistic-like condition in children who received the MMR vaccine (born between 7/1/1980 and 12/31/1984): 6 (32%) | The authors note that if MMR were related to autism, then more than 45% of children born between 7/1/1980 and 12/31/1984 would have developed autism. Similar results were found for children with autistic-like conditions. The authors concluded that the findings did not support a link between MMR and autistic disorder or atypical autism. Ecological nature of the study limits its ability to assess causality. | The study shows no association between MMR and autism. The study design limits its contribution to the causality argument. |
| Patja et al. (2000) | Passive surveillance | 1.8 million Finnish people who received the MMR vaccine 1982–1996 (3 million doses of MMR vaccine) | Since 1982, the MMR vaccine was administered twice to children—at ages 14–18 months and again at 6 years of age. Other groups include military recruits, health care workers, nursing school students, and 11- to 13-year-old girls who received the vaccine only once. From 1988 to 1993, rubellaseronegative women were vaccinated after delivery. | Death, likely allergic reactions, neurologic disorders, miscellaneous events.
Reports of adverse events were submitted to the central office by health care personnel, public health nurses, general practitioners, and pediatricians in primary care and hospitals. Reports were evaluated and, if needed, the hospital/health care center was contacted for additional information. | No cases of autism were reported.
Vaccine associated adverse events 437 173 serious reactions 169 potentially serious 79 hospitalizations | The authors concluded that there was no evidence to support the hypothesis of association between MMR vaccine and ASD or IBD. | The study shows no association between MMR and autism. The study design limits its contribution to the causality argument. |
| Peltola et al. (1998) | Passive surveillance | 1.8 million Finnish people who received the MMR vaccine 1982–1996 (3 million doses of MMR vaccine) | Since 1982, the MMR vaccine was administered twice to children—at ages 14–18 months and again at 6 years of age. Other groups include military recruits, health care workers, nursing school students, and 11- to 13-year-old girls who received the vaccine only once. From 1988 to 1993, rubella-seronegative women were vaccinated after delivery. | Gastrointestinal symptoms or signs lasting 24 hours or longer after receipt of the MMR vaccine. Information was obtained from hospital or health center records or from public health nurses. | Autistic spectrum disorder None reported
Gastrointestinal symptoms 31 children 20 hospitalized Symptoms included diarrhea, gingivostomatitis, vomiting only, and abdominal pains. Time between symptom onset and MMR vaccination: 20 hours to 15 days | The study shows no association between MMR and autism. The study design limits its contribution to the causality argument. | |
| Wakefield et al. (1998) | Case series | 12 children with history of pervasive developmental disorder referred to the Royal Free Hospital and School of Medicine in London, England. Children were between 3 and 10 years of age. | Researchers obtained subject’s immunization history from child’s parent or doctor. | Children underwent gastroenterological, neurological, and developmental assessment and review of developmental records. Ileocolonoscopy and biopsy sampling, MRI, EEG, lumbar puncture were done under sedation. Barium follow- through was conducted when possible. Biochemical, hematological, and immunologic profiles were examined. | Number of cases with behavioral problems Autism: 9
Disintegrative psychosis: 1 Postviral or postvaccinial encephalitis: 2 Number of cases with intestinal abnormalities All 12 children had intestinal abnormalities MMR vaccine In 8 children, onset of behavioral problems was linked to MMR vaccine by parent or physician 6 children had autism 1 postvaccinial encephalitis None of the children had neurological abnormalities on clinical exam. 5 had adverse reaction to MMR vaccine (rash, fever, delirium), 3 cases had convulsions. Average interval between exposure and 1st behavioral symptom: 6.3 days (range 1–14 days) | Authors did not find association between MMR vaccine and syndrome described. They identified a chronic entercocolitis in children that may be related to neuropsychiatric dysfunction. In most cases, onset of symptoms was after MMR vaccination. | This case series is uninformative with respect to causality. |
The reported slope of 4831 seems anomalous.
The authors appear to use the term “linear regression coefficient” in their analyses to refer to the coefficient of determination (R-sq), although these are distinct concepts. Slope (or partial slope) and coefficient are generally interchangeable terms in the context of regression. Data and variables used in these regression models were not included.
The authors used the epidemiologic measure “attributable risk” incorrectly. As noted by Mann (2003), attributable risk is intended to be a measure of the absolute, rather than the relative, difference in risk among the exposed and unexposed groups. Attributable risk percent, known also as the attributable fraction for the exposed or the etiologic fraction, is defined as (risk among the exposed – risk among the unexposed)/(risk among the exposed) and is mathematically equivalent to (RR–1)/RR × 100. Attributable risk percent is interpreted as the proportion of exposed cases whose disease can be attributed to exposure. A second related concept is the population attributable risk percent (also known as attributable fraction for the population) and is defined as (risk in the population – risk for the unexposed)/(risk in the population). Mathematically equivalent formulas are: (prevalence of exposure among cases) × (RR–1)/RR and (prevalence of exposure in the population) × (RR–1)/(1 + prevalence of exposure in the population) × (RR–1). Geier and Geier (2003b) calculate attributable risk by subtracting 1 from the relative risk value (RR–1 = attributable risk). Making attributable risk a fraction of the relative risk provides no information on the absolute or actual risk of thimerosal, nor does it provide information on the proportions of disease either among the exposed or population that can be attributed to exposure.
Percent association, which the authors define as [RR/{RR+1} × 100], is not a recognized epidemiologic measure.
From: Immunization Safety Review: Vaccines and Autism
NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.