Population attributable fractions for three perinatal risk factors for autism spectrum disorders, 2002 and 2008 autism and developmental disabilities monitoring network

Probably not a substantial claim: may not pass S3 (load-bearing).

Conditional and unconditional logistic regression produced very similar results.

“The findings from conditional and unconditional logistic regression were very similar; thus, we only present the former.”

May combine two claims.

For the 1994 birth cohort, the estimated average PAFs were 4.2% for PTB, 0.9% for SGA, and 7.9% for CD.

“For the 1994 cohort, average PAFs were 4.2%, 0.9%, and 7.9% for PTB, SGA, and CD, respectively.”

• Also stated in: § Results · plain

  “For the 1994 cohort, average PAFs for PTB, SGA, and CD were 4.2%, 0.9%, and 7.9% respectively, with a summary PAF of 13.0% (7.7%–19.5%) (Table 4 and Fig. 1).”

May combine two claims.

For the 2000 birth cohort, the estimated average PAFs were 2.0% for PTB, 3.1% for SGA, and 6.7% for CD.

“For the 2000 cohort, average PAFs were 2.0%, 3.1%, and 6.7% for PTB, SGA, and CD, respectively, with a summary PAF of 11.8% (7.5%–15.9%).”

• Also stated in: § Results · plain

  “For the 2000 cohort, average PAFs for PTB, SGA, and CD were 2.0%, 3.1%, and 6.7%, respectively, with a summary PAF of 11.8% (7.5%–15.9%).”

May combine two claims.

Among ASD cases, frequencies of PTB, SGA, and CD were 13-14%, 11-12%, and 27-28%, respectively, in both cohorts.

“Overall, the frequencies of PTB, SGA, and CD among cases were 13%–14%, 11%–12%, and 27%–28%, respectively, for both cohorts (Table 2).”

May combine two claims.

All three factors were present in 0.7% of 1994-cohort cases and 1.0% of 2000-cohort cases.

“The proportion of cases with all three factors was 0.7% and 1.0%, for the 1994 and 2000 cohorts, respectively.”

May combine two claims.

Compared with the full ADDM populations, both study cohorts had somewhat lower PTB and SGA rates and somewhat higher CD rates.

“For both cohorts, our study populations had slightly lower preterm and SGA rates and slightly higher CD rates than the total ADDM populations.”

May combine two claims.

Results comparing males and females were imprecise and inconsistent.

“The findings for males versus females were imprecise and inconsistent.”

May combine two claims.

Nearly all 18 stratum-specific odds ratios were greater than 1.0, although some confidence limits included 1.0.

“Moreover, within both cohorts, nearly all the 18 stratum-specific ORs were greater than 1.0 (although some CLs included 1.0), indicating that children with ASD were more likely to have each risk factor alone or together with other risk factors.”

May combine two claims.

The 2000 birth cohort was comparable with the total ADDM population on race-ethnicity and maternal education, while the 1994 cohort had slightly lower proportions of Hispanics and advanced maternal education.

“The 2000 birth cohort was also comparable on race-ethnicity and maternal education, although proportions of His-panics and advanced maternal education were slightly lower for the 1994 cohort study population.”

May combine two claims.

In both cohorts, controls generally had lower percentages of each risk factor, overall and within strata of the other factors, except for SGA without PTB or CD in the 1994 cohort.

“For both cohorts, the percentages of the risk factors, overall, and in various strata of the other factors was lower among the controls than cases, with only one exception (SGA without PTB or CD in the 1994 cohort).”

May combine two claims.

Subgroup results hinted that the summary PAF might be somewhat larger for ASD cases who also had intellectual disability, but the confidence limits for the ID and non-ID groups overlapped and the difference was hard to interpret.

“Subgroup analyses for both cohorts (Table 4) were suggestive of a slightly higher summary PAF for ASD cases with co-occurring ID; however, CLs for the two subgroups with and without ID overlapped and the large PAF difference for the subgroup with ID unknown is difficult to interpret.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Confidence limits at 95% for the average and summary PAFs were obtained by bootstrap with 1000 resamples.

“We estimated 95% confidence limits (CLs) for average and summary PAFs using a bootstrap technique with 1000 replicates.”

Probably not a substantial claim: may not pass S3 (load-bearing).

Sensitivity analyses were conducted to assess the impact of inadvertently selecting a case as a control.

“We conducted sensitivity analyses to assess the impact of this possible problem (see the following section).”

• Also stated in: § Statistical methods · plain

  “We also performed sensitivity analyses to consider the potential problem that in some case-control sets, one selected control was actually the case.”

May combine two claims.

The selection criteria did not affect internal validity but narrowed generalizability.

“These selection criteria, although not impacting internal validity, did narrow the generalizability.”

May combine two claims.

The average PAF method accounts for co-occurrence among risk factors and adjusts for confounders, avoiding overestimation from separate crude or adjusted PAFs.

“Average PAF methodology accounts for risk factor co-occurrence.”

• Also stated in: § Statistical methods · hedged

  “Average PAFs account for co-occurrence among the three factors while adjusting for other potential confounders; thus, in general, average PAFs address the inherent over-estimation that occurs when computing separate crude or adjusted PAFs [21].”

• cites [21] Rockhill, 1998, Use and misuse of population attributable fractions, Am J Public Health sourceNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim. May combine two claims.

Each case was matched with 20 controls using birth year, sex, maternal race-ethnicity, county of residence, maternal age, and maternal education as matching variables, to control confounding.

“To carefully and efficiently consider confounders, we used a matched design. We matched each case to 20 controls from the same birth year on sex, maternal race-ethnicity (NHW, NHB, Hispanic), county of residence, age (<20, 20–29, 30–34, 35+ years), and education (high school or less, greater than high school) at birth.”

The quote only partly states this claim.

PTB, SGA, and CD together make a notable contribution to ASD risk in the U.S. population studied.

“Three perinatal risk factors notably contribute to ASD risk in a U.S. population.”

• Also stated in: § Discussion · boosted

  “This study establishes that perinatal risk factors (or suboptimal perinatal environment) make a notable contribution to ASD risk in a U.S. population.”

Classified as probably not a substantial claim.

Odds ratios were adjusted for matching factors using conditional logistic regression.

“Given the matched design, we used conditional logistic regression to calculate ORs adjusted for the matching factors.”

May combine two claims.

The matched odds ratios for each perinatal factor (1.1-1.5 overall) agree with earlier studies and fall near the lower end of estimates from a systematic review.

“The risk factor-ASD matched ORs reported here (1.1–1.5, for each perinatal factor overall) are in line with previous studies, albeit toward the more conservative end of the range of estimates we recently reported from a systematic literature review [12].”

• cites [12] Schieve, 2011, Have secular changes in perinatal risk factors contributed to the recent autism prevalence increase? Development and application of a mathematical assessment model, Ann Epidemiol sourceNot checked yet

  Not checked in the free run.

Classified as probably not a substantial claim. May combine two claims.

The findings suggest general risk from a suboptimal perinatal environment but lack the detail needed to recommend specific clinical practices.

“This is perhaps particularly important with regard to CD, as these study findings point to risk contributions from a suboptimal perinatal environment generally but are not sufficiently detailed to make recommendations about any specific clinical practices.”

• AssumptionRests on an assumption: The PAF for CD represents a composite of all reasons for having a CD along with any direct effects.

May combine two claims.

The study used a well-defined consistent population at two time points, a matched design to consider confounders, and a rigorous PAF estimation methodology that addressed inherent biases in most PAF studies.

“This study has several strengths including use of data from a well-defined consistent population at two time points, use of a matched design to efficiently consider potential confounders, and use of a rigorous PAF estimation methodology that addressed inherent biases present in most PAF studies.”

May combine two claims.

The results are useful at the population level but cannot identify which etiologic subtypes of PTB, SGA, or CD contribute most to ASD.

“Thus, although the findings are informative on a population level, they do not indicate which etiologic subgroups of PTB, SGA, and CD most contribute to ASD.”

• AssumptionRests on an assumption: Each of PTB, SGA, and CD represents multiple potential underlying etiologic mechanisms or pathways.

May combine two claims.

A rise in the SGA PAF counterbalanced decreases in the CD and PTB PAFs, leaving the summary PAF stable.

“The increase in the PAF for SGA offset the declines in PAFs for CD and PTB. Thus, the summary PAF, which more globally represents suboptimal perinatal environment, remained stable.”