Table 1 Characteristics of excluded studies (full text articles) and the reason for their ineligibility

Pogoda and Preston-Martin, 1997. [35]

Follow up to population-based case–control study of pediatric brain tumors in Los Angeles county, CA, US, from 1984 through June 1991. Cases, N = 224, Controls, N = 218.

Household pesticide use from pregnancy to diagnosis assessed by phone interviews.

Risk was elevated for flea-tick pesticides, OR = 1.7, 1.1-2.6.

Round up was listed among pesticides (herbicides) used in law-garden. However, only associations between broad classes of pesticides and the outcome (brain tumors) were tested. Exposure assessment encompassed both pre and postnatal development and did not discriminate between the two periods.

Larsen et al., 1998. [36]

Retrospective study of time-to-pregnancy (fecundability) among Danish farmers (N = 1146). Telephone interviews on time-to-pregnancy (How many months did it take your wife to get pregnant?) and reproductive history

Exposure data obtained by phone interviews (use of pesticides the year before the youngest child was born). Exposed groups: traditional farmers, traditional farmers who did not spray pesticides themselves, organic farmers.

Fecundability ratio between “traditional farmers who applied pesticides” and “organic farmers”: 1.3, 0.75 – 1.40.

Although GLY is mentioned in the Introduction section, authors did not investigate whether ‘time-to-pregnancy” was altered specifically by this herbicide. The authors tested only classes of “pesticides”.

Nevison 2014. [37]

Ecological-type study design. Temporal trends (1970–2005) in autism compared to temporal trends in US application of GLY. Autism counts were from the Individuals with Disabilities Education Act (IDEA) database for 50 US states plus District of Columbia.

GLY exposure data were aggregated at the group level. Data were from US department of agriculture.

Increasing trends in application of GLY positively correlated to a marked rise in autism prevalence in the US.

In this study of temporal trends, data on potential exposure were aggregated at the whole US population level and encompasses preconception, pre-natal and postnatal periods of development. Moreover, the ecological study design leaves the door wide open for non-causal explanations for the observed parallel rise in temporal trends in GLY use and autism prevalence.