The title of this study isrelevant to both medical research and public health as it states importantissues such as, nutritional status during delicate life developmental periodswhich are pregnancy and childhood disorder. However, the title does not relateto the aim of the study. The conjuction ‘and’ in the title could have beenreplaced with the words “in association with”. This study was published by theAmerican journal of epidemiology online in 2016 and considerably reliable. The abstractsummarized clearly and concisely: the study aims, objectives, design, studysetting, hypothesis, methodology and results.
The PICO of the study can be seenin the flow chart below and was outlined in the abstract as: FIGURE 1: Showing causal diagram of thePICO for the Steenweg-de Graaff et al (2016) study.The authors could have logicallysummarised each sectionof the article by using subheadings outlining what is reported in each sectionof the journal. This enables readers to comprehend what is being outlined ineach section of the study. The study was designed as a prospective cohort and isappropriate because it is high on the evidence hierarchy pyramid, differentrare exposures can be studied, interventions and time sequence of outcomes canbe measured. Nevertheless, a randomized control trial can be applied to followup participants to improve the theory question and provide placebo that willreduce incidence of the disease.
To achieve the expected outcome,the various methodology applied were appropriate. However, interventions suchas maternal fish intake and child IQ were self- assessed. This can causeinformation bias and validate the accuracy of results. The study sample sizewas large and precise thereby reducing the risk of random error of estimation amongstvariables.
Nevertheless, there was a slight difference in the sample size ofsome subgroup, as some had significant numbers in comparison to othersubgroups. The different interventionsapplied was broadly explained with detailed guidelines, recounting theequipment’s used and procedures followed. Ethically, the study is notquestionable as ethical guidelines were followed and approved by the localmedical ethics committee. Consent and approval was obtained from adultparticipants and caregivers of the participating children.
Follow-up could havebeen applied to explore each of the study intervention rather than focusing ononly child IQ score. The construction of a causal diagram (Fig.1) summarising various variables measuredand sample effect estimates overall is a good attempt and included the sampleexclusion, inclusions and the rationale for inclusion.
However, reasons for exclusionswas not acknowledged. This questions the generalizability and homogeneity ofthe sample selection because the data inputted lacked precision and was looselydefined. For instance, the total number of eligible children who had data on IQscores were excluded from the flowchart but was mentioned as a criterion in thestudy population. The different interventions applied was broadly explainedwith descriptive guidelines, recounting details of the equipment’s andprocedures followed. Fish intake data was excluded from the flowchart.Clinically, important cofounding variables and outcomes such as:psychopathology in mid-pregnancy, alcohol consumption, smoking and folic acidsupplementation was reported in the paper. Considerably, thetraits of other offspring’s born to the same women in the study could becompared to children used in the study for any similarities.
Food frequencyquestionnaire was used to adjust for child fatty acid status. This method has alot of limitations and the authors would have substituted this by using anothermeasure of nutritional status such as a 24-hour dietary recall which is lessprone to information bias. Statistical analyses is reported for eachsubgroup in this paper, all analyses were carried out by using PASW statisticssoftware. This is particularlyimportant in order to answer and test the hypotheses. However, multiplestatistical testing was not performed in this paper, considering the differentsubgroups categories.
Nevertheless, the authors reported the use ofmultivariable linear regression to test the association of overall maternalPUFA status with child autistic traits. The linear graph/plot was not reportedin the article rather a quintile model was used to illustrate the linearresults. Most statistical analysis was adjusted for this could have been thecause of the large increase in power for some of the outcomes. Sample sizecalculations was stated to have been performed for some subgroups and notentirely for other subgroups. Therefore, increasing the risk of accepting thenull hypothesis when it is actually false (type II error). From, Table 1 theauthors have included other types of maternal fatty acid linked with autistictrait that was not mentioned earlier instead of focusing on just the totals of -3fand-6g.
Authors adjusted for multiple covariates that might have affected the resultsas the p value (0.0009), CI (-0.020, -0.009) for -3fwas significant before adjustment and after adjustment was notsignificant as the P value was (0.69), CI (-0.010,0.007). Nevertheless, theresults of -6gbefore and after adjustment remained significant thereby accepting thehypothesis of the study.
Additionally, from table 2 the results differ to table1 as before and after adjustments of both -3fand-6gthe results were associated to child autistic traits. A statistical test known as ‘Ancova’ couldhave been used to adjust for baseline and analyse the covariates which couldhelp minimise data drenching in the result section. Inthe discussion, the authors acknowledged the change in -3fresults when adjusted for child autistic trait in table 1.
This is because of somestudy limitations such as: shortage in some study samples, selection andinformation bias reported by the authors. Also, the use of biomarkers wassuggested for further research in order to confirm the study findings. Inconclusion, I think this paper is a fair attempt and can be improved innumerous ways such as: assessing PUFA status in mid-pregnancy more than once, substitutingself-assessed data in order to minimise information bias and socioeconomic groupassessment. However, the external validity can be generalizable because thetopic seems to be applicable to other programs, population and settings ratherthan the Dutch population. Nevertheless, I will be cautious when generalisingthe internal validity (the results) of the study based on the study sampleselection and participants. Further research would be useful in investigatingPUFA status in pregnancy with child autistic traits.