Commentary: Representativeness is usually not necessary and often should be avoided

Female Smokers Are at Greater Risk of Airflow Obstruction Than Male Smokers. UK Biobank

“…We are aware that the UK Biobank may not be representative of the general population (24). However, our findings are not descriptive, rather they regard to risk of airflow obstruction associated with smoking history and are internally valid, which is of greater relevance than representativeness when inferring causality (25,26).…”

Association of Maternal Hypertension and Chorioamnionitis With Preterm Outcomes

“…39 This line of reasoning is gaining momentum in epidemiologic studies, where a "lack of representativeness" has been shown to provide unbiased estimates of effects under most conditions, and indeed to be often advantageous in terms of efficiency. 40,41 Moreover, although we employed definitions widely used in clinical practice and research (those of the Vermont Oxford Network, the largest neonatal network worldwide), we admit that exposures were defined without any clinical detail, as is always the case in large epidemiologic studies, and consequently there is the possibility of some misclassification. In particular for chorioamnionitis, our classification probably detected the more severe/ manifested cases for the exposures under examination and not the whole spectrum of the disease.…”

“…In particular for chorioamnionitis, our classification probably detected the more severe/ manifested cases for the exposures under examination and not the whole spectrum of the disease. Again, the fact of focusing only on more severe degrees of exposure (or differently severe degrees in the 2 arms of study) does not bias the results, [39][40][41] as long as there is not differential misclassification (that is, different recording of exposure owing to the presence/ absence of outcomes, for example, preferentially recording chorioamnionitis instead of hypertension in case of PVL). Given that we collected our data prospectively, and the outcomes were recorded without knowledge of the aim of this study, diagnostic bias and differential misclassification were probably low.…”

Large panel-survey data demonstrated country-level and ethnic minority variation in consent for health record linkage

“…The inference is that representativeness is therefore as important for the follow-up cohort as it is for the initial survey, where the aim was to estimate the occurrence of a disease or risk factor in the population. However, follow-up studies are more usually concerned with questions of aetiology and the importance of representativeness in this context is hotly debated [47][48][49][50][51][52]. In practice, many [53,54], though not all [55], studies find that estimates of effect size amongst respondents consenting to follow-up/linkage differ very little from those found in the full population.…”