BICYCLE HELMET

RESEARCH

FOUNDATION

cyclehelmets.org

Home page

Main topics

News Headlines

Frequently asked Questions

For Policy Makers

Research evidence

Misleading claims

Helmet laws

Analysis

Search Engine

Australia

Canada

New Zealand

UK

USA

Other countries

Full index

Links

BHRF

Policy statement

The Medical Research Library of Brooklyn provides these helpful explanations of medical research methods:

Randomized controlled studies

Case-control studies

Cohort studies

Odds ratios .

An explanation, by Rebecca Goldin, of one of the most difficult statistical concepts to grasp, and one which many helmet researchers have not properly understood.

One goal of a meta-analysis will often be to estimate the overall, or combined, effect of its constituent studies (such as, the protective effect of bicycle helmet use on head, neck and facial injuries).

If all studies in the analysis were equally precise, it would be possible to simply compute the mean of the individual effects observed in each study. However, if some studies were more precise than others, more weight should be assigned to the studies that carried more information. This is what should be done in a meta-analysis. Rather than compute a simple mean of the effect sizes, a weighted mean is computed, with more weight given to some studies and less weight given to others.

There are two models used in meta-analysis that determine how the weights are assigned: the fixed effect model; and the random effects model. The two make different assumptions, which lead to different mechanisms for assigning weights.

Under the fixed effect model, it is assumed that there is only one true effect size, which is shared by all the included studies.

By contrast, under the random effects model, it is acknowledged that the true effect could vary from study to study; for example, the effect size might be larger if the subjects are older. This is the case in Elvik, 2011, which demonstrates a time trend in the protective effect of helmets found in different studies. The effect is largest in the earliest studies, but progressively falls over time. By taking account of this time trend, Elvik, 2011 concludes that in the most recent studies, helmets have no overall protective effect, since reduced head and facial injuries are matched by increased neck injuries.

**Publication bias**

The tendency not to publish studies if ﬁndings are not statistically signiﬁcant or contradict prior expectations or the vested interests of sponsors of the research.

**Time-trend bias**

This refers to a tendency for study ﬁndings to change over time; if all ﬁndings are pooled independently of when they were published, the trend will be pasted over and the summary estimate of effect will be misleading.

**Zero-count bias**

Bias arising if studies with zero counts are omitted or if inefficient continuity corrections are applied to such studies.

Elvik R, 2011. Publication bias and time-trend bias in meta-analysis of bicycle helmet efficacy: A re-analysis of Attewell, Glase and McFadden, 2001. Accident Analysis & Prevention 2011;43(3):1245-1251.