"Negative" results in cohort studies--how to recognize fallacies

Abstract

Negative studies are important in occupational medicine because knowledge on noneffect levels for harmful exposures is pertinent. A truly negative study must (i) be large, (ii) be sensitive, and (iii) have well-documented exposure data. Small and/or insensitive so-called negative studies are uninformative, and negative results can only be related to the actual or lower exposure levels. Some of the causes for falsely negative results are inappropriate design (eg, cross-sectional instead of longitudinal), crude measuring methods, inappropriate type of examination, wrong categories of exposed workers, inclusion of workers with too short an exposure time and too low an exposure intensity in the exposed series, too short a follow-up for diseases with long latency times (eg, cancer), incomplete follow-up, wrong reference category (eg, the general population), poor precision of measuring methods, and insensitive or wrong statistical methods. Finally, the same data may be interpreted in different ways. Correct interpretation requires both knowledge of the subject and apprehension of the fact that errors in design and measurement often tend to mask existing differences. At the most, a small insensitive study may rule out very strong effects.