participants for your statistics to work out. In the case of Figure 7-6, 210 exposed and 390 unexposed
participants were enrolled.
In cohort studies, all the participants are examined upon entering the study, and those with the outcome
are not allowed to participate. Therefore, at the beginning of the study, all 600 of the participants did
not have the outcome, which is HTN. A cohort study is essentially a series of cross-sectional studies
on the same cohort called waves. The first wave is baseline, when the participants enter the study (all
of whom do not have the outcome). Baseline values of important variables are measured (and criteria
about baseline values may be used to set inclusion criteria, such as minimum age for the study).
Subsequent waves of cross-sectional data collection take place at regular time intervals (such as every
year or every two years). Changes in measured baseline values are tracked over time, and subgroups
of the cohort are compared in terms of outcome status. Figure 7-6 shows the exposure status from
baseline, and the outcome status from the first wave.
Because the exposure is measured in a cohort study before any participants get the outcome, it
is considered the highest level of evidence among the observational study designs. It is far less
biased than the case-control study design. Several measures of relative risk can be used to
interpret a cohort study, including the OR, risk ratio, and incidence rate (see Chapter 14).
Advancing to the clinical trial stage
Higher up the pyramid of evidence shown in Figure 7-2 are experiments. Not all experiments are at
such a high level of evidence — only high-quality clinical trials. These are experiments, not
observational studies. This is where the researcher assigns the participants to engage in a particular
behavior or intervention during the study. There are different types of clinical trials as described in
Chapter 5; however, the highest-quality trials use both double-blinding and randomization. Double-
blinding is where both the researcher and the participant do not know whether the participant was
assigned to an active intervention (one being studied), or a control intervention. Randomization is
where participants are randomly assigned to groups (so there is no bias in selecting participants for
each group).
It is possible to use a 2x2 table to analyze the results of a high-quality clinical trial as long as
the rows are replaced with the intervention groups. You can report the same measure of relative
risk as for a cohort study; however, the difference is that the high-quality clinical trial would be
seen as having much less bias than the cohort study — and stronger causal evidence.
Reaching the top: Systematic reviews and meta-analyses
Imagine a scenario where a new drug for HTN was developed, and several clinical trials were
conducted to see whether this drug was better than the most popular current drug used for HTN. How
would we be able to know whether, on balance, the new drug was actually better when we have so
many different clinical trials on the same drug with different results?
We could ask a similar question about observational studies as well. Imagine that multiple case-
control studies were conducted to determine whether having liver cancer was associated with the