8. 8 Other types of studies
We refer to randomised, double-blind, placebo-controlled studies as the gold standard, but other types of studies are very common, and are often needed first in order to justify the expense of running a randomised controlled study.
Randomised controlled trial (RCT)
These are usually experimental and prospective, and compare two or more groups.
Randomisation is the most important factor, as it should make sure each group is similar at the start. The control group helps confirm whether a real effect is seen, rather than just happening by chance, or from other external factors.
All potential new drugs have to be studied in RCTs before they can be approved.
Cohort studies are usually observational and longitudinal.
They can either follow a group of people prospectively to see the incidence of whatever is being looked for or look backwards (retrospectively) to look for an effect.
They can also look at other related factors.
Cohort studies may include all patients at one or more hospitals (such as the MACS or WIHS cohorts in the US), or patients in one country (such as the UK-CHIC cohort), or can include international collaborations of national cohorts, such as the EuroSIDA or D:A:D studies in Europe.
Cohort studies can provide different types of results to an RCT. They can report on what happens in a regular clinic setting and in a wider group of patients than are usually selected for clinical trials.
People can be followed for longer, and they can look at more than one or two options.
However, because patients are not randomised to different treatments and know which treatment they receive (ie are not blinded), results have to be interpreted carefully to try to rule out other things that might explain the results (called ‘confounding factors’).
These studies are usually observational and retrospective.
A group of patients with a symptom (cases) is compared to similar patients without the symptom (controls) in order to try and identify what factors either caused the symptom or protected against it.
A case-control study could look at a group of people with lipodystrophy compared to a similar group (similar age, gender, duration of HIV infection, smoking status, etc) and see whether there was a pattern in different HIV drug use; or look to see whether genetic factors could be identified.
These studies are usually quick studies to look at the scale of a problem. For example, what percentage of a population are HIV-positive; or what percentage of people have lipodystrophy etc.
They can identify prevalence of an illness (how many people have a disease at any one point in time) but not the incidence of an illness (how many people will develop an illness over time).
The results from cross-sectional studies are limited by not being followed in time. We can see what is seen at and analyse what factors are related or associated to what is seen.
They cannot prove that one thing causes another or whether an intervention improves health.
Case study and case-note review
This is not a strong type of evidence, but can be used to collect data that might lead to other types of studies.
A case study is where an individual patient report is included as evidence.
Even though all sorts of other factors could have caused what was seen, case studies can alert researchers, doctors and patients to something new.
A case-note review is where a group of patient notes are reviewed retrospectively. The quality of the results are dependent on the quality of the data that was recorded.
Literature review and systematic literature review
A literature review can report collected results of selected studies. A systematic review has to include all relevant studies in the area being looked at.
This refers to comparing collected results from several studies.
These results have to be carefully interpreted as different studies usually involve different types of patients and it is not straightforward to just compare a final study result.
Both literature reviews and meta analysis are reliant on the types of studies that are published. For the results to be reliable you need to see the range of studies that were included in the analysis.
One problem is that studies that do not find a postive effect are often never published. This is an example of ‘publication bias’.
Last updated: 21 July 2009.