Searching the medical literature for studies that have already been conducted is a key step in evidence generation – even for those who conduct primary research.
With research projects using existing data, too often the tendency is to browse the data that have been collected and ask “what kind of analysis could I conduct using this data set?” It sounds simplistic, but it is important to also consider what you want to know. What question(s) will, if answered, add to scientific knowledge on the subject? Then examine ways to answer that with existing evidence.
Let’s say your question is, “what treatments for lung cancer have the best chance of increasing long-term survival?” Once you have a research question you can begin to refine it and create a search strategy. A good way to examine and sharpen the aspects of a question is to use the PICO format, which stands for:
Population (or patient)
Let’s map out our example, which naturally brings up more questions:
P >>> patients with lung cancer (any type? Or is the diagnosis more specific?)
I >>> treatments (which treatments?)
C >>> compared to what? Standard care?
O >>> long term survival (how long is long term?)
The “P” term is critical, and if it’s tightly defined, sometimes this will be the only search term you need. In our example, what kind of lung cancer are we interested in? Perhaps it’s metastatic non-small cell lung cancer (NSCLC), or more specifically, metastatic NSCLC with an EGFR mutation.
The “I” in PICO often ends up being the easiest way to begin a search in clinical medicine, just because drug names and surgical procedures are well-indexed in most databases, and are often both unique and specific. Remember that scientific papers rarely use brand names, however, so be sure to look up any relevant generic names. You can start with good ol’ Google, or type brand names into the Drugs@FDA search to find their active ingredient(s).
C – If you are looking for evidence on how well an intervention works (“efficacy”), remember that the most rigorous studies are randomized trials comparing an intervention to a control group. For our example that might mean phase III trials of novel agents against the standard of care. It’s a good idea not to limit your search prematurely to a certain comparison, however, since your assumption about the proper control group or alternative treatment may not match published trials.
O – In our sample question, we are specifically interested in long-term survival, not necessarily progression-free survival or even median survival. This will definitely narrow down the articles we look at, since many studies don’t have a very long follow-up duration. However, because papers often do not report every outcome in the abstract, and the same endpoint may be called different names by different authors, it is rarely advisable to limit your search by outcomes terms. Exceptions might occur when an outcome is particularly infrequently studied, such as cost or specific quality of life instruments.
Here are a few more examples of how to use PICO:
At Frame Research, our name may be a happy coincidence, but it’s true: we have proven expertise in framing research questions – arguably the most critical part of any review. From there we develop a protocol, create and execute a search strategy, screen through all the citations to identify relevant papers, and finally collate and summarize the results. We use a systematic approach to each of these steps in order to minimize bias.