Dear RG members.
When planning a systematic review protocol, one have to build a PICO question.
In the case of systematic reviews of observational studies one can build a PICO question as follows:
P: Patients I(Risk): exposure/risk C: comparators O: outcomes.
If im going to assess risk of some X disease based on some Y exposure how do I operationalize/build a hard epi variable to be the outcome?
For example:
P: Adult patients
Risk: DM2
Comparator: not having DM2
Outcome: risk of bladder cancer...
How do I construct a statistic measure to be the risk of bladder cancer? I know one can use the RR or OR but, do you have any suggestions? Especially in the case of observational studies ??. In the case of observational studies can I make a patients meta-analysis?
I will appreciate your comments and feedback
Systematic reviews - Meta-analysis - Evidence based medicine - Observational studies - Modern epidemiology
Operationalization is the process of strictly defining a variable so that its measurable. In a meta-analysis, you operationalize your variables in your study protocol through inclusion and exclusion criteria. In order to operationalize your outcome as you have it here ("risk of bladder cancer"), you need to be much much more specific about it in a number of ways. I'm going into 2 specific ones:
First: what do you mean by bladder cancer?
There are a number of histopathologic types of bladder cancer, each with their own known associated risk factors, etiologies, and morbidity and mortality prognoses:
- Transitional cell carcinomas?
- Squamous cell carcinomas?
- Adenocarcinoma?
- Sarcomas?
- Neuroendocrine tumors of the bladder?
Further are you going to limit the stage, grade and severity of bladder cancer?
You need to have answers to these questions and definitely more I haven't thought of before you decide what exactly you mean by bladder cancer.
Second: what do you mean by risk?
Risk of disease occurrence is usually thought of in terms of incidence. Incidence proportion of a disease is the number of new cases of disease at a specific time interval divided by the population at risk. (Many people use the term risk to loosely refer to a measure of association [a.k.a. RR and OR], but this is inappropriate and it can confuse you when you try to operationalize a variable for a meta-analysis.)
To operationalize the risk in the outcome of "risk of bladder cancer" for meta-analysis you need to decide what effect measure you are going to extract from included studies. To decide which effect measure (a.k.a. RR, OR, rate ratio, population attributable fractions, prevalence ratios, standardized morbidity ratios, etc.) you need to think about a number of things: what type of data is the outcome measure (is it continuous or categorical data? does it represent incidence or prevalence? is it derived from retrospective case-control studies or prospective cohort studies? etc.)?
And more importantly for you: is the measure consistent across the studies and transferable?
If you find 10 studies for inclusion and 3 present relative risks, 2 present unadjusted odds ratios, 3 present adjusted odds ratios, and 2 presents rate ratios, what are you going to do with those measures to pool them? Can you pool them directly? Absolutely not. RRs, ORs, rate ratios each represent different types of data and have different real-world interpretations. You absolutely cannot pool them without addressing those differences. What if some of them report means and mean differences? or correlation data?
There are many formulas to convert between different measures (that can be found in the Cochrane handbook and in other locations), but you should pause and think before you go straight to statistics to solve problems of study design. Studies using different effect measures often differ in substantial ways in their designs. You need to carefully consider whether it makes sense to combine those studies in one analysis.
I've listed a number of online resources below. Good luck with your project!
MOOSE Guidelines for reporting meta-analysis of observational studies: Stroup DF, Berlin JA, Morton SC, Olkin I, Williamson GD, Rennie D, et al. Meta-analysis of observational studies in epidemiology: a proposal for reporting. Meta-analysis Of Observational Studies in Epidemiology (MOOSE) group. JAMA 2000;283:2008–12. (http://jama.jamanetwork.com/article.aspx?articleid=192614)
Cochrane Handbook for Systematic Reviews of Interventions (version 5.1.0). (http://www.cochrane-handbook.org/)
CDC's Principles of Epidemiology in Public Health Practice, 3rd Edition. http://www.cdc.gov/ophss/csels/dsepd/ss1978/index.html
Institute of Medicine. Finding what works in health care: standards for systematic reviews. March 23, 2011. (http://www.iom.edu/Reports/2011/Finding-What-Works-in-Health-Care-Standards-for-Systematic-Reviews.aspx)
Sutton AJ, et al. Methods for Meta-Analysis in Medical Research. Chichester:John Wiley & Sons, LTD, 2000.
(You can also find free online courses in public health, epidemiology and biostatistics on Coursera and EdX.)
Hi, Ramiro,
I think you got to look at the measure of risk in your finally selected studies. For example, see the attached article for you, the authors used hazard ratio (HR) as the measure of risk. You may also used RR or OR after looking at the details of attributes you can retrieve from those studies.
Victor
- Badges
- Science topic
- Similar topics
- Statistics
- Survey