# [R-meta] do I need to do meta analysis and how to perform meta analysis for observational studies

Yuan Chun Ding ycd|ng @end|ng |rom coh@org
Mon Jun 24 18:33:27 CEST 2019

```Dear meta-analysis users,

I have collected raw data from five observational studies, so a total of 1300 observations or patients, then for each patients there are 17,000 variables, since all five studies used the same method to measure gene expression values for over 16900 genes (plus some phenotypic variables, 17000 variables in total). Then I normalized gene expression data for all 16900 genes, and then calculated an immune score for each patient based on gene expression values for those  16900 genes.   I need to run a multivariable logistic regression to test association between the estimated immune score and a cancer development phenotype by adjusting for  covariates such as age, race etc.  My first question is :   do I really need to use the meta analysis approach?  I originally just ran multivariable logistic regression by considering all 1300 patients are from one combined study (patient phenotype (yes or no) = immune score + 6 covariates, so estimating adjusted OR).  Then I thought about including study ID as a  random factor in the model or calculate OR for each of five study, then pool the five OR to get an overall OR.  What do you think?   The second question:  from reading meta analysis paper,  the following R code to run meta analysis was design to run  clinical trials (randomized design) with confounding variables controlled in each study, so only univariate logistic regression analysis required, no need to adjust covariate; so how to calculate overall adjusted OR by meta analysis?  if I calculate adjusted OR for each study, I will get " Estimate" (for OR) and  "Std. Error" in each study. if using metafor, I guess I need to convert OR to logOR (Y in the example code below), then standard error to variance (V), is that right?  also how to convert from standard error in each study to variance?

Y <- with(dat.bcg, log(tpos * cneg/(tneg * cpos)))
V <- with(dat.bcg, 1/tpos + 1/cneg + 1/tneg + 1/cpos)
cbind(Y, V)

result.or.FE <- rma(yi = Y, vi = V, method = "FE") # Log Odds Ratio
result.or.FE

result.or.DL <- rma(yi = Y, vi = V, method = "DL")
result.or.DL

Thank you very much in advance,

Yuan Ding from City of Hope National Medical center

------------------------------------------------------------
-SECURITY/CONFIDENTIALITY WARNING-

This message and any attachments are intended solely for the individual or entity to which they are addressed. This communication may contain information that is privileged, confidential, or exempt from disclosure under applicable law (e.g., personal health information, research data, financial information). Because this e-mail has been sent without encryption, individuals other than the intended recipient may be able to view the information, forward it to others or tamper with the information without the knowledge or consent of the sender. If you are not the intended recipient, or the employee or person responsible for delivering the message to the intended recipient, any dissemination, distribution or copying of the communication is strictly prohibited. If you received the communication in error, please notify the sender immediately by replying to this message and deleting the message and any accompanying files from your system. If, due to the security risks, you do not wish to receive further communications via e-mail, please reply to this message and inform the sender that you do not wish to receive further e-mail from the sender. (LCP301)
------------------------------------------------------------

[[alternative HTML version deleted]]

```