OHDSI Home | Forums | Wiki | Github

LEGEND - HTN - 101 for beginners

Hello Everyone,

We are currently setting up OMOP CDM at our site and am slowly trying to familiarize myself with how studies are conducted and what methods are utilized etc.

Am not from a healthcare background but trying to pick up the required skills for CDM-based research by reading OMOP CDM related papers. Currently am reading the LEGEND-HTN paper and have few questions.

can help me with the below questions, please?

q1) I am quoting few lines from the paper - “LEGEND-HTN is …representing 4·9 million patients initiating monotherapy across nine databases from four countries, examining all pairwise comparisons between the five first-line drug classes against a panel of 55 health outcomes. This equates to 22 000 traditional observational studies”

May I know how does this equate to 22000 studies? Can anyone let me know how to arrive at this number? Its purely due to my limitation that I don’t understand this… Is it 9 (dbs)* 4 (countries)* 2 (pairwise) * 5 (first line drug classes) * 55 (health outcomes)?

But this comes to about 19800 only. Can help me how it 22000?

q2) Is it okay if I follow the same procedure (posting in the forum) to ask questions on OMOP-CDM related papers for educational purpose or should I only mail the authors? I felt posting it here can help others like me as well (who are new to OMOP CDM and healthcare)

Hi @Akshay, your second question first: the Forums is the PERFECT place to be having this discussion so that everyone can learn and benefit from the conversation, so keep the questions coming!

Q1: The Lancet paper highlights some of the key clinical results in LEGEND, but there are a few other papers in print (and even more working their way through the development and review process) that will help you understand this a bit further.

The one that I think will be most directly helpful for you is @schuemie 's recent paper in JAMIA, which goes through the LEGEND hypertension use case in greater depth. The link is here: https://academic.oup.com/jamia/advance-article/doi/10.1093/jamia/ocaa124/5895564.

You’ll see in Figure 1 the vast array of research questions that were addressed in this design. But note, the theoretical number of questions that could be answered is much larger than the observed number of questions for which we had sufficient sample size to yield an estimate.

The 22,000 studies cited in the Lancet paper comes from: (9 databases + 1 meta-analysis) * 10 pairwise comparisons between drug classes * 55 outcomes * 2 time-at-risk definitions * 2 propensity-score adjustment strategies.