Guideline-driven evidence generation opportunities

Clinical guidelines are extremely helpful, not only for providing guidance not only for providers on how to best treat their patients, but also for highlighting to the research community where there are open questions and evidence gaps. Some of these evidence gaps can be filled reliably through proper analysis of real-world data, as we aim to conduct across the OHDSI Evidence Network.

To help stimulate discussion and prioritize our community’s collaborative activities in 2025, I’m opening up this thread with a specific ask for anyone who is interested in leading or participating in an OHDSI network study, or simply are interested in seeing an OHDSI study conducted by the community on a topic that’s of interest to you:

Post a link to a current clinical guideline about a disease/topic of interest.
Review the guideline and share what evidence gaps you see that could be potentially filled with real-world evidence.

We’ll use the input and interest from the community on this thread to guide us toward collaborative evidence generation activities this year.

4 Likes

Treating obesity is receiving tremendous attention these days. The attached links are clinical guidelines for obesity management.

2013 AHA/ACC/TOS Guideline for the Management of Overweight and Obesity in Adults
2016 AACE/ACE Comprehensive Clinical Practice Guidelines For Medical Care of Patients with Obesity
2025 ADA guidelines: Obesity and Weight Management for the Prevention and Treatment of Type 2 Diabetes: Standards of Care in Diabetes–2025

However, comparisons of the effectiveness and safety of traditional obesity medications with newer ones are limited. ref1 ref2 This is likely the area where OHDSI excels the most.

Great idea @Patrick_Ryan . Are we open to screening, diagnostic, therapeutic monitoring guidelines too, or do we want to limit to therapeutic intervention guidelines, or maybe combinations.

:+1:, @Patrick_Ryan!

We’re interested in studying the necessity, efficacy and safety of interventions (e.g., medications) in frail or very old (say, 80+) individuals. These populations are typically poorly represented or ineligible in RCTs and, therefore, guidelines referring to them are vague, incomplete, or missing altogether.

We’re currently in the process of developing analysis tools that address these gaps in specific disease areas and treatment stages.

We’d love to hear your thoughts!

Thanks @Chungsoo_Kim for the link to the guideline and the suggestion to look at new vs old treatments. Definitely a good potential target, recognizing the challenges with continous capture of weight in our data network.

Hi @Gowtham_Rao , my personal opinion is that any part of a guideline that can be strengthened by reliable rwe should be fair game and a potential target for us. The main issue is whether we can observe the indication, the target and comparator, the outcome. Do you have a specific screening guideline in mind?

Thanks @cyanover , I have noticed that some guidelines include sections for special considerations for the elderly. Itd be nice to identify a guideline that you think needs more solid information for this vulnerable population (maybe as you say, elderly were excluded from rct that forms the basis of the current evidence) and we could seek to fill that gap together.

@Patrick_Ryan as a preventative medicine physician I would love to see adherence and impact of adherence to USPSTF (and similar guidelines ) across the network of OHDSI evidence network, stratified by calendar years to see temporal trends.

https://www.uspreventiveservicestaskforce.org/uspstf/recommendation-topics/uspstf-a-and-b-recommendations

Most of those measures have a T and O, and sometimes TAR (when we review why the committees recommended the screening). There are grades that need strengthening with new evidence. If we able to design an appropriate C, then we will have TCIO and TAR.

Anesthesia

The following recommendations in recent guidelines have been based on weak evidence. There are more evidence gaps in the guidelines, but it is quite hard to design an OHDSI study to tackle these gaps. CPS stands for clinical practice statements.

R5.3: We suggest that patients admitted for hospital care with a previous clinical–radiological diagnosis of SARS-CoV-2 infection that required intensive care unit, or high-dependency unit admission should go through more extensive cardiorespiratory preoperative evaluation (echocardiography, chest CT, cardio-pulmonary exercise testing (CPET). (CPS)

Background and existing evidence: click

Twenty-two prospective and retrospective cohort studies were included in evaluating preoperative assessment of patients with previous moderate-to-severe COVID-19 infection. Unfortunately, there were no randomised controlled studies, and there is a need for more trials, as the large variation in study designs significantly affects the quality of any recommendations. Moreover, specific recommendations in the first waves of COVID-19 infection are not applicable in the post-COVID era.

R12.8: SGLT2 inhibitors (SGLT2i) drugs should be withheld for 3 to 4 days before elective procedures to reduce the risk of euglycemic diabetic ketoacidosis. (CPS)

Background and existing evidence: click

However, alongside their benefits, SGLT2i also pose risks such as genitourinary tract infections, AKI and skeletal fractures. Additionally, they can predispose patients to ketoacidosis, often with blood glucose levels remaining relatively normal, hence the name eDKA. There have been numerous documented instances of eDKA occurring in perioperative patients, highlighting the importance of recognising this as a potential risk factor for individuals undergoing surgical procedures.

Guidelines: Preoperative assessment of adults undergoing elective noncardiac surgery

The role of pre- and postoperative B-type natriuretic peptides (BNP/NT-proBNP) measurement as a prognostic factor is still unclear.

The questions are typically formulated the following way:

Should preoperative B-type natriuretic peptides (BNP/NT-proBNP) be assessed as a prognostic factor before non-cardiac surgery?
Should preoperative B-type natriuretic peptides (BNP/NT-proBNP) be added to clinical scores for the prediction ofpostoperative events?
Should preoperative B-type natriuretic peptide (BNP/NT-proBNP)-enhanced management be implemented in non-cardiac surgery patients to improve outcome?

Guidelines: ESAIC focused guideline for the use of cardiac biomarkers in perioperative risk evaluation

R3.1: In patients undergoing surgery, we do not suggest the use of any drug as a prophylactic measure to reduce the incidence of POD.

No drugs known reduce the incidence rate of the delivery in the postoperative period, but the quality of evidence is low.

Guidelines: Update of the European Society of Anaesthesiology and Intensive Care Medicine evidence-based and consensus-based guideline on postoperative delirium in adult patients

Thanks @Gowtham_Rao , I agree that the USPTF issues important guidelines and it could be potentially use RWE to evaluate both concordance to those guidelines (via characterization analyses) as well as effectiveness (via population-level estimation analyses). Is there one or more specific recommendations that you would like to explore or see the OHDSI community investigate?

Thanks @zhuk , this is really great.

R5.3, interesting to see this recommendation and the note that there are already 22 cohort studies but no RCTs (it’d be a pretty odd RCT to randomize people into the ‘you cant have a more extensive cardiorespiratory preoperative evaluation’ arm :slight_smile: ) It’s a bit unclear to me how new RWE on this topic would impact care, but it seems extremely obvious how evaluating concordance to this recommendation could produce some sobering results.

R12.8, another interesting topic. I suspect it would be hard for us to observe drug non-use for the 3-4 days pre-surgery, but if that were feasible, seems straightforward to compare those who withheld vs. those who did not to see the incidence of outcomes.

R3.1, it’s somewhat surprising to me that there isn’t good evidence on prophylactic treatment for postoperative delirium, but potentially could be studied across OHDSI if we had defined exposures and some level of confidence that the outcome would be observed.

Thanks @Patrick_Ryan, for discussing this very interesting and bright idea. I go with what @Chungsoo_Kim mentioned especially the new standard of care of care guidelines (2025) https://diabetesjournals.org/care/article/48/Supplement_1/S6/157564/Summary-of-Revisions-Standards-of-Care-in-Diabetes . In these guidelines, there were some updated recommendations for the use of new dual GIP and GLP medications. Also, the role of SGLT2 inhibitors in MACE and HF. These might be discussed in the LEGEND studies but also can be studied further especially comparing combinations with mono therapy, or event safety profile.

Another area I guess it is not touched much is the depression treatments, a recent guideline
Canadian Network for Mood and Anxiety Treatments (CANMAT) 2023 Update on Clinical Guidelines for Management of Major Depressive Disorder in Adults discussed the different medications for depression. Treatment Guidelines: Depression | aapp.org
As you know, these medications have benefits but not sure if there were any comparison between different groups like SSRI, TCAs, or SNRI or others. Also, these medications have big concerns about their safety that also can be discussed. I am not really expert in these areas and I am sure there are who know better than me, but I thought I’d open this up for future discussion.

1 Like

Cancer guidelines are often large, intricate, and subject to frequent updates due to multifaceted nature of the disease. Cancer is not a single disease but a collection of hundreds of different conditions, each with its own biology, behavior, and response to treatment. This requires guidelines to account for a wide range of variables such as tumor type, stage, genetic mutations, and patient characteristics. Additionally, cancer treatment often involves multidisciplinary approaches, including surgery, radiation, chemotherapy, immunotherapy, and targeted therapies, further complicating the decision-making process.

Several organizations publish guidelines but let us focus on NCCN guidelines, these are step-by-step decision trees for each cancer type all the way down the possible decision points. Unfortunately, the guidelines are much less specific than one would expect. It’s full of sentences like:

  • “With the recent changes to first-line treatment options for metastatic disease, many providers are moving towards immune checkpoint inhibitor combinations, such as enfortumab - vedotin plus pembrolizumab, as a first-line treatment option. In this evolving paradigm, there is limited evidence to guide the optimal selection of second- and subsequent-line therapies following these new first-line regimens.” — NCCN Bladder Cancer

  • “Clinical trial enrollment is recommended by the NCCN Panel for all patients when appropriate, but is strongly recommended for second-line and subsequent therapies, as data for locally advanced or metastatic disease treated with subsequent-line therapy are highly variable.” — NCCN Bladder Cancer

  • “The optimal sequencing of systemic therapy and resection remains unclear.” — NCCN Colon Cancer

Bottom line: We have large and complicated decision trees, somewhat fuzzy, with each node its own constellation. Most of them are based on clinical trials. With respect to RWE, we could create evidence for each node about:

  1. To which degree are they applied in practice (adherence)?
  2. Are there real-world patients to create a feedback loop to effectiveness and toxicity outcomes (feasibility)?
  3. What are the effectiveness and toxicity (generalizability)?

This is a massive undertaking, and we already know that not all data are rich enough, large enough or longitudinal enough to support these. We can only find out after starting the job. I recommend focusing on two cancers-bladder cancer and colon cancer.

2 Likes

Thanks @Thamir , guidelines for use of antidepressants is an important topic given the uncertainty of effectiveness and patient preference around relative safety concerns about each product/class. Do others in the community have expertise or experience for how this (or other) guidelines for depression are used in practice?

Thanks @agolozar , this seems like a very exciting opportunity. It’d be great to see our community tackle this area, not just within the Oncology WG but across the entire Evidence Network, so that we can learn where real world data can be fit-for-use for which questions. It’d also be a big help if we can push the oncology community to recognize and use real world evidence more actively.

Clinical guidelines vary widely in how heavily they weigh evidence vs professional opinion and guild incentives in their recommendations. We should focus on filling evidence gaps identified in guidelines that base their work on high quality systematic evidence reviews. The USPSTF publishes detailed supplements that detail the evidence reviewed, the relative strength and high priority evidence gaps that need research. So, I agree with those being called out.

We identified oncology evidence gaps identified by the FDA at the bottom of this page under Collaboratons.

I led or co-led several grants quantifying adherence to cancer, cardiovascular, and diabetes prevention and treatment guidelines and estimating the impact on incident disease associated with adherence. We could definitely do that in OHDSI. Doing so in a reproducible way will require a new structure for codifying guideline conformance. That is often complex because decisions based on results and the timing of next steps can be elaborate. I’d love to collaborate with other who want to work on that.

A general conclusion from the work in those grants is that the connection between guideline adherence and outcomes was much greater in management services than in preventive services.

An interesting methodological challenge for reproducing and improving on that work will be to develop a better solution than what I devised for propensity score adjustment when adherence (treatment) varies by degrees rather than by kind of intervention. I used an approach for that proposed by Imai. But it’s generally uncertain territory still and ripe for improvement.

1 Like

Hi All,
Great suggestions so far. To add to these suggestions I am interested in a couple of different issues:

  1. evidence gaps around guidelines for secondary prevention after stroke. Living guidelines in Australia suggest that some issues that require further evidence include concordance with appropriate medicine use after stroke and compliance with that treatment, eg CV medicines (Blood pressure, antiplatelets, AF, cholesterol lowering treatment) and HRT/oral contraception). I think this is one area where there is a real need for robust study designs that can account for immortal time and confounding by indication when understanding the effectiveness and safety of treatments post stroke.

  2. A recent editorial in JACC highlighted the ongoing issue of clopidolgrel (and other antiplatelet therapy) after coronary artery stenting with Drug-eluting stents). Clopidogrel for Long-Term Secondary Prevention After Coronary Artery Stenting∗ | Journal of the American College of Cardiology. The issues here are that guidelines for length of time on treatment and use of dual therapy have changed over time supported by clinical trials however as pointed out by the editorial “clinical trial patients were highly selected, with a low-intermediate ischemic risk, and a high use of intracoronary imaging, and thus may not be representative of routine clinical practice”. I think OHDSI could make a significant contribution here!. First, I would love to understand exactly how long patients are remaining on antiplatelet treatment after DES and how this has changed over time. Second, how may use DAT and lastly, what is the consequence of length of treatment on outcomes (balancing of bleeding/stent thrombosis, myocardial infarction) and in specific subgroups at high risk.

Hello everyone,
here are some additional thoughts in the field of psychiatry.

The field may not be as rapidly developing as some other medical disciplines, and many of the guidelines have aged quite a bit.
On the topic of treatment of the first psychotic episode, national and associations’ guidelines agree on the fact that ‘The inconsistency of findings argues against established clinical superiority for a specific antipsychotic in first-episode schizophrenia or, in fact, antipsychotic class (i.e., second-generation antipsychotic [SGA] vs. first-generation antipsychotic [FGA])’ (from the Canadian Guidelines for the Pharmacotherapy of Schizophrenia in Adults, 2017).
Similar conclusions are made in the APA guidelines (2020), mentioning significant heterogeneity of clinical trial designs and limited number of head-to-head comparisons, as well as in the BAP (2019) and HPFT (2023) guidelines.

So, head-to-head comparisons of SGAs and FGAs based on their effectiveness and safety is something that can help bridge this knowledge gap.
As for the possible groups of interest, one option could be to look at the first-episode psychosis and rehospitalizations as a measure of relapse prevention.