Guideline-driven evidence generation opportunities

I’d like to contribute to this important discussion by highlighting a critical evidence gap in the treatment of acute coronary syndrome (ACS) patients undergoing percutaneous coronary intervention (PCI). Current clinical guidelines provide detailed recommendations on the use of antithrombotic agents (e.g., aspirin, P2Y12 inhibitors) post-PCI. However, there remain unanswered questions that could be addressed with high-quality real-world evidence.

Prior landmark trials such as PLATO (2009) and TRITON-TIMI 38 (2007) established the efficacy of ticagrelor and prasugrel, respectively, compared to clopidogrel, with prasugrel demonstrating superior efficacy at the cost of increased bleeding risks. The 2021 ACC/AHA guidelines, based on these trials, include the following recommendations:

In patients with ACS undergoing PCI, it is reasonable to use ticagrelor or prasugrel in preference to clopidogrel to reduce ischemic events, including stent thrombosis (COR 2a, LOE B-R)

In patients undergoing PCI who have a history of stroke or transient ischemic attack, prasugrel should not be administered (COR 3, LOE B-R)

However, direct comparisons between prasugrel and ticagrelor remain controversial. PRAGUE-18 (2016), which demonstrated comparable composite outcomes between prasugrel and ticagrelor, was an underpowered study, limiting its reliability. ISAR-REACT 5 (2019), a sufficiently powered study that demonstrated prasugrel’s superior efficacy over ticagrelor with comparable bleeding risks, also faced criticism for potential bias.

However, the 2021 ACC/AHA guidelines did not include a recommendation on the direct comparison between ticagrelor and prasugrel due to the controversy surrounding ISAR-REACT 5, while its findings were incorporated into the 2023 ESC guidelines, in the recommendation below. These contradicting guidelines highlight a major evidence gap in this subject.

Prasugrel should be considered in preference to itcagrelor for ACS patients who proceed to PCI (Class IIa, Level B)

Furthermore, another controversy related to this subject recently flared up after the BMJ openly expressed doubts over the PLATO study (https://doi.org/10.1136/bmj.q2550). Both ACC/AHA and ESC guidelines’ current recommendations prioritizing ticagrelor are based on this trial. With the study now being directly challenged, the foundation for these recommendations may be called into question, potentially shifting the treatment landscape.

I believe a well-designed OHDSI study could provide valuable real-world evidence to address this issue.

1 Like

Wonderful initiative!

A study I wanted to do for a while but lacked the right push :slight_smile::

We know there is a huge variation in how community-acquired pneumonia is treated, but nobody ever quantified it/looked at adherence to guidelines at scale. Large-scale treatment pathways is something OHDSI is well equiped to do and is generally a lighter lift.

A gap in guidelines associated with it is lack of knowledge on effectiveness of glucocorticosteroids (for which I think you first need to quantify how many patients receive corticosteroids for pneumonia). Ref1 and ref2.

Thanks @nicolepratt !

  1. Can you provide the link to which guideline for stroke prevention is commonly referenced in Australia?

  2. I agree, characterizing antiplatelet use duration after DES seems a straightforward yet useful analysis, and could help to guide subsequent analyses about impact of duration on outcomes.

Thank you @TatianaSkugarevskaya ! I think old guidelines are really good targets to go after, because presumably practice should be aware of them, and we should have real world data for guideline-concordant care that we can use to assess effectiveness. And head-to-head comparisons of antipsychotics for some defined outcomes, (like rehospitalization and safety outcomes) seems quite aligned to how we’ve done prior studies following the LEGEND principles.

1 Like

Thanks @SCYou ! I agree this sounds like a great opportunity to help resolve an active controversy. Given your important role with JACC, do you think a publication of a proper RWE study that provides explicit comparative effectiveness and safety prasugrel vs. itcagrelor could inform the next iteration of the ACC/AHA and ESC guidelines?

Thanks @aostropolets. I definitely agree that a treatment pathway to characterize pneumonia treatment seems like a reasonable analysis that many data partners should be able to contribute to.

As it relates to effectiveness of pneumonia treatments, what outcomes would you consider relevant to inform guidelines? (I’m wondering whether we think showing reduced hospital length of stay is an acceptable proxy for infection cessation)

That’s a great question. I think I wouldn’t even go to com effectiveness and stop at pathways to, among others things, compute the # of ppl on gcs with antibiotics. One reason for it is that for comparative effectiveness we would need to look at antibiotic A vs Antibiotic A + gcs, which means that we need to know which antibiotics A, B … E are the most common to be studied individually. Plus, there seems to be more methodological considerations there…

It feels that this is something @Christophe_Lambert and Psychiatry WG would be interested in participating :slight_smile:

Similar to @nicolepratt post. I think it would be great to see time varying treatment pathways as guidelines change over time. Treatment pathways are great, an ability to look at treatment pathways changing over time is next level. Treatment pathways with duration of exposure and time to switch would add to interpretation and of course having the ability to monitor these changes over time would add to the exposure target (T) stories we tell.

As an aside when we do our comparative effectiveness studies for either benefit or safety many times our comparators (C) suggested are therapies that would be guideline recommended earlier in disease course. These are not good Cs for our Ts, as we are not in equipoise. If insurance coverage forces us through C before we get to prescribe T that’s not a good C for T. Visualizing within an Indication (I) these Ts and Cs, the time spent exposed to these Ts and Cs, and how these change over time particularly as treatment guidelines change over time would aid in appropriate comparator selection.

Then all of this not only varies over time but also space (geography). Thus we need a Facilitating Longitudinal Utilization aX (across) resources, a FLUX Capacitor if you will to treatment pathways.

The immunology space has several agents and several changing guidelines over time. UC and CD are copied below but there are several in this space.
https://www.gastrojournal.org/article/S0016-5085(20)30018-4
https://www.gastrojournal.org/article/S0016-5085(21)00645-4

Hi all - this is a great thread!

In the autoimmune/rheumatology space, two areas where numerous clinical knowledge gaps exist are (1) risk stratification/prophylaxis for infectious complications and (2) vaccination strategies. Guidelines are regularly published by US (American College of Rheumatology) and Europe (European Union League Against Rheumatism). I imagine many data sources contain the relevant information to address at least some of these gaps: the exposure (autoimmune disease itself), risk factors (most of which are medications) and outcomes (diagnosis of infection).

I’ve included two such guidelines below. They both include ‘research agenda’ questions (Box 1 in EULAR guidelines and Box 7 in ACR guidelines).

Infectious Disease:
2022 EULAR recommendations for screening and prophylaxis of chronic and opportunistic infections in adults with autoimmune inflammatory rheumatic diseases | Annals of the Rheumatic Diseases

Some example research questions:

  • Does the risk of opportunistic and chronic infections differ between the different classes of disease-modifying antirheumatic drugs (DMARDs) or immunosuppressive drugs?
  • Is screening and prophylaxis for opportunistic and chronic infections in people with AIIRD receiving antirheumatic therapies cost-effective?

Tuberculosis

  • Should patients starting immunosuppressants (eg, cyclophosphamide) be screened routinely for latent tuberculosis (TB)?
  • How often should patients who have already been tested for tuberculosis, be rescreened? In relation to that, is there a need to rescreen patients who switch biological DMARDs or targeted synthetic-DMARDs?

Hepatitis

  • When should hepatitis antiviral treatment be started in people living with AIIRD commencing antirheumatic treatment found to be at risk of hepatitis reactivation?
  • For how long should hepatitis antiviral prophylaxis be continued in patients at risk for hepatitis reactivation after antirheumatic treatment is stopped?

Other viruses

  • Is it safe to treat people living with HIV with antirheumatic treatments?
  • When should antiviral prophylaxis be considered in people with AIIRD who have recurrent herpes zoster infections?
  • Is postexposure prophylaxis for patients non-immune to VZV who are exposed to VZV beneficial?

Pneumocystis jirovecii pneumonia (PCP)

  • Does the risk of Pneumocystis jirovecii pneumonia (PCP) differ according to underlying AIIRD (eg, giant cell arteritis, lupus, ANCA-associated vasculitis, etc)?
  • What is the added risk of PCP in patients treated with combination glucocorticoids/immunosuppressive therapies compared to those receiving glucocorticoids alone?
  • What is the safest and most effective regimen for PCP prophylaxis?
  • How long should patients at risk for PCP receive prophylaxis?

Vaccination:
2022 American College of Rheumatology Guideline for Vaccinations in Patients With Rheumatic and Musculoskeletal Diseases - Bass - 2023 - Arthritis & Rheumatology - Wiley Online Library

@Benjamin_Martin @wkelly19 @Azza_Shoaibi @Gowtham_Rao @Joel_Swerdel

1 Like

In terms of guidelines, I like to use phenotypes as trigger part of the Clinical Decision Support rules. The phenotype will reveal how many people trigger a given rule based on various versions of the trigger logic.

I like to focus on consumer and PHR (personal health record) focus. Provide rules for consumers.
Because alerting clinicians is great but alerting patients sometimes may work better.

The problem with guidelines that if we do anything in OMOP, it needs translation into FHIR and CDS Hooks and EHRs
I spent 4.5 years of my PhD doing guidelines.

E.g.,

  • if you wear glasses, see ophthalmologist every x years. (or prevent glaucoma if history of increased intraocular pressure)
  • if you had high cholesterol, you should have a repeated measurement of your cholesterol level
  • if you had 3+ high blood pressure results at some point, are you managing it long term correctly
  • if you had low vitamin D and took vitamin D supplement, check again your level several years later
  • (this could be generalized to any abnormal lab test history that needs later monitoring, e.g., thyroid function tests)
  • Barrett’s esophagus prevention (repeat endoscopy) if history of abnormal findings found in endoscopy.

CDS inspired phenotype to show the value of OMOPed data and power of structured data.
Target group is healthcare consumers.

So not truly provide evidence for guideline improvements.

@Kevin_Haynes , I definitely support treatment pathway over time to monitor progress toward adherence to guidelines. It does make me wonder how to evaluate that the guideline changes over time and so too does practice, with some clear lag in between.

The link you posted to the guideline didn’t work for me, could you please repost so that I can take a look?

Thanks @Christopher_Mecoli , I remember at our Rheumatoid Arthritis study-a-thon in Barcelona in 2019 that the guidelines were a clear motivator for evidence needs about comparative safety of DMARDs, and still seems opportunities to continue to contribute there (tagging @Daniel_Prieto @jweave17 ).

Specific subpopulations of interest are another area where RCTs often fail to deliver evidence, yet RWE can fill the gap, so I like the thought of exploring treatment effect heterogeniety in particular vulnerable populations, like persons with HIV who need an immunosupressive drug for their RA.

@Vojtech_Huser , I’m not suggesting that we aim to implement guideline decision support in clinical practice. I’m positing that the guidelines can provide us inspiration for very specific evidence gaps that, if we can produce reliable evidence, could be filled within our community. So I’m much less ambitious than trying to model a full guideline, I’d be quite happy if we take one single recommendation that we think has impact and could be strengthened with better evidence, and make it happen.

AGA Clinical Practice Guidelines on the Management of Moderate to Severe Ulcerative Colitis - PubMed UC
AGA Clinical Practice Guidelines on the Medical Management of Moderate to Severe Luminal and Perianal Fistulizing Crohn's Disease - PubMed CD

I think inflammatory conditions such as UC and CD lend themselves to treatment pathway analyses to characterize guideline adoption over time. Such an analysis could build the phenotypes of drug treatment pathways to evaluate the evidence gaps noted in the guidelines. Only through treatment pathways will we even be able to identify the number of patients that look like this: “In adult outpatients with moderate to severe CD, the AGA makes no recommendation regarding the use of, ustekinumab or vedolizumab in combination with thiopurines or methotrexate over biologic drug monotherapy for the induction and maintenance of remission.”

Hello OHDSI Community, a warm greeting from the SciForce team!

Thanks, @Patrick_Ryan, for organizing this discussion, and to @agolozar for bringing up the challenges and limitations of the NCCN recommendations.

One related area where real-world evidence could be valuable is determining the best duration for PD-1/PD-L1 inhibitor therapy.

Checkpoint inhibitors such as pembrolizumab, nivolumab, atezolizumab and durvalumab have changed the treatment landscape in oncology, but their prolonged use poses problems.

  1. Medical problem: Prolonged therapy increases the risk of severe autoimmune adverse events that can be life-threatening and require significant medical interventions. And sometimes can significantly impair quality of life if these side effects require lifelong insulin therapy or thyroid hormone replacement therapy.
  2. Economic challenge: These drugs are expensive, and prolonged treatment can be a significant financial burden for patients, health systems and insurers.
  3. Social problem: For many patients, completing a full 1-2 year course of therapy is simply not feasible for a variety of reasons.

NCCN guidelines generally recommend up to 1-2 years of therapy for various cancers, such as melanoma, NSCLC, renal cancer, high-risk triple-negative breast cancer etc. However, there is no robust evidence to support the idea that all patients benefit equally from this duration. Real-world cases indicate that many patients stop sooner because of side effects, cost, or stable disease. This is raising the question: could a shorter course work just as well?

Given that there are many patients in real-world clinical practice who have not completed a full long-term course of immunotherapy (some due to toxicity, some due to financial considerations, some who have run out of insurance, etc.), their data can be included in the RWE study.

This provides an opportunity for the OHDSI community to study progression-free survival, overall survival, and toxicity rates in patients receiving shorter (3 or 6 or 9 months) and longer treatment (1-2 years). We can also study subgroups by cancer types (I prefer to focus on melanoma and lung cancer), by therapy settings (adjuvant or metastatic), or by medical history and baseline condition (such as patients with minimal disease or comorbidities), and also find out whether biomarkers such as ctDNA or PD-L1 expression help to personalize therapy.

We can also clarify the clinical and economic implications of long-term immunotherapy, which will help generate more patient-centered and cost-effective treatment recommendations.

Looking forward to the discussion!

1 Like

Wonderful topics!!! I can see some methodological challenges that one needs to consider for some of them. For example, PCP relies on microbiology data, which is not always converted to OMOP and when it is I know the conversion is done differently across sites. Similarly, TB testing is usually poorly captured as helath deaparments do that and one usually needs to have access to health exchange data to get good capture.

Hello Patrick,

Thank you for opening up this space for discussion and collaboration on evidence generation. One gap that I find particularly important in clinical guidelines, especially in the context of acute heart failure, is the need for a better definition and classification of patient phenotypes.

Accurate classification of phenotypes can facilitate more targeted and effective treatments, allowing healthcare professionals to tailor interventions based on the specific characteristics of each patient. This would not only improve the quality of care but could also lead to better clinical outcomes.

This gap is highlighted in the 2021 ESC Guidelines for the diagnosis and treatment of acute and chronic heart failure. Additionally, the article titled “Machine Learning Methods Improve Prognostication, Identify Clinically Distinct Phenotypes, and Detect Heterogeneity in Response to Therapy in a Large Cohort of Heart Failure Patients” attempts to address this phenotyping issue.

I believe that real-world data analysis could be a valuable tool in addressing this gap, helping to identify patterns and characteristics that are not fully captured in current guidelines. I am interested in participating in an OHDSI network study that explores this issue.

I look forward to the community’s contributions and the opportunities for collaboration in 2025!

Best regards,
João Silva

1 Like

Thank you @Bohdan_Khilchevskyi for joining the conversation, the topic of duration of treatment is quite interesting, Looking at event rates as a function of duration of exposure seems like a reasonable piece of evidence that RWE would be better equipped to characterize than RCTs or other data sources.

Thank you @JoaoSilva , I agree that proper discrimination of clinical subtypes within a disease, like acute heart failure, is a good phenotyping problem and real world data offer the opportunity for us to characterize the heterogeneity within the disease and subpopulations. How are you thinking this evidence would inform the 2021 ESC Guideline?