The pharmaceutical industry spends approximately $157 million per pivotal cardiovascular Phase III trial, with timelines averaging 4–7 years from protocol design to regulatory submission. Meanwhile, the FDA's December 2025 guidance removed the requirement for identifiable patient-level data in real-world evidence (RWE) submissions, and the Inflation Reduction Act (IRA) Medicare Drug Price Negotiation Program is expanding to 25 drugs by 2028. The convergence of these forces has created an unprecedented demand for regulatory-grade causal evidence from observational data.
Yet submissions continue to fail. The FDA has explicitly rejected RWE due to “major methodological issues including immortal time bias, selection bias, misclassification, confounding, and missing data.” The gap is not in the data — it is in the methodology. Companies that master causal inference methods for RWE stand to capture a cost advantage of 60–90% per indication compared to running additional Phase III trials, while reducing timelines from years to months.
This white paper presents a decision framework — the Causal RWE Playbook — that maps the three highest-value regulatory use cases to specific causal architectures, identifies the five biases that kill FDA submissions, and provides an operational methodology selection algorithm. It draws on analysis of 218 FDA labeling expansions (2022–2024), published regulatory rejection rationales, and methodology comparisons from a real-world cardiovascular cohort of 24,788 participants on the All of Us Research Platform.
A single well-designed causal RWE study costs $2–5M and takes 6–12 months. The equivalent Phase III trial costs $50–157M and takes 3–7 years. For label expansions, post-marketing confirmatory evidence, and IRA value demonstration, causal RWE is not a compromise — it is a strategic advantage. But only if the methodology survives regulatory scrutiny. This playbook shows how.
1. The $157 Million Problem
Cardiovascular disease remains the leading cause of death worldwide, accounting for 17.9 million deaths annually. The therapeutic market is massive: the top 10 IRA-negotiated drugs include multiple cardiovascular agents — Eliquis, Jardiance, Xarelto, Entresto — each generating billions in annual revenue. Yet the evidence base for expanding these drugs to new populations, confirming long-term effectiveness, and demonstrating real-world value to CMS faces a structural bottleneck.
1.1 The Cost Reality
For context, a well-designed causal RWE study using de-identified EHR or claims data costs approximately $2–5 million and can be completed in 6–12 months. The cost differential is not marginal — it is an order of magnitude.
1.2 The IRA Pressure Multiplier
The Inflation Reduction Act has fundamentally changed the economic calculus. Under the IRA, CMS now negotiates prices for high-expenditure single-source drugs. For the 10 drugs selected in the first cycle, manufacturers must justify pricing based on clinical benefit, therapeutic alternatives, and real-world effectiveness. The program expands to 15 additional drugs in 2028, including Part B physician-administered drugs.
This creates a new category of evidence demand: IRA value demonstration. Manufacturers must prove that their drug delivers differentiated clinical benefit in real-world populations — not just the idealized trial cohort. The evidence that supports this demonstration is, by definition, real-world evidence. And the methodology that makes this evidence credible is causal inference.
CMS considers “the extent to which the drug addresses unmet medical need” and “data on comparative effectiveness.” For drugs facing price negotiation, every percentage point of demonstrated real-world effectiveness translates directly into negotiating leverage. A causal RWE study showing 15% MACE reduction in a real-world population could be worth hundreds of millions in preserved pricing over a 3-year negotiation cycle.
2. The Five Biases That Kill FDA Submissions
Analysis of FDA review documents for RWE-containing submissions reveals a consistent pattern of methodological deficiencies. In one documented case, the FDA stated: “Due to major methodological issues (including immortal time bias, selection bias, misclassification, confounding, and missing data), the FDA does not consider these results adequate to support regulatory decision making.” These five biases are predictable and diagnosable — but each demands a distinct, case-by-case causal methodology to address.
Immortal time bias is perhaps the most insidious. It inflates treatment benefit by misclassifying pre-treatment survival time as time spent under treatment. When a study defines treated patients based on a future event — such as filling a prescription — it guarantees that those patients survived long enough to receive the drug. The treated group appears to have lower early mortality, but the advantage is entirely artifactual. The FDA has rejected submissions specifically citing this bias.
Confounding by indication operates in the opposite direction. In cardiovascular prevention, sicker patients are more likely to receive aggressive treatment — statins, polypills, combination therapy. This creates a paradox where treated patients have worse outcomes, not because the drug is harmful, but because they were sicker to begin with. Crude and even partially adjusted analyses can make an effective drug look dangerous. Addressing this requires careful study design choices that vary depending on the treatment strategy, comparator, and data structure involved.
Selection bias arises when loss to follow-up is non-random. Patients who drop out of care — or out of the dataset — often do so for reasons related to their health trajectory. If attrition differs between treatment arms, the surviving population no longer represents the original cohort. The FDA has flagged differential selection between trial populations and real-world data sources as a specific concern in RWE submissions.
Misclassification is the silent corruptor of claims and EHR data. Inaccurate coding of exposures or outcomes — particularly composite endpoints like MACE — introduces measurement error that can bias results toward or away from the null. When outcome rates in your data diverge from known population incidence, or exposure prevalence looks inconsistent across sources, misclassification is the likely culprit. The appropriate correction depends on the nature of the error and must be evaluated on a study-by-study basis.
Time-varying confounding is the most methodologically challenging of the five, and the one that standard analytical approaches are least equipped to handle. When a covariate changes over time and is simultaneously affected by prior treatment — for example, blood pressure that responds to antihypertensive therapy but also influences future prescribing decisions — standard regression cannot be used without inducing collider bias. Adjusting for the covariate blocks the causal pathway; not adjusting introduces confounding. Breaking this cycle requires advanced longitudinal causal methods that go well beyond conventional multivariable adjustment.
Each of these five biases requires a tailored methodological response — there is no single correction that addresses all of them simultaneously. The appropriate causal architecture depends on the specific bias structure, data availability, treatment strategy, and regulatory context of the submission. Identifying which biases threaten your study — and selecting the right combination of design and analytic choices to address them — is the core challenge that separates regulatory-grade RWE from evidence that fails on review.
3. Three Use Cases Where Causal RWE Generates Maximum Value
3.1 Label Expansion (sNDA/sBLA)
Between January 2022 and May 2024, the FDA granted 218 labeling expansions for drugs and biologics. Of these, 25.2% had available real-world evidence in FDA documents or the published literature. RWE was most commonly found in oncology (43.6%), infection (9.1%), and dermatology (7.3%). In 76.5% of cases where RWE served as primary evidence, no new clinical trial was required.
The opportunity in cardiovascular medicine is largely untapped. For drugs already on the IRA negotiation list, a well-executed label expansion supported by causal RWE simultaneously (a) expands the addressable market, (b) strengthens the CMS negotiation position, and (c) demonstrates value to payers — at a fraction of the cost of a new trial.
3.2 Post-Marketing Confirmatory Evidence
Under the PDUFA VII framework, the FDA's Advancing RWE Program explicitly seeks to improve the acceptability of RWE-based approaches for meeting post-approval study requirements. For drugs granted accelerated approval, confirmatory evidence requirements can cost $50–100M in traditional Phase IV trials. A causal RWE study using target trial emulation can provide equivalent confirmatory evidence at $3–8M with appropriate sensitivity analyses.
3.3 IRA Value Demonstration
This is the emerging frontier. As CMS negotiates maximum fair prices for selected drugs, manufacturers must present evidence of real-world clinical benefit. The IRA mandates that CMS consider comparative effectiveness, unmet medical need, and data on the drug's impact on specific populations. Causal RWE is the only methodology that can generate this evidence at scale, across diverse populations, in the timeframes required by the negotiation calendar.
For cardiovascular drugs, the stakes are enormous. Eliquis (apixaban) alone generated $20.3 billion in global sales in 2023. A 5% difference in negotiated price, supported by robust causal RWE showing real-world effectiveness in underrepresented populations, could preserve over $1 billion in revenue over a 3-year period.
4. The Methodology Selection Algorithm
Selecting the correct causal architecture is not a matter of statistical preference — it determines whether your submission survives regulatory review. The following decision framework maps your research question, data structure, and regulatory goal to the appropriate causal method.
Use when: Cross-sectional exposure, no time-varying confounding, adequate measured covariates.
Use when: Point treatment, no time-varying confounding affected by prior treatment, strong overlap assumption.
Use when: Defining a clear time zero, mimicking randomization through protocol design, active comparator available.
Use when: Time-varying treatment strategies, time-varying confounding affected by prior treatment (e.g., cardiovascular polytherapy).
Use when: Complex causal structure with multiple mediators, effect modifiers (e.g., genetic variants), and competing pathways. Need for explainability.
When multiple causal methods with different assumptions converge on the same estimate, confidence in the result is high. When they diverge, the divergence itself is informative and can reveal hidden effect modification or model misspecification. FDA reviewers increasingly value methodological triangulation as evidence of robustness. A submission that presents concordant results across multiple causal architectures is substantially more defensible than one relying on a single method.
5. The Financial Case: ROI of Causal RWE vs. Traditional Pathways
| Use Case | Traditional Cost | Causal RWE Cost | Savings | Time Saved |
|---|---|---|---|---|
| Label Expansion (CV) | $50–157M | $2–5M | $48–152M | 3–5 years |
| Post-Marketing Confirmatory | $50–100M | $3–8M | $42–92M | 2–4 years |
| IRA Value Demonstration | N/A (new req.) | $2–5M | Revenue preservation: $100M–$1B+ per drug | 6–12 months |
| Phase III Sample Size Optimization | $157M baseline | $94M (40% reduction) | $63M | 6 months |
5.1 Aggregate Opportunity Per Cardiovascular Asset
| Activity | Value Created |
|---|---|
| Avoided Phase III for label expansion (1 new indication) | $100–150M saved |
| Accelerated post-marketing confirmatory evidence | $50–90M saved |
| IRA negotiation leverage from real-world effectiveness data | $100M–$1B+ preserved |
| Phase III optimization through RWE-informed design | $30–63M saved |
6. Implementation Roadmap
Foundation
Define the target trial protocol for your primary use case. Specify eligibility criteria, treatment strategies, time zero, outcomes, and causal contrasts with the same rigor you would apply to a Phase III SAP. Identify the RWD source (All of Us, Optum, MarketScan, Flatiron, or institutional EHR) that best captures your population. Build the DAG specification with clinical domain experts to make all causal assumptions explicit and testable.
Execution
Implement the study design and run the primary causal analysis using the methodology best suited to your bias structure and data characteristics. Conduct pre-specified sensitivity analyses: quantitative bias analysis for unmeasured confounding, E-values, and probabilistic sensitivity analysis for exposure and outcome misclassification. Where genomic data is available, evaluate pharmacogenomic effect modification to identify subpopulations with differential treatment response.
Regulatory Engagement
Prepare the Type B meeting request and briefing document for FDA. Present methodological triangulation results demonstrating concordance or informative discordance across analytic approaches. Include transparent assumption documentation with the DAG and corresponding statistical tests. Engage the FDA's Office of Biostatistics early — their feedback on study design is more valuable than post-hoc methodological justification.
The FDA's December 2025 guidance on de-identified RWE takes effect February 16, 2026. Companies that initiate causal RWE programs now will be among the first to submit under the new framework — with a first-mover advantage in methodology acceptance and reviewer familiarity.
7. Conclusion
The pharmaceutical industry is entering a new era of evidence generation. The FDA's evolving posture on RWE, the IRA's demand for real-world value demonstration, and the prohibitive cost of cardiovascular Phase III trials have created a structural opening for causal inference methods. But the opening is narrow: submissions that fail to address the five core biases will be rejected, and the reputational cost of a failed RWE submission will set back an organization's regulatory strategy by years.
The causal RWE playbook presented here is not theoretical. It is grounded in the methods that the causal inference community has developed over four decades and applies them to the specific regulatory, economic, and clinical context of cardiovascular medicine in 2026.
The companies that invest in this capability now will not merely save costs. They will build an evidence moat: a strategic advantage in regulatory submissions, payer negotiations, and clinical differentiation that competitors relying solely on traditional trial designs cannot match.
References
- Moore TJ, Zhang H, Anderson G, Alexander GC. Estimated Costs of Pivotal Trials for Novel Therapeutic Agents Approved by the US Food and Drug Administration, 2015–2016. JAMA Intern Med. 2018;178(11):1451–1457.
- Eisenstein EL, Collins R, Cracknell BS, et al. Reducing the cost of phase III cardiovascular clinical trials. Am Heart J. 2005;149(3):482–488.
- Alipour-Haris G, Liu X, Acha V, Winterstein AG, Burcu M. Real-world evidence to support regulatory submissions: A landscape review and assessment of use cases. Clin Transl Sci. 2024;17:e13903.
- Alipour-Haris G, et al. Real-World Evidence in FDA Approvals for Labeling Expansion of Small Molecules and Biologics. Ther Innov Regul Sci. 2025.
- U.S. Food and Drug Administration. Use of Real-World Evidence to Support Regulatory Decision-Making for Medical Devices. Guidance Document, December 2025.
- U.S. Food and Drug Administration. FDA use of Real-World Evidence in Regulatory Decision Making. fda.gov, accessed February 2026.
- CTDSU. Real-World Evidence in Pharma: 2025 Industry Report. ctdsu.com, August 2025.
- Hernan MA, Robins JM. Causal Inference: What If. Chapman & Hall/CRC, 2020.
- Robins JM. A new approach to causal inference in mortality studies with a sustained exposure period. Math Model. 1986;7:1393–1512.
- Pearl J. Causality: Models, Reasoning, and Inference. 2nd ed. Cambridge University Press, 2009.
- Sendor R, Sturmer T. Core concepts in pharmacoepidemiology: Confounding by indication and the role of active comparators. Pharmacoepidemiol Drug Saf. 2022;31(3):261–269.
- Lund JL, Richardson DB, Sturmer T. The active comparator, new user study design in pharmacoepidemiology. Curr Epidemiol Rep. 2015;2(4):221–228.
- PwC. What is the value of real-world data in pharma? pwc.com, 2023.
- Martina R, Jenkins D, Bujkiewicz S, et al. The inclusion of real world evidence in clinical development planning. Trials. 2018;19:468.
- Centers for Medicare & Medicaid Services. Medicare Drug Price Negotiation Program Final Guidance, IPAY 2028. cms.gov, September 2025.
- Inflation Reduction Act of 2022. Pub. L. 117–169, 136 Stat. 1818 (2022), Sections 11001–11002.
- FDA. Regulatory Submissions with Real-World Evidence: Successes, Challenges, and Lessons Learned. FDA Speech, September 23, 2025.
- ISPOR. Solving the Real-World Evidence Challenge for US Payers: A Call to Action for Pharma. Value Outcomes Spotlight, 2024.
- ENCePP. Methods to Address Bias and Confounding. European Network of Centres for Pharmacoepidemiology and Pharmacovigilance Guide, Chapter 6.
- Gandjour A. Inclusion of phase III clinical trial costs in health economic evaluations. BMC Health Serv Res. 2024;24:1158.