ResearchPod Summary
The INCREASE trial previously established that inhaled treprostinil improves exercise capacity and reduces the risk of a first clinical worsening event in patients with pulmonary hypertension associated with interstitial lung disease (PH-ILD). However, clinical trials often ignore subsequent events once a patient experiences their first sign of disease progression. This study conducted a post hoc analysis of the original 326-patient dataset to determine if continued treatment with inhaled treprostinil provides ongoing benefits after an initial progression event occurs. The researchers expanded the definition of disease progression to include six clinical markers: a 15% or greater decline in 6-minute-walk distance (6MWD), a 10% or greater decline in forced vital capacity (FVC), acute exacerbation of lung disease, cardiopulmonary hospitalization, lung transplantation, or death.
The study found that patients treated with inhaled treprostinil experienced significantly fewer total disease progression events compared to those on placebo (147 vs. 215 events). Furthermore, patients in the treatment group were significantly less likely to experience multiple progression events; only 22% of the treprostinil group had more than one event, compared to 36% in the placebo group. The treatment was associated with a lower risk of both the first and second progression events, and these benefits were observed across most major disease subgroups, including idiopathic interstitial pneumonias and connective tissue disease-associated ILD.
In clinical practice, physicians and patients often face the difficult decision of whether to discontinue or switch therapies when a patient shows signs of worsening. This analysis provides evidence that inhaled treprostinil maintains its protective effect even after a patient has already experienced a decline in their condition. By demonstrating that the therapy continues to reduce the frequency of subsequent clinical events, the study supports the continued use of inhaled treprostinil as a long-term management strategy for PH-ILD patients, rather than abandoning the treatment at the first sign of progression.
[[RP_SECTION:increase-trial-post-hoc-analysis|INCREASE trial post hoc analysis]]
Alex: Welcome to another episode of ResearchPod. Today we're looking at a post hoc analysis of the INCREASE trial — a study of inhaled treprostinil in patients with pulmonary hypertension associated with interstitial lung disease, or PH-ILD. [[RP_SECTION:limitations-of-time-to-first-event|Limitations of time-to-first-event]]
Sam: The central puzzle here is how we handle clinical trial data in chronic, progressive diseases. The standard approach uses a time-to-first-event analysis, which effectively censors a patient the moment they experience a decline. After that first event, they disappear from the count.
Alex: So the paper's argument is that treating patients as binary — either they worsen or they don't — misses the reality of their clinical trajectory?
Sam: Exactly. In PH-ILD, clinical worsening isn't a one-time event; it's a series of setbacks. The original INCREASE trial met its primary endpoint, but once a patient hit that first worsening event, the analysis stopped counting. This study asks: what happens if you keep counting?
Alex: So the core question isn't just whether the drug works, but whether our measurement framework is actually capturing the burden of the disease over time.
Sam: Right. Think of it like a leaky bucket. The standard approach only registers when the bucket first starts to leak. This analysis measures the total volume lost across the entire 16-week period — even after that first hole appears. And that changes what you see.
Alex: How do they handle the statistical side of that? Recurrent events introduce a lot of noise. [[RP_SECTION:statistical-methodology-for-recurrent-ev|Statistical methodology for recurrent events]]
Sam: They moved away from simple time-to-event models and used negative binomial regression — designed specifically for count data where variance exceeds the mean. The key insight is that it lets every subsequent event contribute signal rather than getting absorbed into censoring. You're reclaiming data that the standard framework discards.
Alex: And what does that reclaimed data actually show? [[RP_SECTION:clinical-findings-and-disease-momentum|Clinical findings and disease momentum]]
AI-generated third-party summary by ResearchPod. Not official content or an endorsement by the paper authors or affiliated organizations.
Sam: The main finding is that patients on inhaled treprostinil were significantly less likely to experience further progression after an initial event. In the placebo group, once a patient worsened, they were much more likely to keep sliding. About 22% of the treatment group experienced multiple progression events, compared to 36% in the placebo arm. That's not just a delay — it's a meaningful reduction in the disease's momentum.
Alex: You mentioned they also expanded the definition of disease progression for this analysis. Does that complicate the interpretation?
Sam: It's worth flagging. The original trial used a four-component composite: six-minute walk distance, hospitalization, transplant, or death. This post hoc analysis added FVC decline and acute exacerbations — a broader, arguably more clinically complete picture. But any time you expand your endpoint definition post hoc, you have to ask whether the result is an artifact of the new definitions.
Alex: So did they test that?
Sam: They did. The sensitivity analysis restricted to the original four endpoints still showed a clear benefit in reducing total event burden. That's the load-bearing robustness check here — the finding doesn't depend on the expanded definition holding. [[RP_SECTION:sensitivity-and-study-limitations|Sensitivity and study limitations]]
Alex: What are the harder limitations? Post hoc analyses always carry some baggage.
Sam: The biggest is that this is post hoc by design, so you can't fully escape the inference constraints that come with that. Beyond that, the acute exacerbations were investigator-reported rather than centrally adjudicated — which introduces heterogeneity in how events were classified across sites. And the 16-week follow-up is genuinely short for a chronic, progressive disease. You're getting a snapshot of trajectory, not a long-term picture of disease modification or survival.
Alex: So this shouldn't be read as evidence for durability of effect beyond the trial window. [[RP_SECTION:clinical-and-methodological-implications|Clinical and methodological implications]]
Sam: Correct. What it does support is a reframing of how clinicians interpret a setback during treatment. If a patient shows a meaningful decline in walk distance, the instinct might be to assume the drug has failed and switch strategies. This analysis suggests that instinct may be premature — the drug appears to provide sustained value even after an initial event, and the overall trajectory remains more stable than in the placebo arm.
Alex: That's a meaningful shift in clinical reasoning. From "did the drug prevent the first drop?" to "is the drug keeping the overall trajectory more controlled?"
Sam: That's exactly the reframe. And there's a methodological implication worth noting: if future trials adopt total event burden as a primary endpoint rather than time-to-first-event, they may actually improve statistical power without inflating sample size — because they're extracting more information from each patient's longitudinal course.
Alex: Which would be a genuine design improvement for this class of disease.
Sam: It would. PH-ILD is a condition where the clinical reality is cumulative deterioration, not a single threshold crossing. Aligning the statistical framework to that reality is overdue. This analysis is a reasonable proof of concept for what that looks like in practice.
Alex: Thanks for walking through the mechanism, Sam. It's a clear example of how the choice of statistical lens can reveal a substantially different story in the same dataset.
Sam: And a reminder that the analysis plan is as much a design decision as the intervention itself. Thanks for listening to ResearchPod.