Review Reports - Efficacy and Safety of Inhaled Treprostinil in Idiopathic Pulmonary Fibrosis: A Systematic Review and Meta-Analysis

Round 1

Reviewer 1 Report

Comments and Suggestions for Authors

This systematic review and meta-analysis addresses an important and highly current question regarding inhaled treprostinil in IPF. The decision to separate the two direct-IPF TETON trials from INCREASE and other PH-ILD datasets is methodologically appropriate and avoids double-counting overlapping populations. arm-4541865-peer-review-v1.pdf The findings are clinically relevant. However, several methodological issues should be addressed before the conclusions can be considered sufficiently robust.

Major comments

Systematic-review methodology is insufficiently reproducible.
The manuscript states that PubMed, Google Scholar, and CENTRAL were searched, but provides only general search concepts rather than complete database-specific strategies. The exact final search date is also unclear (“January 2000 to 2026”). arm-4541865-peer-review-v1.pdf Please provide the complete search strings, exact search dates, restrictions applied, and preferably the strategies as supplementary material.
Please clarify whether the review protocol was prospectively registered.
I could not identify a PROSPERO registration or another prespecified protocol. If no protocol was registered, this should be explicitly stated and acknowledged as a limitation, particularly because several decisions regarding primary versus secondary evidence and pooling were made during the review.
Study selection and data extraction procedures need clarification.
The manuscript does not clearly state whether title/abstract screening, full-text assessment, data extraction, and risk-of-bias assessment were performed independently by two reviewers, how disagreements were resolved, or whether extraction was verified. These are important safeguards against reviewer-selection and extraction errors. Although author contributions indicate involvement in screening and extraction, they do not establish independent duplicate assessment.
The statistical approach requires reconsideration because only two trials are pooled.
The principal meta-analysis contains only TETON-1 and TETON-2. With k=2, estimates of between-study heterogeneity are highly unstable, and I² values of 0% or 8.5% should not be interpreted as demonstrating meaningful homogeneity. The authors recognize this limitation later in the Discussion. However, DerSimonian-Laird random-effects estimation is particularly fragile with only two studies. Please justify this estimator and consider an appropriate sensitivity analysis (for example, fixed-effect and/or Hartung-Knapp-type approaches), emphasizing the limited information available for estimating between-study variance
The statement regarding clinical meaningfulness of the 111-mL effect needs more caution.
The manuscript states that the pooled 111.28-mL difference “exceeds commonly cited thresholds for meaningful FVC change.” arm-4541865-peer-review-v1.pdf A between-treatment difference in FVC decline is not directly equivalent to an individual-patient minimal clinically important difference. I suggest revising this wording and distinguishing statistical treatment effect from patient-level MCID.
Multiplicity and hierarchical testing should be handled consistently throughout the manuscript.
The authors appropriately acknowledge that the TETON testing hierarchy stopped after the nonsignificant acute-exacerbation endpoint and that subsequent endpoints should be interpreted nominally. arm-4541865-peer-review-v1.pdf However, the Abstract describes pooled acute exacerbation as simply “lower” (HR 0.52, p=0.022), which can be read as confirmatory. arm-4541865-peer-review-v1.pdf These results should consistently be described as exploratory/nominal and not as establishing efficacy.
A formal certainty-of-evidence assessment would strengthen the review.
The manuscript presents risk-of-bias assessments but does not appear to provide a GRADE or equivalent certainty-of-evidence assessment. Given that this is presented as a systematic review/meta-analysis and only two independent IPF RCTs constitute the quantitative evidence, a structured certainty assessment would substantially improve interpretation.
Risk-of-bias assessment needs additional methodological detail.
The manuscript reports “some concerns” for both TETON trials, largely due to attrition and incomplete sequence-generation information. arm-4541865-peer-review-v1.pdf Please provide the domain-level RoB 2 assessments, explain who performed them, and describe the “adapted non-randomized study framework” used for open-label studies. If an adapted NOS was used, the adaptations should be specified rather than reporting only a global judgment.
The limitations of the literature search deserve greater emphasis.
Embase was not searched because access was unavailable, while Google Scholar contributed a very large proportion of retrieved records. arm-4541865-peer-review-v1.pdf Google Scholar searches are difficult to reproduce exhaustively. The authors should describe exactly how Google Scholar was searched, how many results were examined, and acknowledge the possibility of missed studies.

Minor comments

Consider changing statements such as “the randomized direct IPF evidence answers yes” to more measured language. arm-4541865-peer-review-v1.pdf The two phase III trials provide strong supportive evidence, but a systematic review should avoid binary language.
The manuscript correctly separates INCREASE from direct-IPF evidence; this is one of its strongest features and should be retained.
Safety interpretation is appropriately balanced: discontinuations, cough, and headache were increased despite no statistically significant increase in overall or serious AEs.
Consider consolidating the numerous individual forest plots if journal space is limited; several could potentially be moved to supplementary material.

Author Response

Response to Reviewers

Manuscript: Efficacy and Safety of Inhaled Treprostinil in Idiopathic Pulmonary Fibrosis: A Systematic Review and Meta-Analysis

Manuscript ID: arm-4541865

We thank the reviewers for their careful and constructive assessment of the manuscript. We have revised the manuscript to improve methodological transparency, statistical interpretation and separation of direct IPF evidence from supportive PH-ILD evidence. Revisions incorporated in the manuscript are shown in blue font. The supplementary document has also been expanded to include the exact database search strategies, protocol chronology, reviewer workflow, detailed extraction and pooling rules, small-k sensitivity analyses and a GRADE certainty-of-evidence assessment.

Reviewer 1 - Major Revision

Overall Assessment

This systematic review and meta-analysis addresses an important and clinically relevant question regarding the efficacy and safety of inhaled treprostinil in idiopathic pulmonary fibrosis (IPF). The topic is timely and potentially valuable; however, several methodological and interpretative issues require clarification before the manuscript can be considered for publication.

Response: We appreciate the reviewer's positive assessment of the clinical relevance of the question and agree that the methodological and interpretative points required full explanation. The revision therefore focuses on reproducibility of the search, protocol chronology, independent screening and data verification, the limitations of meta-analysis with only two direct-IPF trials, additional sensitivity analyses, GRADE certainty assessment, clearer risk-of-bias interpretation and more cautious statements regarding clinical significance and mechanism.

Major Comment 1. Systematic review methodology and reproducibility

The search strategy requires greater detail to ensure reproducibility. Please provide the complete search strategies for each database, including search terms, Boolean operators, filters, and the date on which the final search was conducted.

Response: We agree. The Methods now state that the final search was run on 3 July 2026 and identify the principal databases as PubMed, Google Scholar and the Cochrane Library/CENTRAL. The complete database-specific search strategies have been reproduced in the supplementary material rather than summarized only by search concepts. Supplementary Table S1 and Supplementary Search Appendix S1A provide the exact PubMed strings, MeSH and title/abstract fields, Boolean operators, publication-date syntax, the four Google Scholar queries with the custom year range 2000-2026, and the three Cochrane searches.

For transparency, we also distinguish the principal searches used for PRISMA identification from focused verification searches. The PRISMA identification total remains 1552 records from the three principal searches: PubMed 449, Google Scholar 913 and Cochrane Library/CENTRAL 190. Focused IPF, PH-ILD, TETON and INCREASE searches were used to verify retrieval and were not added again to the PRISMA total. We also clarify that although the PubMed syntax contained an upper publication-date limit of 31 December 2026, the searches were actually executed on 3 July 2026 and could not retrieve records indexed after that date. Embase was not searched because database access was unavailable, and this remains stated as a limitation.

Major Comment 2. Protocol registration

Please clarify whether a review protocol was prospectively developed and registered. If the review was not registered before study screening/data extraction, this should be transparently acknowledged as a methodological limitation rather than retrospectively assigning a registration date.

Response: The protocol was developed and finalized before study screening commenced. It specified the eligibility criteria, outcomes and planned analytical approach. It was not, however, registered prospectively in PROSPERO. PROSPERO registration was completed retrospectively. We have revised the methods to state this chronology explicitly and have added retrospective registration as a methodological limitation. We have not described the registration as prospective or assigned a retrospective registration date as if it preceded screening.

Major Comment 3. Study selection and data extraction

The manuscript should provide greater detail regarding the number of reviewers involved in screening, full-text selection, and data extraction, including whether these processes were performed independently and how disagreements were resolved.

Response: We have expanded the Methods and author-contribution statements. Title and abstract screening and full-text eligibility assessment were performed independently by the first and fourth authors. Disagreements were recorded and then adjudicated in meetings involving all authors until consensus was reached.

Data extraction was divided equally between the first and second authors. Each independently cross-checked the data extracted by the other against the corresponding source publication before the dataset was finalized. Any discrepancies identified during verification were resolved by discussion. This workflow is now also documented in Supplementary Methods S1B.

Major Comment 4. Meta-analysis based on limited studies

The quantitative synthesis appears to be driven primarily by two TETON trials. The authors should justify the appropriateness of pooling only two studies and discuss the limitations of estimating between-study heterogeneity with such a small evidence base.

Response: We agree that the small number of independent trials is the principal statistical limitation of the review. We retained pooling because TETON-1 and TETON-2 are independent replicate phase 3 randomized trials addressing the same direct-IPF question, using the same intervention, placebo comparator and 52-week outcome framework. Pooling therefore provides a useful summary estimate across two clinically comparable trials, but it does not make estimation of between-study heterogeneity reliable.

The revision now states explicitly that I squared and tau squared are imprecisely estimated when k = 2 and should be interpreted cautiously. We also added sensitivity analyses comparing the primary DerSimonian-Laird random-effects model with fixed-effect inverse-variance and REML random-effects models, together with modified Hartung-Knapp inference as a conservative small-k analysis. For absolute FVC change, the point estimate was essentially unchanged: MD 111.28 mL (95% CI 77.61 to 144.95) with DerSimonian-Laird, MD 111.14 mL (95% CI 78.96 to 143.31) with fixed effect and MD 111.28 mL (95% CI 77.61 to 144.95) with REML. The modified Hartung-Knapp interval was much wider, MD 111.28 mL (95% CI -106.99 to 329.56), which illustrates the uncertainty of random-effects inference when only two trials are available. Full results for all 14 pooled outcomes are provided in Supplementary Table S13.

Major Comment 5. Interpretation of the FVC effect

The reported approximately 111-mL difference in FVC is clinically interesting; however, the manuscript should avoid implying definite clinical significance solely on the basis of statistical significance. Please discuss the magnitude of effect in relation to established clinically meaningful thresholds and the uncertainty surrounding these estimates.

Response: We agree and have moderated the interpretation. The revised Results state that the pooled between-group effect was approximately 111 mL, but its 95% CI of 77.61 to 144.95 mL spans values below and above the commonly cited approximately 100-mL patient-level threshold. We now explicitly distinguish a trial-level between-group treatment effect from a patient-level minimal clinically important difference.

The Discussion has also been revised to state that the pooled result supports preservation of FVC at the trial level but should not be interpreted as evidence that every treated patient experiences a clinically meaningful improvement. The GRADE assessment downgrades certainty for FVC by one level for imprecision, in part because of this range of plausible effects and the much wider small-k adjusted interval.

Major Comment 6. Risk of bias

Please provide a detailed study-level risk-of-bias assessment and explain how the identified methodological limitations influenced the interpretation of the pooled results.

Response: We have expanded the risk-of-bias reporting at study level in Table 3 and Supplementary Table S9. RoB 2 was used for randomized trials. TETON-1 and TETON-2 were judged as having some concerns, chiefly because of attrition before week 52 and incomplete reporting of sequence-generation details in the main articles. We also describe features that reduce the likelihood of material distortion of the primary outcome, including double blinding, central spirometry review, prespecified analysis procedures and supportive sensitivity analyses. INCREASE was also judged as some concerns, with specific attention to investigator-reported clinical events.

Open-label and single-arm evidence was assessed separately using an adapted non-randomized/NOS framework because the RoB 2 domains are not appropriate for uncontrolled evidence. Post-hoc INCREASE analyses were treated as exploratory and assessed for reuse of the parent cohort, retrospective subgrouping, multiplicity and post-randomization selection. These limitations now directly inform the interpretation: direct randomized evidence is kept separate from supportive observational or post-hoc evidence, and the GRADE note explains why the RoB 2 judgment of some concerns was retained even where no additional GRADE downgrade for risk of bias was applied.

Major Comment 7. Certainty of evidence

Consider incorporating a GRADE assessment for the major outcomes. This would help readers understand the certainty and clinical reliability of the available evidence.

Response: We have added a formal GRADE certainty-of-evidence assessment for the major direct-IPF outcomes. The Methods now describe assessment of risk of bias, inconsistency, indirectness, imprecision and publication bias. A summary is provided in Table 7 of the revised manuscript, with full domain-level judgments and reasons for downgrading in Supplementary Table S14.

The resulting certainty ratings are moderate for absolute FVC change and clinical worsening, low for acute exacerbation of IPF and all-cause mortality, low for serious adverse events and moderate for adverse events leading to treatment discontinuation. The principal reasons for downgrading are imprecision with only two direct trials, sparse events for acute exacerbation and mortality, and inconsistency for serious adverse events. INCREASE and other PH-ILD evidence were not used to increase certainty for the direct-IPF question because of population and follow-up indirectness. Formal publication-bias testing was considered uninformative with only two direct trials.

Major Comment 8. Direct IPF versus PH-ILD evidence

The manuscript should clearly distinguish evidence derived directly from IPF populations from evidence obtained from pulmonary hypertension associated with interstitial lung disease. These populations should not be interpreted as equivalent, and the possibility of indirectness should be explicitly discussed.

Response: We agree and have strengthened this distinction throughout the manuscript. The primary quantitative synthesis is restricted to TETON-1 and TETON-2, the two independent randomized trials conducted directly in IPF. INCREASE enrolled PH-ILD and had a 16-week follow-up; its IPF subgroup FVC findings are therefore retained as secondary supportive/sensitivity evidence rather than pooled with the 52-week direct-IPF trials. The open-label extension and Japanese PH-ILD studies remain narrative only.

The Discussion now treats these PH-ILD data as biologically and clinically supportive but indirect for the direct-IPF treatment question. The GRADE assessment is restricted to TETON-1 and TETON-2 and explicitly excludes PH-ILD evidence from the certainty ratings. Supplementary Methods S2 also maps each publication to its parent dataset and pooling role so that population indirectness and overlapping evidence are transparent.

Reviewer 1 - Minor Comments

Minor Comment 1. Definitions of clinical worsening and acute exacerbation of IPF

Clarify definitions of clinical worsening and acute exacerbation of IPF.

Response: We have clarified the endpoint definitions used in the TETON trials. Clinical worsening is the first occurrence of death from any cause, respiratory hospitalization or a relative decline of at least 10% in percentage predicted FVC. Acute exacerbation of IPF is treated as the trial-defined time-to-event endpoint and was adjudicated according to the trial protocol. The underlying diagnostic framework requires an acute clinically significant respiratory deterioration in a patient with IPF, typically developing within less than one month, with new bilateral ground-glass opacity and/or consolidation superimposed on a UIP background and not fully explained by cardiac failure or fluid overload. The meta-analysis used the published trial hazard ratios and did not redefine this endpoint.

Minor Comment 2. Pulmonary-function terminology

Ensure consistent terminology for FVC, DLCO, and other pulmonary function parameters.

Response: We have standardized terminology across the text, tables and supplementary material. Absolute FVC change is reported in milliliters, percentage predicted FVC is reported separately, and percentage predicted DLCO is distinguished from absolute DLCO. The same terminology is used in the pooled-results table, sensitivity table and figure labels.

Minor Comment 3. Effect estimates and confidence intervals

Please standardize reporting of effect estimates and confidence intervals throughout the manuscript.

Response: Effect estimates are now reported in a consistent format using the effect measure followed by the estimate and 95% confidence interval, for example HR 0.69 (95% CI 0.57 to 0.84) and RR 1.38 (95% CI 1.06 to 1.79). Mean differences are reported with the original units. Supplementary Tables S11 and S12 define the null value and interpretation for MD, HR and RR and state explicitly that SMD and OR were not used.

Minor Comment 4. Safety terminology

The safety outcomes should clearly distinguish treatment-emergent adverse events from serious adverse events.

Response: We agree. Safety categories are now kept separate throughout the Results and supplementary extraction tables. Any adverse event, serious adverse events, adverse events leading to treatment discontinuation, deaths and individual common adverse events are reported as distinct outcomes. Where the source trial uses treatment-emergent adverse-event terminology, that terminology is retained rather than treating all safety categories as interchangeable. The interpretation also distinguishes treatment tolerability from serious-event risk.

Minor Comment 5. Cautious conclusion

The conclusion should be appropriately cautious and reflect the limited number of direct randomized studies.

Response: The conclusion has been moderated. It now states that the findings are based on two independent phase 3 direct-IPF trials, that a mortality benefit has not been established, and that the findings do not establish a disease-modifying antifibrotic effect. The PH-ILD subgroup, open-label, and Japanese data are described as supportive and are kept separate from the primary direct-IPF meta-analysis.

Reviewer 2 Report

Comments and Suggestions for Authors

Overall, the manuscript is well structured, clearly presented, and addresses clinically relevant question; however, some aspects of the methodology, interpretation of the findings, and language should be clarified or refined to strengthen the manuscript.

Major

The search strategy should be described in greater detail to ensure reproducibility. Please provide the complete search strategies for each database, including the exact search terms, Boolean operators, and search dates. The authors should also clarify the rationale for not searching Embase, as this may affect the completeness of the evidence retrieval.
The eligibility criteria and approach to evidence synthesis could be clarified. The Methods include direct-IPF randomized trials, PH-ILD studies, post-hoc analyses, open-label extensions, and single-arm studies, but it is not fully clear how these different evidence types were handled in the review. Please clarify a priori which study designs were eligible for the primary analysis, secondary analyses, and narrative synthesis.
Please provide additional justification for the statistical approach. In particular, the authors should clarify the choice of the DerSimonian–Laird random-effects model given that the primary meta-analysis includes only two trials, and describe how heterogeneity was interpreted in this context.
The handling of multiple reports from the same patient-level dataset should be clarified. Although the authors state that the 11 reports represent four independent datasets and that secondary INCREASE and Japanese reports were not treated as independent studies, it would be helpful to explain more explicitly how these reports were handled in the quantitative and narrative syntheses to ensure that no patient population was counted more than once.
The risk-of-bias assessment for TETON-1 and TETON-2 should be better justified. In particular, please clarify how the authors determined that attrition and incomplete reporting of sequence generation resulted in an overall judgment of “some concerns” under RoB 2, and indicate the specific RoB 2 domains supporting this judgment.
The meta-analysis includes only two direct-IPF phase 3 trials, and several secondary endpoints did not meet the trials' hierarchical testing criteria. Statements such as “the randomized direct IPF evidence answers yes” and “treprostinil reduces lung-function decline and clinical worsening” may therefore be somewhat definitive. Please consider more cautious wording that reflects the limited number of trials and the nominal nature of some secondary analyses.

Minor:

On the line 49 of abstract add the full name of PH-ILD
The rationale and aim of the systematic review could be clarified. Please state the primary objective and main efficacy and safety outcomes more explicitly, and clarify the distinction between the direct-IPF evidence from TETON-1/TETON-2 and the supportive evidence from INCREASE and other PH-ILD studies.
The authors report “some concerns” for both TETON-1 and TETON-2 because of attrition and limited reporting of sequence generation. Please clarify whether these judgments are based on the published reports alone and provide the specific RoB 2 domains supporting the overall judgment.
The interpretation of secondary outcomes should be consistent with the statistical hierarchy. Since the authors appropriately note that these outcomes are nominal because the hierarchical testing procedure stopped after the non-significant acute exacerbation endpoint, this limitation should also be clearly reflected in the Abstract and Discussion to avoid overinterpretation of these statistically significant pooled estimates.
Please clarify whether the 6-minute walk distance and other outcomes reported from INCREASE are being presented only as supportive evidence, and avoid combining these findings with the direct-IPF TETON evidence when drawing overall conclusions about efficacy.
The authors describe the 111 mL FVC difference as clinically relevant, but the discussion should more clearly distinguish the statistical effect size from established thresholds for clinically meaningful change at the individual-patient level. This is particularly relevant because the pooled estimate represents a between-group treatment effect rather than an individual patient's change in FVC.
“The TETON findings point to possible prostacyclin pathway effects beyond pulmonary vasodilation.” is interesting, but this remains a mechanistic hypothesis based partly on preclinical evidence. I would suggest making this more cautious to avoid implying that the clinical trials demonstrate an antifibrotic mechanism.

Author Response

Reviewer 2 - Major Revision

Overall Assessment

The manuscript evaluates the efficacy and safety of inhaled treprostinil in IPF and addresses an important therapeutic question. The manuscript is potentially publishable; however, clarification of the statistical methodology, handling of overlapping evidence, outcome interpretation, and limitations is required.

Response: We thank the reviewer for this assessment. We have revised the statistical methods, added model-based sensitivity analyses, made the handling of overlapping datasets explicit, strengthened the discussion of multiplicity and small-k uncertainty, added GRADE and expanded the limitations. The revision also separates direct-IPF evidence from indirect PH-ILD evidence more clearly.

Major Comment 1. Statistical methodology

The authors have used a random-effects DerSimonian-Laird model. Given that only a very small number of studies contribute to several pooled analyses, please justify this approach and discuss the limitations of estimating heterogeneity and pooled effects with only two studies.

Response: We retained the DerSimonian-Laird random-effects model as the primary analysis because it was the prespecified/submitted approach and because the two TETON trials are independent replicate trials of the same direct-IPF question. We now explicitly state that with only two studies, tau squared and I squared are poorly estimated and should not be treated as precise measures of between-study heterogeneity.

To test model dependence, we repeated all 14 pooled outcomes using fixed-effect inverse variance and REML random effects and added modified Hartung-Knapp inference as a conservative small-k sensitivity analysis. Point estimates were nearly identical under DerSimonian-Laird, fixed-effect and REML models, while modified Hartung-Knapp intervals were substantially wider. This distinction is now emphasized in both Results and Discussion: the estimated effect is stable across conventional models, but the precision of random-effects inference is limited by k = 2.

Major Comment 2. Handling of potentially overlapping datasets

Please clearly demonstrate that individual trials or publications reporting overlapping patient populations were not counted more than once. A study-level table identifying the source of each outcome would improve transparency.

Response: We have made the parent-dataset structure explicit. The 11 included reports represent four independent patient-level datasets: TETON-1, TETON-2, INCREASE and the Japanese PH-ILD cohort. TETON-1 and TETON-2 alone contribute to the primary quantitative synthesis. A combined TETON analysis is not pooled with the individual trials. The INCREASE parent trial and its post-hoc publications are recognized as a single shared cohort, and the Japanese initial and long-term publications are recognized as the same Japanese cohort.

Supplementary Methods S2 now provides a dataset-level map showing the reason for each analytical decision and the double-counting bias avoided. Supplementary Table S3 also identifies the role of every included report as main meta-analysis, secondary/sensitivity, narrative only or design background.

Major Comment 3. Outcome hierarchy and multiplicity

Multiple efficacy and safety outcomes are analysed. Please clarify whether primary and secondary outcomes were prespecified and discuss the potential issue of multiplicity, particularly where several nominally statistically significant secondary outcomes are reported.

Response: The review protocol was finalized before screening and specified the primary and secondary outcomes. The primary review outcome was absolute FVC change at week 52. The revision now distinguishes this review-level prespecification from the hierarchical testing structure within the TETON trials.

We specifically note that the TETON statistical hierarchy stopped after the acute-exacerbation endpoint failed to meet the required significance criterion in the individual trial hierarchy. Consequently, pooled estimates for subsequent secondary outcomes such as K-BILD and DLCO are described as nominal rather than confirmatory, even when their pooled 95% confidence intervals exclude the null. Safety analyses are interpreted descriptively and in the context of multiple endpoints rather than as a series of independent confirmatory tests.

Major Comment 4. Continuous and time-to-event outcomes

Please provide sufficient methodological detail regarding the use of mean differences, standardized mean differences, risk ratios, and hazard ratios. Explain how missing variance data were handled and whether any assumptions or transformations were required.

Response: We have expanded the statistical methods in both the manuscript and Supplementary Methods S3. Continuous efficacy outcomes were on the same measurement scale across TETON-1 and TETON-2 and were therefore pooled as mean differences using generic inverse variance. The trial-reported Hodges-Lehmann location-shift estimates and their 95% confidence intervals were used; standard errors were derived as (upper confidence limit - lower confidence limit)/(2 x 1.96). Standardized mean differences were not required.

Hazard ratios for time-to-event outcomes were pooled on the log scale, with standard errors derived as [ln(upper confidence limit) - ln(lower confidence limit)]/(2 x 1.96). Safety outcomes were pooled as risk ratios from raw event counts, using the log-risk-ratio variance based on events and randomized/safety denominators. Odds ratios were not used. A continuity correction of 0.5 was planned only if a zero cell occurred. These derivations, assumptions and the interpretation of the null value for each effect measure are now shown in Supplementary Tables S11 and S12.

Major Comment 5. Sensitivity analyses

Additional sensitivity analyses would strengthen the manuscript, particularly analyses examining the influence of individual trials, different statistical assumptions, and exclusion of indirect/supportive evidence.

Response: We have added sensitivity analyses for all 14 pooled outcomes. The primary DerSimonian-Laird results were compared with fixed-effect inverse-variance and REML random-effects models and with modified Hartung-Knapp inference. The point estimates were essentially unchanged across the conventional models. The much wider Hartung-Knapp intervals are reported as evidence of small-k uncertainty rather than as evidence of disagreement between the TETON trials.

A conventional leave-one-out meta-analysis was not performed because removing either of two trials leaves only one study and therefore no pooled estimate. Instead, we assessed trial influence by comparing each trial-specific estimate with the pooled estimate. For absolute FVC change, TETON-1 showed 130.1 mL (95% CI 82.2 to 178.1), TETON-2 showed 95.6 mL (95% CI 52.2 to 139.0) and the pooled estimate was 111.28 mL (95% CI 77.61 to 144.95). For clinical worsening, the corresponding HRs were 0.67, 0.71 and 0.69.

Indirect/supportive evidence was already excluded from the primary pooling by design. INCREASE, its post-hoc analyses, the open-label extension and the Japanese PH-ILD cohort therefore do not contribute to the direct-IPF pooled estimates. This exclusion is now made explicit as part of the sensitivity and evidence-architecture discussion.

Major Comment 6. Risk-of-bias assessment

Please explain how risk of bias was assessed for randomized versus non-randomized/open-label evidence. If different assessment tools were used, the rationale and domains should be explicitly described.

Response: We have clarified the tool selection and rationale. Randomized trials were assessed with RoB 2 because its domains address the randomization process, deviations from intended interventions, missing outcome data, outcome measurement and selection of the reported result. Open-label extension and single-arm studies were assessed with an adapted non-randomized/NOS framework because they lack random allocation and concurrent randomized comparators; the assessment therefore focused on selection, comparability where applicable, outcome ascertainment and follow-up/attrition.

Post-hoc INCREASE analyses required a separate methodological critique because their main threats arise from shared-cohort reuse, retrospective subgroup definitions, multiplicity and post-randomization grouping. Study-level judgments and the principal reasons are presented in Table 3 and Supplementary Table S9, and these considerations are carried forward into the GRADE certainty assessment.

Major Comment 7. Clinical interpretation of secondary outcomes

Changes in quality-of-life measures, 6-minute walk distance, DLCO, NT-proBNP, and other secondary endpoints should be interpreted cautiously. Statistical significance should not automatically be equated with clinically meaningful benefit.

Response: We agree. The revision separates statistical evidence from clinical importance. K-BILD is explicitly interpreted against published patient-level meaningful-change thresholds, and the pooled effect is not presented as proof of a clinically important quality-of-life improvement. Percentage predicted DLCO is described as a nominal physiologic finding because of the TETON outcome hierarchy.

Six-minute walk distance and NT-proBNP derive from INCREASE PH-ILD rather than the direct-IPF TETON trials and are therefore described only as supportive PH-ILD evidence. We no longer use statistical significance in these secondary outcomes as a substitute for demonstrated patient-centered benefit in direct IPF.

Major Comment 8. Safety analysis

The safety section should provide a balanced assessment of adverse events, treatment discontinuation, serious adverse events, and treatment-related mortality. Particular attention should be given to adverse effects known to influence adherence to inhaled treprostinil.

Response: We have retained a balanced safety presentation. Any adverse event and serious adverse events did not show statistically significant pooled differences, whereas adverse events leading to treatment discontinuation were more frequent with inhaled treprostinil. Cough and headache were the clearest tolerability signals and are discussed in relation to treatment persistence. Dyspnoea and diarrhoea did not differ significantly in pooled analysis.

Deaths are reported separately from adverse events and serious adverse events. The manuscript does not attribute a treatment-related mortality benefit or harm where the source data do not support such a conclusion. The Discussion now emphasizes that an efficacy signal must be considered together with tolerability and the ability of patients to continue inhaled treatment in routine practice.

Major Comment 9. Mechanistic interpretation

Statements regarding antifibrotic or disease-modifying effects should be presented as hypotheses unless directly supported by clinical evidence. The observed preservation of FVC does not by itself establish that inhaled treprostinil modifies the underlying fibrotic disease process.

Response: We agree. Mechanistic statements have been moderated. Preclinical findings regarding fibroblast proliferation, extracellular matrix pathways and YAP/TAZ signaling are retained as biological rationale, but the clinical TETON results are not presented as proof of an antifibrotic or disease-modifying mechanism. The Discussion now describes possible prostacyclin-pathway effects beyond pulmonary vasodilation as a hypothesis requiring direct mechanistic clinical evidence. The revised conclusion explicitly states that the present findings do not establish a disease-modifying antifibrotic effect.

Major Comment 10. Conclusions

The conclusion should be moderated to reflect the relatively limited direct evidence in IPF. The manuscript should distinguish between evidence supporting preservation of lung function and evidence demonstrating improvement in patient-centered outcomes or long-term survival.

Response: The conclusion has been revised to reflect the limited evidence base. It now identifies the evidence as two independent phase 3 direct-IPF randomized trials, supports preservation of FVC and reduction in clinical-worsening events, and states that a mortality benefit has not been established. It also avoids equating physiologic preservation with proven long-term survival or a disease-modifying effect. PH-ILD subgroup, open-label and Japanese data remain supportive but separate from the direct-IPF meta-analysis.

Reviewer 2 - Minor Comments

Minor Comment 1. Study and participant counts in forest plots

Please ensure that all forest plots identify the number of studies and participants contributing to each analysis.

Response: We agree. The forest-plot set for resubmission will explicitly identify the number of contributing studies and participants. The primary direct-IPF pooled analyses contain two independent studies, TETON-1 and TETON-2, with 1191 randomized participants in total; outcome-specific safety denominators are retained where they differ because of the safety populations. The legends and plot labels will be synchronized with the corresponding Results and supplementary tables.

Minor Comment 2. Abbreviations

Define all abbreviations at first appearance.

Response: We have performed an abbreviation audit and expanded abbreviations at first use in the main text and supplementary material, including IPF, PH-ILD, FVC, DLCO, K-BILD, 6MWD, NT-proBNP, RR, HR, MD, REML, GRADE and other statistical abbreviations. Repeated use thereafter follows the abbreviated form consistently.

Minor Comment 3. Follow-up duration

Clarify the rationale for the selected follow-up duration for each outcome.

Response: The follow-up periods reflect the design of the source evidence rather than an arbitrary review choice. The primary direct-IPF outcomes use 52 weeks because this was the prespecified primary assessment time in TETON-1 and TETON-2. INCREASE outcomes are reported at 16 weeks because this was the randomized PH-ILD trial period, and these data are kept secondary because both the population and follow-up differ from TETON. Open-label and Japanese long-term observations are reported at the follow-up available in those studies and are not pooled with the direct-IPF randomized evidence.

Minor Comment 4. Consistency between manuscript, tables and supplementary material

Ensure consistency between the text, tables, and supplementary material.

Response: We have cross-checked the numerical results, effect measures, confidence intervals, p values, study roles and dataset identities across the manuscript and supplementary material. The supplementary document contains the extraction tables that underpin the pooled results, the effect-measure interpretation table and the sensitivity-analysis table. We also corrected the non-zero tau squared values in the pooled-results summary where needed: 50.7213 for absolute FVC change, 0.0767 for percentage predicted FVC, 0.0470 for serious adverse events and 0.0066 for cough. These corrections do not alter the pooled effect estimates or confidence intervals.

Minor Comment 5. Overstatement in abstract and conclusion

Please review the manuscript for overstatement of efficacy in the abstract and conclusion.

Response: We have reviewed the efficacy wording throughout the manuscript. The revised conclusion is explicitly limited to the two direct-IPF randomized trials and states that the findings do not establish a mortality benefit or a disease-modifying antifibrotic effect. The abstract and Discussion are aligned with this interpretation: the data support preservation of lung function and fewer clinical-worsening events over 52 weeks, while patient-centered and long-term survival benefits remain uncertain. Secondary and PH-ILD findings are described as nominal, supportive or hypothesis-generating as appropriate.

Closing statement

We thank both reviewers for comments that materially improved the transparency and interpretation of the review. The revised manuscript and supplementary data now provide a more reproducible search record, a transparent protocol and reviewer workflow, explicit handling of overlapping datasets, small-k sensitivity analyses, GRADE certainty ratings and more cautious clinical and mechanistic interpretation. We hope that these revisions adequately address the concerns raised.

Round 2

Reviewer 1 Report

Comments and Suggestions for Authors

The revision is substantially improved and responsive to the prior major-revision points, but a small number of issues remain that should be resolved before acceptance — most importantly a factual/chronology inconsistency around the search date, an unfilled registration number, and a conceptual point about what this meta-analysis adds beyond the trials' own prespecified pooled analysis. The extracted trial data are accurate, and the statistical handling is now appropriately cautious. On balance this warrants minor revision, not another major round.

Verification of the core data (accurate)

The numbers the authors extracted match the primary NEJM reports published 9 July 2026:

TETON-2 between-group FVC difference 95.6 mL (95% CI 52.2–139.0), clinical worsening HR 0.71, cough 48.3% vs 24.1% — all correct.
TETON-1 clinical worsening 95/299 (31.8%) vs 133/299 (44.5%), HR 0.67 (0.52–0.88); acute exacerbation HR 0.50 (0.23–1.08); death 22 vs 32, HR 0.65 — all correct.
Their pooled FVC estimate of 111.28 mL essentially reproduces the trialists' own prespecified combined-dataset difference of 111.8 mL (95% CI 79.7–144.0).

That last point is reassuring for accuracy but raises the central scientific question below.

Points adequately resolved

Small-k statistics (Major 4): The addition of fixed-effect, REML, and modified Hartung–Knapp analyses, plus explicit caution that I²/τ² are uninterpretable at k=2, is exactly right. The Hartung–Knapp FVC interval (−106.99 to 329.56) honestly conveys the fragility.
FVC interpretation (Major 5): Separating a trial-level between-group effect from a patient-level MCID is correct; the ~2–6% / ~100 mL MCID framing is consistent with du Bois et al..
GRADE, RoB 2, direct-IPF vs PH-ILD separation (Majors 6–8): Well handled, and keeping INCREASE/OLE/Japanese data as narrative-only is the correct architecture.
Endpoint definitions, terminology, safety categorization (Minors): Adequately addressed; the clinical-worsening composite now matches the trial definition.

Remaining issues to fix

Search date precedes publication of the two pivotal trials (must clarify). The final search is stated as 3 July 2026, yet TETON-1 and TETON-2 (refs 29–30) were published 9 July 2026. As written, the searches could not have retrieved the very trials that drive every pooled estimate. The authors must explain the capture mechanism (online-first/preprint, or a later verification search) and reconcile the dates — otherwise the PRISMA identification is internally inconsistent.
PROSPERO registration still shows "[TO BE INSERTED]." The retrospective-registration disclosure is handled honestly, but the placeholder cannot remain; the actual number is required.
Google Scholar counted as a primary identification source (913 of 1552 records). This remains methodologically nonstandard — Google Scholar is generally recommended as a supplementary, not primary, database because its results are non-reproducible and poorly deduplicated. This should at minimum be named as a limitation, and ideally the PRISMA count should distinguish it. The absence of Embase compounds this.
Added value over the trials' own prespecified combined analysis (conceptual). Because TETON-1 and TETON-2 are from a single coordinated program (same sponsor, protocol family, and a prespecified pooled dataset already published), pooling them independently largely re-derives an existing result rather than synthesizing independent evidence. The framing of the two trials as "independent replicate" studies for random-effects purposes overstates their independence. The Discussion should state plainly what the meta-analysis contributes beyond the published combined analysis (e.g., formal safety pooling, GRADE, small-k sensitivity) and temper "independent" language.
Incomplete reference 29 ("Nathan SD, et al." with no full author list, volume, or pages) — complete it to match ref 30.
Nerandomilast omission (context). Since acceptance, nerandomilast (FIBRONEER-IPF, FVC difference 68.8 mL vs placebo) has been published and FDA-approved for IPF. A sentence situating inhaled treprostinil among newly available options would keep the Discussion current, though this is optional.

None of these require re-analysis except possibly the search-date reconciliation; the rest are text-level corrections.

Author Response

Response to Reviewers

Manuscript: Efficacy and Safety of Inhaled Treprostinil in Idiopathic Pulmonary Fibrosis: A Systematic Review and Meta-Analysis

Manuscript ID: arm-4541865

Reviewer 1 - Major Revision

Overall Assessment

Response: We appreciate the reviewer's positive assessment of the clinical relevance of the question and agree that the methodological and interpretative points required fuller explanation. The revision therefore focuses on reproducibility of the search, protocol chronology, independent screening and data verification, the limitations of meta-analysis with only two direct-IPF trials, additional sensitivity analyses, GRADE certainty assessment, clearer risk-of-bias interpretation and more cautious statements regarding clinical significance and mechanism.

Major Comment 1. Systematic review methodology and reproducibility

Response: We agree. The Methods now state that the final search was run on 3 July 2026 and identify the principal search sources as PubMed, Google Scholar and the Cochrane Library/CENTRAL, with Google Scholar clarified as a broad supplementary identification source within the search strategy. The complete database-specific search strategies have been reproduced in the supplementary material rather than summarized only by search concepts. Supplementary Table S1 and Supplementary Search Appendix S1A provide the exact PubMed strings, MeSH and title/abstract fields, Boolean operators, publication-date syntax, the four Google Scholar queries with the custom year range 2000-2026, and the three Cochrane searches.

Major Comment 2. Protocol registration

Response: The protocol was developed and finalized before study screening commenced. It specified the eligibility criteria, outcomes and planned analytical approach. It was not, however, registered prospectively in PROSPERO. PROSPERO registration was completed retrospectively as CRD420261497663. We have revised the Methods to state this chronology explicitly and have added retrospective registration as a methodological limitation. We have not described the registration as prospective or assigned a retrospective registration date as if it preceded screening. The registered PROSPERO number is CRD420261497663 and has been inserted in the revised manuscript and supplementary material.

Major Comment 3. Study selection and data extraction

Major Comment 4. Meta-analysis based on limited studies

Response: We agree that the small number of contributing trials is the principal statistical limitation of the review. We retained pooling because TETON-1 and TETON-2 are two separately randomized phase 3 trials within the coordinated TETON program addressing the same direct-IPF question, using the same intervention, placebo comparator and 52-week outcome framework. We now avoid describing them as independent replicate programs and explicitly recognize the prespecified combined TETON analysis. Pooling therefore provides a useful summary estimate across two clinically comparable trials, but it does not make estimation of between-study heterogeneity reliable.

Major Comment 5. Interpretation of the FVC effect

Major Comment 6. Risk of bias

Please provide a detailed study-level risk-of-bias assessment and explain how the identified methodological limitations influenced the interpretation of the pooled results.

Major Comment 7. Certainty of evidence

Consider incorporating a GRADE assessment for the major outcomes. This would help readers understand the certainty and clinical reliability of the available evidence.

Major Comment 8. Direct IPF versus PH-ILD evidence

Response: We agree and have strengthened this distinction throughout the manuscript. The primary quantitative synthesis is restricted to TETON-1 and TETON-2, the two separately randomized direct-IPF trials within the coordinated TETON program. INCREASE enrolled PH-ILD and had a 16-week follow-up; its IPF subgroup FVC findings are therefore retained as secondary supportive/sensitivity evidence rather than pooled with the 52-week direct-IPF trials. The open-label extension and Japanese PH-ILD studies remain narrative only.

Reviewer 1 - Minor Comments

Minor Comment 1. Definitions of clinical worsening and acute exacerbation of IPF

Clarify definitions of clinical worsening and acute exacerbation of IPF.

Minor Comment 2. Pulmonary-function terminology

Ensure consistent terminology for FVC, DLCO, and other pulmonary function parameters.

Minor Comment 3. Effect estimates and confidence intervals

Please standardize reporting of effect estimates and confidence intervals throughout the manuscript.

Minor Comment 4. Safety terminology

The safety outcomes should clearly distinguish treatment-emergent adverse events from serious adverse events.

Minor Comment 5. Cautious conclusion

The conclusion should be appropriately cautious and reflect the limited number of direct randomized studies.

Response: The conclusion has been moderated. It now states that the findings are based on two separately randomized phase 3 direct-IPF trials within the coordinated TETON program, that a mortality benefit has not been established and that the findings do not establish a disease-modifying antifibrotic effect. PH-ILD subgroup, open-label and Japanese data are described as supportive and are kept separate from the primary direct-IPF meta-analysis.

Reviewer 2 - Major Revision

Overall Assessment

Major Comment 1. Statistical methodology

Response: We retained the DerSimonian-Laird random-effects model as the primary analysis because it was the prespecified/submitted approach and because the two TETON trials are separately randomized trials within the same coordinated phase 3 program. We now explicitly state that with only two studies, tau squared and I squared are poorly estimated and should not be treated as precise measures of between-study heterogeneity.

Major Comment 2. Handling of potentially overlapping datasets

Response: We have made the parent-dataset structure explicit. The 11 included reports represent four principal patient-level datasets: TETON-1, TETON-2, INCREASE and the Japanese PH-ILD cohort. TETON-1 and TETON-2 are separately randomized trial datasets within the coordinated TETON program. TETON-1 and TETON-2 alone contribute to the primary quantitative synthesis. A combined TETON analysis is not pooled with the individual trials. The INCREASE parent trial and its post-hoc publications are recognized as a single shared cohort, and the Japanese initial and long-term publications are recognized as the same Japanese cohort.

Major Comment 3. Outcome hierarchy and multiplicity

Major Comment 4. Continuous and time-to-event outcomes

Major Comment 5. Sensitivity analyses

Major Comment 6. Risk-of-bias assessment

Major Comment 7. Clinical interpretation of secondary outcomes

Major Comment 8. Safety analysis

Major Comment 9. Mechanistic interpretation

Major Comment 10. Conclusions

Response: The conclusion has been revised to reflect the limited evidence base. It now identifies the evidence as two separately randomized phase 3 direct-IPF trials within the coordinated TETON program, supports preservation of FVC and reduction in clinical-worsening events, and states that a mortality benefit has not been established. It also avoids equating physiologic preservation with proven long-term survival or a disease-modifying effect. PH-ILD subgroup, open-label and Japanese data remain supportive but separate from the direct-IPF meta-analysis.

Reviewer 2 - Minor Comments

Minor Comment 1. Study and participant counts in forest plots

Please ensure that all forest plots identify the number of studies and participants contributing to each analysis.

Response: We agree. The forest-plot set for resubmission will explicitly identify the number of contributing studies and participants. The primary direct-IPF pooled analyses contain the two separately randomized TETON trials, with 1191 randomized participants in total; outcome-specific safety denominators are retained where they differ because of the safety populations. The legends and plot labels will be synchronized with the corresponding Results and supplementary tables.

Minor Comment 2. Abbreviations

Define all abbreviations at first appearance.

Minor Comment 3. Follow-up duration

Clarify the rationale for the selected follow-up duration for each outcome.

Minor Comment 4. Consistency between manuscript, tables and supplementary material

Ensure consistency between the text, tables, and supplementary material.

Minor Comment 5. Overstatement in abstract and conclusion

Please review the manuscript for overstatement of efficacy in the abstract and conclusion.

Closing statement

Additional final-round feedback

Search date precedes publication of the two pivotal trials

Response: We clarified the publication chronology without changing the search date or PRISMA counts. The final search was run on 3 July 2026. TETON-2 had been published online ahead of issue on 11 March 2026 and TETON-1 on 18 May 2026, so both reports were available before the final search. Their 9 July 2026 date is the later issue publication date. The Methods and Supplementary Search Appendix now state this explicitly, with references 29 and 30 providing the publication details.

PROSPERO registration still shows "[TO BE INSERTED]"

Response: Corrected. The retrospective PROSPERO registration number is CRD420261497663 and has been inserted in the Methods and supplementary material. The manuscript continues to state clearly that the protocol was finalized before screening but registered retrospectively.

Google Scholar counted as a primary identification source

Response: We retained the original PRISMA counts and did not reconstruct the identification set. We revised the terminology to describe PubMed, Google Scholar and Cochrane Library/CENTRAL as principal search sources, and clarified that Google Scholar functioned as a broad supplementary identification source within the search strategy. We also added a limitation noting that Google Scholar has dynamic ranking and limited reproducibility, and that Embase was not searched because access was unavailable.

Added value over the trials' own prespecified combined analysis

Response: We agree that TETON-1 and TETON-2 are separately randomized trials within one coordinated development program and should not be framed as independent external replication. We therefore replaced "independent" and "replicate" language with "two separately randomized phase 3 trials within the coordinated TETON program." The Discussion now states that the TETON program had already reported a prespecified combined analysis and clarifies the added value of this review: outcome-specific synthesis across the broader evidence base, formal safety pooling, small-k sensitivity analyses, GRADE certainty assessment, and explicit separation of direct-IPF evidence from overlapping or indirect PH-ILD evidence. No re-analysis was required.

Incomplete reference 29

Response: References 29 and 30 were both completed against the published NEJM records. Reference 29 now includes Nathan SD, Smith P, Deng C, et al., the full published title, volume 395 and pages 115-126. Reference 30 now includes volume 395 and pages 127-137.

Nerandomilast omission

Response: We added one brief Discussion sentence placing inhaled treprostinil in the evolving IPF treatment landscape. FIBRONEER-IPF is cited to note that nerandomilast reduced 52-week FVC decline versus placebo. Nerandomilast was not added to the systematic-review evidence set, PRISMA diagram or meta-analysis because it is a different intervention and falls outside the review PICO.

Author Response File: Author Response.pdf