No transportability-first PAIC workflow matches MAIC’s usability

Adjacent-field methods

Very high priority

Overstated

PAIC has already been restated in transportability and estimand terms: Chandler and Ishak (arXiv:2602.17041) reframe PAICs as transportability problems and show that pairwise PAIC identifies an effect defined in the comparator population, so applying it elsewhere is an additional transport step, and Remiro-Azocar (arXiv:2210.01757) treats the transportability of marginal versus conditional estimands in evidence synthesis. The estimators the source calls underused also exist for PAIC: Park et al. (doi:10.1016/j.jval.2023.11.011) implement and compare inverse odds weighting, regression adjustment and a doubly robust method for unanchored time-to-event comparisons, Campbell and Remiro-Azocar (arXiv:2505.00113) give augmented doubly robust MAIC, and drMAIC 0.1.0 ships it. What has not been done is the usability half: no transportability-first workflow, guidance document or reporting template matches MAIC’s one-page explanation, and routine HTA submissions still do not use these estimators. One auditor read that remainder as a workflow and communication objective rather than a well-posed statistical open problem.

ADJ-01 Overstated Very high priority Promising

Source note. This entry derives in part from a source document whose inline citations were ChatGPT interface tokens rather than references, and so resolve to nothing. Claims below are attributed to that document and were checked independently where possible; anything that could not be independently sourced is marked in the verification trail.

Statement

Why it is open

The recasting exists on paper but nothing carries it into practice. The transport framing requires nuisance models, an explicit target covariate distribution and a diagnostic suite that MAIC’s single weighting step conceals, so the explainable option keeps winning over the more rigorous one. No guidance, training material or reporting template has been rebuilt around the transport vocabulary, so a submission written in transport terms costs the analyst more and buys no procedural credit. The case where the framing matters most, a target described only by published summaries, is also the case where standard transport estimators do not apply directly, so the umbrella cannot simply be imported.

What has been tried

Work	What it contributes
Chandler & Ishak, Reframing Population-Adjusted Indirect Comparisons as a Transportability Problem: An Estimand-Based Perspective and Implications for Health Technology Assessment (2026)	Performs the recasting this entry called open: formalizes transportability in PAIC from an estimand perspective, shows pairwise PAIC identifies a comparator-population effect and that using it elsewhere is an extra transport step. Does not address the accessibility half.
Remiro-Azocar, Transportability of model-based estimands in evidence synthesis, Stat Med 2024;43(22):4217-4249	Transportability of marginal versus conditional estimands across studies in evidence synthesis, including the non-collapsibility obstruction.
Park et al., Unanchored Population-Adjusted Indirect Comparison Methods for Time-to-Event Outcomes Using Inverse Odds Weighting, Regression Adjustment, and Doubly Robust Methods, Value in Health 2024	Implements and compares inverse odds weighting, regression adjustment and a doubly robust method for unanchored PAIC with time-to-event outcomes; contradicts the claim that PAIC methodology underuses all three.
Campbell & Remiro-Azocar, Doubly robust augmented weighting estimators for the analysis of externally controlled single-arm trials and unanchored indirect treatment comparisons (2025)	Adds doubly robust augmentation (outcome model plus balancing weights) to MAIC for unanchored and external-control comparisons.
drMAIC 0.1.0: Doubly Robust Matching-Adjusted Indirect Comparison for HTA	Ships doubly robust MAIC with NICE DSU TSD18-aligned outputs, so the doubly robust option is available to practitioners rather than only in theory.

Probable solution or research direction

The derivation work is done, so the direction is dissemination and tooling rather than a new framework: convert the published estimand-based recasting into a workflow, checklist and reporting template whose inputs, outputs and one-page explanation are as simple as MAIC’s. Validate the doubly robust and inverse-odds estimators that already exist for PAIC in realistic HTA settings and get them into guidance, so the causally principled option becomes the operationally convenient one. Measure success by whether a committee can follow the analysis, not by whether the estimator is new.

Verification

Verdict. Overstated. 2 independent auditors judged the claim as stated too strong. PAIC has now been formally restated in transportability/estimand terms in the methods literature, but that restatement has not been converted into a workflow, guidance and reporting template with MAIC’s operational simplicity, and routine HTA submissions still do not use inverse-odds, standardisation or doubly robust estimators.

The source titled this. “Recasting PAIC as a transportability problem without losing usability” The title poses the transportability recasting as the open work, but the audit found it already published (arXiv:2602.17041, arXiv:2210.01757), leaving only the usability half.

The source said. “Population-adjusted indirect comparison is a restricted case of the causal transportability problem, yet it is practised as a set of bespoke HTA estimators with their own vocabulary. The open question is whether PAIC can be rebuilt on an explicit source/target transport framework while keeping the practical accessibility that made MAIC attractive.”

What was wrong with it. The source cited nothing and claimed both that the recasting was undone and that PAIC underuses inverse-odds weighting, standardization and doubly robust estimators; the literature auditor found the recasting published (arXiv:2602.17041, arXiv:2210.01757) and all three estimators implemented for PAIC (doi:10.1016/j.jval.2023.11.011, arXiv:2505.00113, drMAIC 0.1.0). The claim was unattributed in the source and is now sourced; what survives is the usability half, which GPT-5.6 Sol read as a workflow objective rather than a statistical problem.

Dissent. grok dissented (supported-with-caveat): Inverted-prior search finds a near-literal hit: Chandler and Ishak (arXiv:2602.17041, 2026) reframe PAICs as transportability problems from an estimand-based HTA perspective, distinguish conditional versus marginal transport, and clarify when SEMA alone fails for non-collapsible measures. That paper formalizes and diagnoses the dual-transport structure of pairwise MAIC/STC; it does not ship a transportability-first workflow with MAIC-like one-page accessibility, software, training, or reporting templates. Separately, Campbell and Remiro-Azócar (arXiv:2505.00113, 2025) import doubly robust augmentation into entropy-balancing MAIC for unanchored and external-control settings, so C4 is already partly outdated for pairwise unanchored work. C2 is supported as a restricted-case relationship; C3 remains broadly true of HTA practice; C6 is only partly true because target-AgD transport methods exist but still struggle where only published summaries are available; C8 and C9 remain largely open on the usability half of the dual requirement. Queries: PAIC transportability Remiro/Phillippo; Chandler PAIC transportability; DR MAIC; TransportHealth; RSM/SiM 2024–2026 PAIC transport; NICE DSU/EUnetHTA 2025–2026.

Literature and prior-art check: partially-solved / overstated. prior_work is empty, yet a preprint whose title is almost this entry’s title exists and is cited elsewhere in this same corpus (EST-02, EST-03): Chandler & Ishak, ‘Reframing Population-Adjusted Indirect Comparisons as a Transportability Problem: An Estimand-Based Perspective and Implications for HTA’ (arXiv:2602.17041, Feb 2026, rev. Jul 2026). Its abstract: ‘We examine and formalize transportability in PAICs from an estimand-based perspective’ and ‘pairwise PAIC approaches typically identify effects defined in the comparator population and … applying these estimates to other populations entails an additional, often implicit, transport step.’ Remiro-Azocar (arXiv:2210.01757, Stat Med 2024) does the same for transportability of estimands. The recast half is done. The ‘underuses inverse-odds weighting, standardisation and doubly robust estimators’ claim is also weak: Park et al. (Value in Health 2024) implement exactly those three for unanchored PAIC, Campbell & Remiro-Azocar (arXiv:2505.00113) give augmented doubly robust MAIC, and drMAIC 0.1.0 ships it. What genuinely remains open is the usability half: no transportability-first workflow matches MAIC’s one-page explanation. Cites arxiv:2602.17041: Performs the recasting this entry calls open: formalizes transportability in PAIC from an estimand perspective, shows pairwise PAIC identifies a comparator-population effect and that using it elsewhere is an extra transport step. Does NOT address the accessibility/usability half.. Cites arxiv:2210.01757: Transportability of marginal vs conditional estimands across studies in evidence synthesis, including the non-collapsibility obstruction.. Cites doi:10.1016/j.jval.2023.11.011: Directly contradicts ‘PAIC underuses inverse-odds weighting, standardisation and doubly robust estimators’ at the methods level; all three are implemented and compared for time-to-event PAIC.. Cites cran:drMAIC 0.1.0: Ships doubly robust MAIC with NICE DSU TSD18-aligned outputs, so the doubly robust option is available to practitioners, not only in theory..
GPT-5.6 Sol, technical and source-code lens: not-a-problem / overstated. As phrased, this is a workflow and communication objective rather than a well-posed statistical open problem. C2 is defensible only after specifying the source, target, causal estimand, treatment versions, conditional transportability and positivity; PAIC procedures are not automatically valid transport estimators. C3-C5 and C7-C10 are unsourced empirical, causal or normative assertions, not mathematical consequences. Aggregate-only targets make direct transport estimation harder; that does not establish C6’s claim that the framing matters “most” there. The asserted very-high severity is therefore unsupported.
Grok 4.5, recency lens: partially-solved / supported-with-caveat. Inverted-prior search finds a near-literal hit: Chandler and Ishak (arXiv:2602.17041, 2026) reframe PAICs as transportability problems from an estimand-based HTA perspective, distinguish conditional versus marginal transport, and clarify when SEMA alone fails for non-collapsible measures. That paper formalizes and diagnoses the dual-transport structure of pairwise MAIC/STC; it does not ship a transportability-first workflow with MAIC-like one-page accessibility, software, training, or reporting templates. Separately, Campbell and Remiro-Azócar (arXiv:2505.00113, 2025) import doubly robust augmentation into entropy-balancing MAIC for unanchored and external-control settings, so C4 is already partly outdated for pairwise unanchored work. C2 is supported as a restricted-case relationship; C3 remains broadly true of HTA practice; C6 is only partly true because target-AgD transport methods exist but still struggle where only published summaries are available; C8 and C9 remain largely open on the usability half of the dual requirement. Queries: PAIC transportability Remiro/Phillippo; Chandler PAIC transportability; DR MAIC; TransportHealth; RSM/SiM 2024–2026 PAIC transport; NICE DSU/EUnetHTA 2025–2026. Cites arxiv:2602.17041: Conceptual and estimand-level recasting of PAIC as transport (conditional versus marginal, SEMA limits, HTA two-step transport). Closes the pure-vocabulary gap; does not close the MAIC-level accessibility product gap.. Cites arxiv:2505.00113: Imports DR (outcome model plus balancing weights) into MAIC-style PAIC for unanchored/external-control pairwise settings. Does not rebuild a full transportability-first HTA stack or network workflow..
Decision path: R3:partially-solved(n=2) → R8:overstated-majority(n=2)

Source

Derived from A:L50, A:L65, A:L80, A:L223 of the reviewed corpus.