No assumption map spans identification, transport and consistency
Statement
‘Similarity’, ‘transitivity’, ‘exchangeability’, ‘conditional constancy’ and ‘shared effect modification’ are used loosely and often interchangeably in applied ITC and PAIC writing, and no single agreed assumption map ties within-study identification, cross-study transport, target positivity, treatment and outcome consistency and network coherence together for a given estimator. Formal notation for the individual restrictions already exists: Phillippo et al. (doi:10.1177/0272989X17725740, doi:10.1177/0272989X221117162) write conditional constancy and the shared effect modifier assumption as explicit restrictions tied to a named estimator and a declared target population. The five headings proposed here are one useful non-exhaustive taxonomy rather than an exact decomposition, since the categories overlap and they subsume or omit sampling exchangeability, positivity, interference, censoring, missingness, measurement and model or integration assumptions.
Why it is open
In applied writing the terms function as slogans, so an analysis can satisfy one restriction while silently violating another and a reader cannot tell which restriction the estimator actually needs. The restrictions also sit at different levels, inside one trial, between two trials and across a whole network, and are protected by different design features, so a single verbal claim that transitivity holds cannot be stated precisely enough to refute. The two works the source cited state the anchored and unanchored assumption asymmetry but not a level-by-level map, so nothing currently obliges a submission to say which restriction each part of its analysis leans on.
What has been tried
| Work | What it contributes |
|---|---|
| Phillippo et al. 2018, Medical Decision Making | sets out population-adjustment methods for HTA and distinguishes anchored from unanchored comparisons according to whether a common comparator arm is used, stating that unanchored comparisons make much stronger assumptions that are widely regarded as infeasible |
| NICE DSU Technical Support Document 18: Methods for Population-Adjusted Indirect Comparisons in Submissions to NICE | official guidance stating that unanchored MAIC and STC assume all effect modifiers and prognostic factors are accounted for, an assumption it calls very strong and largely considered impossible to meet, and that anchored approaches are preferred whenever a connected network exists |
| Phillippo et al. 2023, Validating the Assumptions of Population Adjustment: Application of Multilevel Network Meta-regression to a Network of Treatments for Plaque Psoriasis, Medical Decision Making | assesses population-adjustment assumptions inside a nine-study network: splits the common class interaction into treatment-specific interactions one covariate at a time, checks conditional constancy through residual heterogeneity and inconsistency in sufficiently large networks, and reports that the shared-interaction restriction was required to identify the model where a treatment was informed only by aggregate data |
Probable solution or research direction
Since the notation for individual restrictions is settled, the task is assembly and adoption rather than invention: build a map that lists, for each estimator, which restriction it requires, at which level, and against which declared target population and effect scale, treating the five headings as a checklist that must also carry positivity, missingness, measurement and interference. Test it as a reporting device on real submissions and check whether readers using it identify the binding restriction more reliably than prose does. Attach negative controls, held-out-trial falsification and bias functions to each restriction separately rather than to the analysis as a whole.
Verification
Verdict. Overstated. 2 independent auditors judged the claim as stated too strong. The ITC/PAIC literature uses ‘similarity’, ‘transitivity’, ‘exchangeability’, ‘conditional constancy’ and ‘shared effect modification’ loosely and often interchangeably in applied writing, and no single agreed assumption map ties within-study identification, across-study transport, target support, treatment-version consistency and network consistency together for a given estimator; formal notation for individual restrictions (notably conditional constancy and shared effect modification, each tied to a declared target population) does already exist.
The source titled this. “Assumption vocabulary conflates five distinct causal restrictions” The title asserts an exact decomposition into five restrictions, which both auditors rejected as an overlapping, non-exhaustive taxonomy.
The source said. “Similarity, transitivity, exchangeability, conditional constancy and shared effect modification are used in the ITC/PAIC literature as if interchangeable, when they denote different causal restrictions. The report proposes writing assumptions instead as an explicit five-part decomposition: within-study causal identification, across-study effect transport, target support, outcome and treatment-version consistency, and network consistency.”
What was wrong with it. The source claimed the assumptions decompose into exactly five restrictions and that no standard formal notation maps a restriction to an estimator and a target population. Both auditors rejected that: the notation exists for conditional constancy and shared effect modification (doi:10.1177/0272989X17725740, doi:10.1177/0272989X221117162), and the five boxes overlap while leaving out positivity, interference, missingness and measurement, so what is genuinely absent is a single map spanning all levels rather than notation as such.
Revised by the full-text reading. Applied 2026-07-26 full-text reading of the consolidated library (687 papers).
- closure-upheld
- 13 findings from the reading assert this problem is open; the closure was re-examined and stands Put to an independent reviewer with the closing auditors’ reasoning, the work they relied on, and the new findings in view together. The new evidence strongly documents unsettled terminology, inconsistent reporting, and the absence of an agreed comprehensive map, all of which the closure already concedes. It does not contradict the closure’s counterpoint that conditional constancy and shared effect modification have formal estimator- and target-specific expressions in the PAIC literature. Surveys showing that applied authors conflate transitivity, similarity, consistency, and exchangeability establish the practical need for better reporting, but do not establish that formal notation for individual restrictions is absent. The present entry already incorporates the proper residual by treating its five headings as a useful but non-exhaustive taxonomy rather than an exact decomposition. (reviewer confidence: high)
- Flagged for reopening
-
19 findings from the reading assert this problem is open while the registry carries
overstatedFlagged, not applied: reversing an earlier audit’s closure needs the closure’s own evidence re-examined alongside this, which the reading did not do. See documentation/audit/reading/PROPOSED.md. - Literature and prior-art check
- open / overstated. Both cited works check out. Phillippo et al. 2018 (abstract, retrieved): ‘We describe the assumptions required for population-adjusted indirect comparisons, and demonstrate how these may be used to generate comparisons in any given target population. We distinguish between anchored and unanchored comparisons according to whether a common comparator arm is used or not.’ NICE DSU TSD 18 is the guidance companion. Neither states a five-part decomposition, so both citations are decorative with respect to the five-restriction taxonomy, which is the entry’s own proposal. Two claims overreach. ‘Exactly these five restrictions’ is a taxonomy assertion no retrieved source makes. ‘There is currently no standard formal notation in the ITC/PAIC literature that maps each of these restrictions to a specific estimator and target population’ is too strong: Phillippo et al. 2018 and 2023 both write conditional constancy of relative effects and the shared effect modifier assumption as explicit restrictions tied to a named estimator and a named target population, and the transportability literature does the same for S-indexed estimands. What is genuinely absent is a single agreed assumption map spanning all five levels.
- GPT-5.6 Sol, technical and source-code lens
- open / overstated. The proposed five-part map is a useful reporting device, not an exact mathematical decomposition. Its categories overlap and compress multiple restrictions, while omitting or subsuming sampling exchangeability, positivity, interference, censoring, missingness, measurement, and model or integration assumptions. Similarity, transitivity, and exchangeability are also overlapping or synonymous under some formalizations, unlike conditional constancy and shared effect modification, which are more clearly distinct. C3 is estimand-dependent: randomization identifies an assigned-treatment effect without an extra intercurrent-event assumption under a treatment-policy strategy when outcomes are complete; additional assumptions arise for hypothetical strategies or missing outcomes. C4 also needs consistency and target positivity. C5 describes nonparametric positivity but mixes identification with finite-sample precision. C7 is a literature-wide absence claim outside the assigned lens. Explicit assumption mapping remains useful, but the exact five-box taxonomy is a roadmap convention rather than a very-high-priority identification result.
- Decision path
-
R5:default-open → R8:overstated-majority(n=2)
How the evidence falls in time
19 papers in the reviewed corpus bear on this problem, published 2013 to 2026. Progress and a fresh assertion of openness land in the same year.
This problem is recurrent: three or more papers spanning at least eight years assert it independently. Sustained restatement by authors who mostly do not cite each other is the strongest evidence this reading can offer that a gap is real rather than one group’s framing.
It is also being closed in pieces: partial results come from more than one paper in more than one year, each covering a different part.
| Year | Says | Paper |
|---|---|---|
| 2013 | partly addresses it | Is network meta-analysis as valid as standard pairwise meta-analysis? It all depends on the distribution of effect modifiers |
| 2014 | confirms it is open | The quality of reporting methods and results in network meta-analyses: an overview of reviews and suggestions for improvement |
| 2016 | partly addresses it | Assumptions of Mixed Treatment Comparisons in Health Technology Assessments - Challenges and Possible Steps for Practical Application |
| 2017 | partly addresses it | Methods for Population-Adjusted Indirect Comparisons in Health Technology Appraisal |
| 2018 | confirms it is open | Mapping the characteristics of network meta-analyses on drug therapy: A systematic review |
| 2018 | confirms it is open | Network meta-analysis explained |
| 2018 | confirms it is open | Network meta-analysis, a new statistical technique at urologists’ disposal to improve decision making |
| 2021 | partly addresses it | Transportability of Causal Effects: Completeness Results (Bareinboim & Pearl, Proceedings of the Twenty-Sixth AAAI Conference on Artificial Intelligence, pp. 698-704) |
| 2021 | confirms it is open | Concepts and emerging issues of network meta-analysis |
| 2021 | confirms it is open | Reporting of Cochrane systematic review protocols with network meta-analyses-A scoping review |
| 2022 | confirms it is open | Knowledge user survey and Delphi process to inform development of a new risk of bias tool to assess systematic reviews with network meta-analysis (RoB NMA tool) |
| 2024 | partly addresses it | Twenty years of network meta-analysis: Continuing controversies and recent developments |
| 2024 | confirms it is open | Methods for Indirect Treatment Comparison: Results from a Systematic Literature Review |
| 2024 | confirms it is open | Low awareness of the transitivity assumption in complex networks of interventions: a systematic survey from 721 network meta-analyses |
| 2024 | confirms it is open | Mastering meta-analysis in Microsoft Excel with MetaXL add-in: A comprehensive tutorial and guide to meta-analysis |
| 2025 | confirms it is open | Reporting and evaluation of assumptions and certainty of evidence in network meta-analyses |
| 2025 | confirms it is open | Selection of indirect treatment comparisons for health technology assessments: a practical guide for health economics and outcomes research scientists and clinicians |
| 2026 | partly addresses it | Reframing Population-Adjusted Indirect Comparisons as a Transportability Problem: An Estimand-Based Perspective and Implications for Health Technology Assessment (Chandler & Ishak; arXiv:2602.17041v3). The batch file points at media/media/Table_1_v2.pdf; the full LaTeX source Transportability_PAIC.tex sits in the same arXiv package and was read. |
| 2026 | confirms it is open | Evidence Synthesis via Indirect Treatment Comparisons in the European Framework of Joint Clinical Assessment |
Source
Derived from C:L222, C:L380-392 of the reviewed corpus.