No decision-value method chooses which IPD or bridging trial to get
Statement
Methods for choosing the next data collection by expected decision value exist and are mature: conditional trial design sizes and configures a new trial against an existing network to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value (Salanti et al. 2018, doi:10.1186/s13063-018-2740-2; Nikolakopoulou et al. 2014, pmid:25225031), and expected value of sample information has practical tutorials and worked prioritization examples (doi:10.1177/0272989X211026292; doi:10.1136/bmjopen-2024-092158). Neither has been applied to deciding which individual patient dataset to acquire for a population-adjusted comparison, nor to sizing and allocating a bridging trial that would connect two disconnected subnetworks. Whether acquisition in ITC practice is in fact opportunistic is undocumented: no auditor found a survey of it.
Why it is open
The general machinery is decision-theoretic and only produces an answer once the decision, utility, costs, access probability, model uncertainty and target population are specified; for a population-adjusted comparison none of these is currently packaged as an input to acquisition. The value of an additional IPD study depends on the target population, the bridge in use and the decision threshold, and no published formulation ties those to a request for a specific dataset. For disconnected networks the harder version, choosing which bridge treatment to run and how to allocate sample across the gap, was not found by any auditor. What is open is the ITC-specific operational formulation and its robustness, not the existence of a principled criterion.
What has been tried
| Work | What it contributes |
|---|---|
| Salanti et al. 2018, Trials | Four-step framework identifying the optimal design for a new trial that updates an existing network while minimizing required sample size, allowing the new trial to omit some interventions and still contribute direct and indirect evidence. |
| Nikolakopoulou et al. 2014, Biometrical Journal | Derives conditional power of a network meta-analysis to inform the design of future trials; earliest of the conditional trial design papers and establishes that sample size can be computed conditional on the existing network. |
| Heath et al. 2022, Medical Decision Making | Practical guidance and recommendations for four model-based methods of computing expected value of sample information efficiently; shows the machinery for deciding what data to collect is mature and tutorialized. |
| Glynn et al. 2025, BMJ Open (DiaDeM) | Worked example of value-of-information analysis used to prioritize which data to collect, showing the acquisition-as-decision-object framing is present in the wider literature. |
Probable solution or research direction
Adapt the existing value-of-information and conditional trial design machinery to the ITC setting rather than build new criteria: specify the decision, utility, costs, access probability and target population, then rank candidate IPD acquisitions by expected decision value. Extend conditional trial design from adding a trial to a connected network to the disconnected case, where the objects of choice are the bridge treatment, the sample size and the allocation across the gap. Establishing whether current acquisition is opportunistic requires a survey of practice, which no auditor could find, so that claim should be tested before it is used to justify the programme.
Verification
Verdict. Partially addressed. Auditors disagreed and the disagreement was evidence-bearing on both sides. Reopened by the full-text reading: 3 findings from 3 papers survived both the reader and an independent second reviewer, so the verdict moves from unverifiable to partially-addressed.
The source titled this. “Data acquisition is opportunistic rather than decision-driven” The title asserts that acquisition is opportunistic, which the audit recorded as undocumented since no survey of ITC practice was found.
The source said. “Decisions about which individual patient dataset to obtain, or which bridging trial to run, are made opportunistically rather than by any principled criterion linking the acquisition to the decision it would improve.”
What was wrong with it. The source said no formal object exists to quantify what a dataset would contribute and listed bridging-trial design as a research direction; conditional trial design for networks has been published since 2014 (pmid:25225031; doi:10.1186/s13063-018-2740-2) and EVSI tooling is mature (doi:10.1177/0272989X211026292), as the literature auditor documented and GPT-5.6 Sol confirmed on decision-theoretic grounds. Grok agreed the ITC-specific application is absent but read the entry as supported, so the disagreement here is over framing rather than over what exists.
Dissent. literature dissented (overstated): Claim 3 is the weak point. Salanti et al. 2018 Trials present a four-step conditional trial design framework identifying the optimal design for a new trial updating an existing network while minimizing required sample size, explicitly allowing the new trial to omit some interventions while still contributing direct and indirect evidence. Nikolakopoulou et al. 2014 Biometrical Journal derive conditional power of an NMA to inform future trial design; Nikolakopoulou 2019 Trials surveys methodologists on it; Siontis 2022 F1000Research applies conditional sample size calculation. Designing the next trial from an existing network is an established literature, not a proposed research direction. What I could not find in any retrieved source is the specific case the entry names: sizing and allocating a bridging trial across the gap between disconnected subnetworks. Claim 2 is similarly half-built: EVSI machinery for deciding what data to collect is mature (Heath 2022 MDM, Vervaart 2023/2024 MDM, Glynn 2025 BMJ Open), but I found no application to choosing which IPD to acquire for an ITC. Claim 1 concerns current practice and I found no survey of it; Clayton 2017 Trials reports only 6% of trialists using VOI, which is adjacent but a different population.
Revised by the full-text reading. Applied 2026-07-26 full-text reading of the consolidated library (687 papers).
- Corrected
- statement: corrected a factual claim Salanti et al. 2018 minimizes sample size for a target conditional power or precision. The words ‘net benefit’ and ‘utility’ do not appear in it. The entry’s other citation, Heath et al. 2022 on expected value of sample information, is decision-value and is correctly described. Previously: “Methods for choosing the next data collection by expected decision value exist and are mature: conditional trial design sizes and configures a new trial against an existing network (Salanti et al. 2018, doi:10.1186/s13063-018-2740-2; Nikolakopoulou et al. 2014, pmid:25225031), and expected value of sample information has practical tutorials and worked prioritization examples (doi:10.1177/0272989X211026292; doi:10.1136/bmjopen-2024-092158). Neither has been applied to deciding which individual patient dataset to acquire for a population-adjusted comparison, nor to sizing and allocating a bridgi”
- Verdict changed
- unverifiable -> partially-addressed 3 surviving findings from: Connecting a disconnected trial network with a new trial: op; Continuously updated network meta-analysis and statistical m; OP22 Using Threshold Analysis To Guide Searches For Addition. New citations: 10.1017/s0266462322000812, 10.1177/0962280216659896, 10.1186/s12874-023-01896-7 Previously: “Methods for choosing the next data collection by expected decision value exist and are mature: conditional trial design sizes and configures a new trial against an existing network to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision rather than by expected decision value to reach a target power or precision ra”
- Literature and prior-art check
- partially-solved / overstated. Claim 3 is the weak point. Salanti et al. 2018 Trials present a four-step conditional trial design framework identifying the optimal design for a new trial updating an existing network while minimizing required sample size, explicitly allowing the new trial to omit some interventions while still contributing direct and indirect evidence. Nikolakopoulou et al. 2014 Biometrical Journal derive conditional power of an NMA to inform future trial design; Nikolakopoulou 2019 Trials surveys methodologists on it; Siontis 2022 F1000Research applies conditional sample size calculation. Designing the next trial from an existing network is an established literature, not a proposed research direction. What I could not find in any retrieved source is the specific case the entry names: sizing and allocating a bridging trial across the gap between disconnected subnetworks. Claim 2 is similarly half-built: EVSI machinery for deciding what data to collect is mature (Heath 2022 MDM, Vervaart 2023/2024 MDM, Glynn 2025 BMJ Open), but I found no application to choosing which IPD to acquire for an ITC. Claim 1 concerns current practice and I found no survey of it; Clayton 2017 Trials reports only 6% of trialists using VOI, which is adjacent but a different population. Cites doi:10.1186/s13063-018-2740-2: Directly contradicts the framing of claim 3 as a proposed direction: optimal design of the next trial given an existing network is published methodology with worked examples.. Cites pmid:25225031: Earliest of the conditional-trial-design papers; establishes that sample size for a future trial can be computed conditional on the existing network.. Cites doi:10.1177/0272989X211026292: Confirms claim 2’s machinery exists and is tutorialized; the ITC-specific application of choosing which IPD to acquire is what remains unbuilt.. Cites doi:10.1136/bmjopen-2024-092158: Recent worked example of VOI used to prioritise which data to collect, showing the acquisition-as-decision-object framing is not absent from the literature generally..
- GPT-5.6 Sol, technical and source-code lens
- open / overstated. C2 and C3 are technically valid research directions: expected-utility or value-of-information criteria can rank candidate IPD acquisitions, and optimal-design criteria can choose bridge treatments, sample size, and allocation. The central empirical claim C1, that current decisions are generally opportunistic, is unsupported here. The literal statement that no formal object exists is false because decision theory already supplies such objects once the decision, utility, costs, predictive model, access probability, and relevant uncertainties are specified. What remains open is an ITC-specific operational formulation and its robustness, not the existence of a principled criterion. This is generally a roadmap and resource-allocation issue rather than a universal threat to identification or decision-grade validity.
- Grok 4.5, recency lens
- open / supported. Search assumed VOI and optimal design had already made IPD acquisition decision-driven. General value-of-information methods for HTA research prioritization and study design are mature (ISPOR VOI task force reports; Jackson et al. reviews of VOI in policy models; EVPI/EVPPI/EVSI tooling). That infrastructure means C2 is a defensible direction rather than science fiction. However, no retrieved 2020–2026 paper or software release implements a standard, decision-linked criterion for which IPD dataset to request for PAIC/ITC, or optimal sample allocation of a bridging trial across a gap between disconnected subnetworks, as a routine acquisition object. Dose-response and other NMA extensions discuss reconnecting disconnected networks when common agents exist; that is model-based reconnection, not VOI-guided bridge-trial design. No CRAN/GitHub signal for multinma/maicplus/netmeta/outstandR closed this acquisition gap. C1 (opportunistic access in practice) remains descriptive and unfalsified by a competing formal acquisition framework. Queries: VOI/EVSI IPD meta-analysis; VOI network meta research prioritization; optimal bridge trial disconnected NMA; active IPD acquisition strategy. Vote open with moderate-low confidence because VOI is adjacent and may be applied without being published as ITC-specific method.
- Decision path
-
R3:partially-solved(n=1) → R7:counter-and-support-both-present
How the evidence falls in time
7 papers in the reviewed corpus bear on this problem, published 2016 to 2023. The most recent paper to touch this reports progress on it, and every paper calling it open is older.
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.
Does the later work answer the earlier work? Put to an independent reviewer with the older and newer claims side by side, it judged that the later work answers a component of it, leaving a named part (high confidence). The 2023 disconnected-network paper gives a substantive, network-dependent rule for selecting the treatment pair in a new bridging trial. It fixes equal allocation and does not formulate acquisition or trial design by decision value, costs, or utility, so it resolves only one component of the earlier gap. What remains: Choosing which IPD to acquire; decision-value or cost-based selection and sample allocation for a bridging trial; and evidence on whether such methods are used in practice.
| Year | Says | Paper |
|---|---|---|
| 2016 | partly addresses it | Continuously updated network meta-analysis and statistical monitoring for timely decision-making |
| 2016 | confirms it is open | An overview of methods for network meta-analysis using individual participant data: when do benefits arise? |
| 2017 | confirms it is open | The INVEST project: investigating the use of evidence synthesis in the design and analysis of clinical trials |
| 2018 | confirms it is open | Living network meta-analysis compared with pairwise meta-analysis in comparative effectiveness research: empirical study |
| 2019 | confirms it is open | Synthesizing existing evidence to design future trials: survey of methodologists from European institutions |
| 2022 | partly addresses it | OP22 Using Threshold Analysis To Guide Searches For Additional Sources Of Evidence |
| 2023 | partly addresses it | Connecting a disconnected trial network with a new trial: optimizing the estimation of a comparative effect in a network meta-analysis |
Source
Derived from C:L214-215, C:L261 of the reviewed corpus.