No versioned intervention ontology maps onto CNMA component matrices
Statement
No component ontology is wired into component NMA or component PAIC software. Versioned, machine-readable intervention ontologies built for evidence synthesis do exist and record dose, delivery, and setting: the Behaviour Change Intervention Ontology upper level specifies 42 entities, the Mode of Delivery Ontology gives 65 modes across 15 upper classes with measured inter-rater reliability, and the Intervention Setting Ontology adds 72 entities. Quantitative dose is also no longer always a silent string match, since dose-response NMA models dose continuously and ships on CRAN as netdose. What is missing is a reusable, version-preserving and software-supported mapping from structured intervention ontologies onto CNMA component matrices, and any hierarchical equivalence-class model that lets related component versions borrow strength without being forced to share a parameter. The same vocabulary gap has its sharpest consequence on the comparator side, where placebo, sham, waiting list, no treatment, treatment as usual and standard care are routinely merged into one node and the merge is often what makes the network connect at all. Control-arm content is systematically unreported, so the mapping from arm to node is made from an outcome-blind guess at study content; in a 123-trial psychotherapy network the odds ratio for response between the strongest and weakest control condition is 5.42 (3.35 to 8.77), larger than several of the active contrasts the network exists to estimate, and design-by-treatment and side-splitting tests do not detect it because it is a property of the node’s definition rather than a disagreement between evidence paths.
Why it is open
The bridge rests on the claim that a component means the same thing in both subnetworks, and that claim is currently established by matching treatment-label strings rather than by linking to any of the existing structured standards. Published trial reports rarely record dose intensity, schedule, or delivery fidelity in a machine-comparable way, so even a careful analyst usually cannot verify component equivalence from the available evidence, and once two implementations share a label the model treats their effects as one parameter. Storing the component matrix makes a coding decision reproducible but does not make it justified, and no model exists that lets a version of a component share information with its siblings without being forced equal to them.
What has been tried
| Work | What it contributes |
|---|---|
| Behaviour Change Intervention Ontology, upper level, 2021 | a versioned, machine-readable OWL ontology of intervention components built for evidence synthesis, specifying 42 upper-level entities including intervention content and dose |
| Mode of Delivery Ontology 2021 | supplies the delivery, formulation, and device metadata layer in OWL, with 65 modes across 15 upper classes and tested inter-rater reliability |
| Intervention Setting Ontology 2020 | supplies the context and setting metadata layer, 72 entities structured hierarchically |
| Pedder et al. 2021 | shows dose-response NMA can link disconnected networks by treating dose as quantitative shared structure rather than a treatment label |
| Petropoulou, Rucker and Schwarzer 2026 | introduces frequentist dose-response NMA with linear and nonlinear dose functions |
| netdose | released R package for dose-response NMA, covering the dose and intensity dimension of component version heterogeneity |
| cpaic 0.1.0 (development version, choxos/cpaic@9d150e9) | unreleased development R package on GitHub, not on CRAN; builds the component matrix by tokenizing treatment labels and stores that matrix in the fitted network object, with no structured dose, schedule, delivery, or equivalence metadata |
| Caldwell et al. 2025, Prevention Science | Separates six control types in a school-based anxiety and depression prevention network and reports that the apparent benefit in targeted populations disappears, with waiting list acting as a nocebo relative to no intervention and attention control, concluding that conflating controls in meta-analysis is not appropriate. |
| Pratt et al. 2020, Systematic Reviews | Scoping review of 89 complementary and alternative medicine network meta-analyses that names the merging of sham, placebo and waiting-list controls as requiring consideration and as previously shown to introduce bias into network results, and reports that only 34.8 percent adequately reported their methods for exploring network geometry. |
| Michopoulos et al. 2021, Journal of Clinical Epidemiology | In a 123-trial psychotherapy network for depression, the odds ratio for response between the strongest and weakest control condition is 5.42 (3.35 to 8.77), larger than several of the active contrasts the network exists to estimate, and care as usual had to be excluded because its content could not be pinned down. Concludes that various control conditions should no longer be lumped into one comparison group. |
| Holper and Hengartner 2020, BMC Psychiatry | Treats placebo arms of short-term antidepressant trials as drug-specific nodes rather than one node, which shows the cost of splitting: comparisons disconnect and need an exchangeability bridge, and placebo-versus-placebo contrasts become purely indirect so consistency cannot be checked. |
| Molloy et al. 2018, Health Psychology Review | Primer on network meta-analysis for complex interventions that names control-group content as a considerable threat to transitivity which is difficult to identify because of poor reporting, and places lumping and splitting at the data-extraction step of its own recommended workflow, where planned approaches may be modified in light of the collected data. |
Probable solution or research direction
Map the existing intervention ontologies onto CNMA component matrices rather than building a new ontology, so that dose, schedule, delivery, and setting attributes become the source of the component coding and the equivalence decisions become auditable data that can be rerun under alternative groupings. Use dose-response models where dose is the dimension that varies, since that machinery is released and validated. The genuinely new work is a hierarchical equivalence-class or embedding model in which related component versions borrow strength with a shrinkage parameter rather than being forced to share one.
Verification
Verdict. Overstated. 2 independent auditors judged the claim as stated too strong. No component ontology is wired into component NMA or component PAIC software: existing versioned, machine-readable intervention ontologies (BCIO and its mode-of-delivery and setting modules) record dose, schedule and delivery, but nothing maps them onto CNMA component matrices, and no hierarchical equivalence-class model lets related component versions borrow strength without being forced to share a parameter.
The source titled this. “Component ontology is label-based, not versioned or quantitative” The title says component ontology is not versioned or quantitative, but the audit found three published versioned OWL ontologies with dose and delivery metadata plus continuous dose modeling in netdose.
The source said. “Components in current component NMA and component PAIC are defined by text labels. The same label can hide differences in dose, schedule, formulation, device, adherence, provider skill, and cointerventions, so two studies nominally sharing a component may not share a parameter.”
What was wrong with it. The source said no versioned component ontology with dose and intensity metadata exists and that equivalence decisions are analyst judgment buried in code. The literature auditor found three published, versioned OWL ontologies covering exactly that metadata (BCIO, Mode of Delivery, Intervention Setting); Grok 4.5 added that dose can be modeled continuously via dose-response NMA and the CRAN package netdose; and GPT-5.6 Sol found that choxos/cpaic@9d150e9 stores the chosen C.matrix in the fitted network object, so the coding is recorded and rerunnable even though no ontology backs it.
Dissent. grok dissented (supported-with-caveat): Label-based component coding is still the default in CNMA practice (Veroniki et al. 2026 treat component definition/coding as a critical node-making decision and point to TIDieR for richer trial reporting, not a machine-versioned ontology). Partial prior art does exist on one dimension of versioning: quantitative dose. Pedder et al. 2021 show dose–response MBNMA can link disconnected networks via shared dose-response assumptions; Petropoulou, Rücker, and Schwarzer 2026 introduce frequentist DR-NMA with linear/nonlinear dose functions and ship the CRAN package netdose (2026). Watt et al. 2022 discuss hierarchical dose nodes. Wigle et al. arXiv:2605.15142 (2026) address estimability and ranking hierarchies of components/treatments, not versioned implementation attributes. No versioned component ontology covering dose, schedule, formulation, device, adherence, provider skill, and cointerventions with auditable equivalence classes was found for CNMA/PAIC. Hierarchical embeddings that share strength without forcing equality of versions were not found in CNMA software. Caveat: dose-response methods partially solve quantitative intensity; they do not close the broader ontology claim.
Revised by the full-text reading. Applied full-text reading.
- merged-in
- absorbed a proposal the reading raised separately: No rule for lumping heterogeneous control or comparator conditions 5 citation(s) added. The drafter disagreed with placing it here: The item’s merge target list also names item-2 of this batch, which is not a catalog identifier, so CMP-05 is taken as the external target. Two of the four readers had this accepted as a distinct problem, and there is a real case for that. CMP-05 is scoped to a component vocabulary for component network meta-analysis, whereas this concerns the identity of the anchor node in an ordinary anchored indirect comparison: a lumped anchor breaks the cancellation the anchored design rests on, which is an identification failure rather than a labeling failure. An ontology that recorded control content would also not supply what the readers asked for, namely an admissibility rule for the merge, a diagnostic computable from the assembled data, and a sensitivity analysis that splits the node and reports the cost in connectivity. It is recorded as a merge as directed.
- scope-narrowed
- statement: the closure stands, but its boundary was drawn in the wrong place A 2024 eHealth CNMA protocol (doi:10.1136/bmjopen-2024-084774) maps the BCTTv1 hierarchy into its component scheme by hand, so the absolute claim that no mapping exists was drawn too broadly. An occasional manual study-level mapping is not a reusable one carried into analysis software. Replacement wording reviewer; reviewer confidence high. Previously: “No component ontology is wired into component NMA or component PAIC software. Versioned, machine-readable intervention ontologies built for evidence synthesis do exist and record dose, delivery, and setting: the Behaviour Change Intervention Ontology upper level specifies 42 entities, the Mode of Delivery Ontology gives 65 modes across 15 upper classes with measured inter-rater reliability, and the Intervention Setting Ontology adds 72 entities. Quantitative dose is also no longer always a silent string match, since dose-response NMA models dose continuously and ships on CRAN as netdose. What”
- Flagged for reopening
-
42 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
- partially-solved / overstated. The claim ‘No versioned component ontology with dose and intensity metadata currently exists’ is false as a field-level statement. The Behaviour Change Intervention Ontology, built inside the Wellcome-funded Human Behaviour-Change Project explicitly for evidence synthesis, is a versioned, machine-readable OWL ontology whose upper level specifies 42 entities covering intervention content and delivery; intervention content is decomposed into behaviour change techniques plus dose, and the companion Mode of Delivery Ontology gives 65 modes across 15 upper classes with measured inter-rater reliability (kappa 0.80 for trained annotators), with a further Setting Ontology of 72 entities. The TIDieR reporting checklist independently mandates dose, schedule and delivery reporting (I did not retrieve it, so I give no locator). What is genuinely absent is any linkage of these ontologies to CNMA/CPAIC component coding, and any equivalence-class model that lets related component versions share a prior without being forced equal. The entry has zero prior_work, which is the more serious defect: this is an area with a decade of published ontology engineering. Cites doi:10.12688/wellcomeopenres.15902.2: A versioned, machine-readable ontology of intervention components built for evidence synthesis already exists; it is not yet connected to CNMA.. Cites doi:10.12688/wellcomeopenres.15906.2: Supplies the ‘delivery/formulation/device’ metadata layer the entry says does not exist, in OWL, with tested inter-rater reliability.. Cites doi:10.12688/wellcomeopenres.15904.1: Context/setting metadata layer, same programme; further evidence the ontology infrastructure exists..
- GPT-5.6 Sol, technical and source-code lens
- partially-solved / overstated. C1 is source-confirmed for cpaic, not a mathematical necessity of component NMA: build_C_matrix() splits treatment strings on sep.comps and creates a binary C matrix. C2 is technically correct. cpaic has no structured fields for dose, schedule, formulation, device, adherence, delivery, or hierarchical equivalence, supporting the package-specific versions of C3 and C4; software-wide claims are outside this audit. C5 applies when component coefficients receive a causal interpretation; purely statistical prediction requires the coding to support the assumed common additive surface, not necessarily a separately manipulable causal component. The report overstates that equivalence decisions cannot be recorded or audited: cpaic stores the treatment labels, component names, separator, and exact C.matrix in the network object, and these inputs can be version-controlled or rerun under alternative labels. Distinct versions can already be encoded as distinct component labels, albeit without structured metadata or borrowing. C6 is a reasonable data-governance direction. For C7, embeddings could support graded borrowing, but an embedding alone neither defines causal equivalence nor guarantees valid effect sharing. Cites choxos/cpaic@9d150e9: The exact chosen component mapping is retained in the fitted network object, making the coding reproducible and auditable even though no formal ontology exists..
- Grok 4.5, recency lens
- partially-solved / supported-with-caveat. Label-based component coding is still the default in CNMA practice (Veroniki et al. 2026 treat component definition/coding as a critical node-making decision and point to TIDieR for richer trial reporting, not a machine-versioned ontology). Partial prior art does exist on one dimension of versioning: quantitative dose. Pedder et al. 2021 show dose–response MBNMA can link disconnected networks via shared dose-response assumptions; Petropoulou, Rücker, and Schwarzer 2026 introduce frequentist DR-NMA with linear/nonlinear dose functions and ship the CRAN package netdose (2026). Watt et al. 2022 discuss hierarchical dose nodes. Wigle et al. arXiv:2605.15142 (2026) address estimability and ranking hierarchies of components/treatments, not versioned implementation attributes. No versioned component ontology covering dose, schedule, formulation, device, adherence, provider skill, and cointerventions with auditable equivalence classes was found for CNMA/PAIC. Hierarchical embeddings that share strength without forcing equality of versions were not found in CNMA software. Caveat: dose-response methods partially solve quantitative intensity; they do not close the broader ontology claim. Cites 10.1177/0272989X20983315: Shows dose–response modeling can reconnect or strengthen disconnected networks by treating dose as a quantitative shared structure rather than a pure treatment label.. Cites 10.1186/s12874-025-02754-4: Implements frequentist DR-NMA with multiple dose-response functions; associated CRAN package netdose makes quantitative dose modeling available for multi-intervention networks.. Cites cran/netdose: Released R software for dose-response NMA, addressing the dose/intensity dimension of component/version heterogeneity..
- Decision path
-
R3:partially-solved(n=3) → R8:overstated-majority(n=2)
How the evidence falls in time
42 papers in the reviewed corpus bear on this problem, published 2018 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.
| Year | Says | Paper |
|---|---|---|
| 2018 | confirms it is open | Dismantling cognitive-behaviour therapy for panic disorder: a systematic review and component network meta-analysis |
| 2018 | confirms it is open | An innovative and comprehensive technique to evaluate different measures of medication adherence: The network meta-analysis |
| 2018 | confirms it is open | Network meta-analysis in health psychology and behavioural medicine: a primer |
| 2019 | confirms it is open | Network meta-analysis of multicomponent interventions |
| 2020 | confirms it is open | Supplementary online content to Miklowitz et al., Adjunctive psychotherapy for bipolar disorder: a systematic review and component network meta-analysis |
| 2021 | confirms it is open | Behavioural interventions for smoking cessation: an overview and network meta-analysis |
| 2021 | confirms it is open | Optimal perioperative nutrition therapy for patients undergoing pancreaticoduodenectomy: a systematic review with a component network meta-analysis |
| 2021 | confirms it is open | Supplement 4 (eResults) to Fujii et al., Effect of adjunctive vitamin C, glucocorticoids, and vitamin B1 on longer-term mortality in adults with sepsis or septic shock: a systematic review and a component network meta-analysis |
| 2021 | confirms it is open | Steroid, ascorbic acid, and thiamine in adults with sepsis and septic shock: a systematic review and component network meta-analysis |
| 2022 | confirms it is open | Impact of program characteristics on weight loss in adult behavioral weight management interventions: systematic review and component network meta-analysis |
| 2022 | confirms it is open | Combination of antiplatelet and anticoagulant therapy, component network meta-analysis of randomized controlled trials |
| 2022 | confirms it is open | Psychological interventions for inflammatory bowel disease: a systematic review and component network meta-analysis protocol |
| 2022 | confirms it is open | Meta-Research: Methods and Protocols (Evangelou E, Veroniki AA, eds), Methods in Molecular Biology vol. 2345 |
| 2023 | confirms it is open | Dismantling cognitive-behavioural therapy components for attention-deficit hyperactivity disorder in adolescents and adults: protocol for a network meta-analysis |
| 2023 | confirms it is open | Data visualisation approaches for component network meta-analysis: visualising the data structure |
| 2023 | confirms it is open | Dismantling and personalising task-sharing psychosocial interventions for common mental disorders: a study protocol for an individual participant data component network meta-analysis |
| 2024 | confirms it is open | Lifestyle modification interventions for adults with intellectual disabilities: systematic review and meta-analysis at intervention and component levels |
| 2024 | confirms it is open | Adherence-enhancing interventions for pharmacological and oxygen therapy in patients with COPD: a systematic review and component network meta-analyses |
| 2024 | confirms it is open | Relative efficacy of prehabilitation components on postoperative outcomes: systematic review with network and component network meta-analyses |
| 2024 | confirms it is open | Components and Delivery Formats of Cognitive Behavioral Therapy for Chronic Insomnia in Adults: A Systematic Review and Component Network Meta-Analysis |
| 2024 | confirms it is open | Network meta-analysis examining efficacy of components of cognitive behavioural therapy for insomnia |
| 2024 | confirms it is open | Comparing interventions for early psychosis: a systematic review and component network meta-analysis |
| 2024 | confirms it is open | Pharmacological Treatments in Heart Failure With Mildly Reduced and Preserved Ejection Fraction: Systematic Review and Network Meta-Analysis |
| 2024 | confirms it is open | Comparative effects of behaviour change techniques using eHealth and mHealth in promoting dietary behaviour: protocol for a systematic review and component network meta-analysis |
| 2024 | confirms it is open | A collaborative network trial to evaluate the effectiveness of implementation strategies to maximize adoption of a school-based healthy lunchbox program: a study protocol |
| 2024 | confirms it is open | The effectiveness of delivery modalities of non-pharmacological diabetes prevention programs: A systematic review and component network meta-analysis |
| 2024 | confirms it is open | Effects of non-pharmacological interventions for adults with subjective cognitive decline: a network meta-analysis and component network meta-analysis |
| 2024 | confirms it is open | A Complex Meta-Regression Model to Identify Effective Features of Interventions From Multi-Arm, Multi-Follow-Up Trials |
| 2025 | confirms it is open | Pain science education and exercise interventions for people with knee or hip osteoarthritis: a systematic review, content and meta-analysis |
| 2025 | confirms it is open | Understanding the effectiveness and underlying mechanisms of lifestyle modification interventions in adults with learning disabilities: a mixed-methods systematic review |
| 2025 | confirms it is open | Dismantling behavioural weight management interventions: component network meta-analysis of randomised controlled trials and real-world services |
| 2025 | confirms it is open | Evaluation of the Aspects of Digital Interventions That Successfully Support Weight Loss: Systematic Review With Component Network Meta-Analysis |
| 2025 | confirms it is open | Deconstructing resuscitation training for healthcare providers: a protocol for a component network meta-analysis |
| 2025 | confirms it is open | Pre-endoscopy erythromycin versus metoclopramide for upper gastrointestinal bleeding: a systematic review and network meta-analysis |
| 2025 | confirms it is open | Regional anaesthesia modalities for primary total hip arthroplasty: a systematic review and component network meta-analysis |
| 2025 | confirms it is open | Impact of pulmonary rehabilitation programme design on effectiveness in COPD: a systematic review and component network meta-analysis |
| 2026 | partly addresses it | Mapping components of behavioural weight management interventions using electronic survey and component selection by expert consensus: the BE:COME Study |
| 2026 | confirms it is open | Components of cognitive-behavioral interventions to prevent the onset of depression among the high-risk population: a systematic review and component network meta-analysis |
| 2026 | confirms it is open | Effectiveness of interventions to increase vaccine uptake: component network meta-analysis |
| 2026 | confirms it is open | Components and Delivery Formats of Cognitive Behavioral Therapy for Chronic Fatigue Syndrome/Myalgic Encephalomyelitis: A Systematic Review and Component Network Meta-Analysis |
| 2026 | confirms it is open | Comparative Effectiveness of Inspiratory Muscle Training Approaches Within Intensive Care Unit-Acquired Weakness Prevention and Rehabilitation Strategies: A Systematic Review and Component Network Meta-Analysis |
| 2026 | confirms it is open | Analysing complex interventions using component network meta-analysis |
Source
Derived from C:L1358, C:L214-215, C:L239, scope:field-level (evidence-base construction for component methods) of the reviewed corpus.