Cointervention

Cointervention in Trials and Economic Evaluation

Participants in a randomised controlled trial rarely receive the study treatment and nothing else. Background medicines, rescue drugs and physiotherapy reach patients during follow-up, and each is a cointervention, also written co-intervention. Care used equally in both arms becomes part of the setting in which the new treatment is tested, but care that differs between arms shifts the health gain and the trial costs at the same time, so the ICER inherits the error on both axes. The same question returns in decision models, where the background therapy in the source trials has to match the care patients would receive in the health system making the decision.

What counts as a cointervention

CONSORT 2025 calls the same thing concomitant care and describes it as any additional treatments, interventions or medications received during the trial period in addition to the study interventions, the relevant part being care that could have affected the outcome. The ICH E9(R1) addendum on estimands notes that trials often label such treatments as background treatment, rescue medication or prohibited medication.

The ROBINS-I tool for non-randomised studies narrows the idea to what matters for bias. Sterne and colleagues define relevant co-interventions as those that people might receive with or after starting the intervention of interest and that are both related to the intervention received and prognostic for the outcome. Chapter 25 of the Cochrane Handbook concludes that co-interventions are a type of confounder, listed separately to highlight their importance. The table gives illustrative examples.

When a cointervention biases the treatment effect

RoB 2, the current Cochrane tool for randomised trials, deals with cointerventions in its domain for bias due to deviations from intended interventions, which covers additional interventions inconsistent with the trial protocol. The Handbook states that non-protocol interventions likely to affect the outcome can bias estimated effects.

Whether a cointervention counts as bias depends on the effect being estimated, as set out under performance bias. For the effect of assignment only cointervention that departs from the protocol because of the trial context and affects the outcome counts, while for the effect of adhering the question is whether non-protocol interventions were balanced across groups. Where blinding is impossible, the Handbook notes that open trials can treat patients according to strict criteria that prevent non-protocol interventions.

Bührer and colleagues reviewed 164 cardiovascular drug trials published from 2011 to 2021 in five high-impact journals and found that 124 (74%) did not adequately report cointerventions. Those trials had larger treatment estimates for the primary end point, with a ratio of odds ratios of 1.08 (95% CI 1.01 to 1.15).

Cointerventions as intercurrent events

ICH E9(R1) treats the use of an additional or alternative therapy, including a change to background or concomitant therapy, as an intercurrent event, an event after treatment starts that affects how outcomes are interpreted. Such events also occur in clinical practice, so the addendum treats them as part of the clinical question rather than a drawback to be avoided.

Under the treatment policy strategy, outcomes are used whether or not the patient takes additional medication, and changes in background or concomitant treatments become part of the treatments compared. Under a hypothetical strategy the effect of interest might concern outcomes if the additional medication had not been available, for example where it must be offered for ethical reasons. The Cochrane Handbook regards the effect of assignment as the most appropriate basis for a health policy question, such as whether to prescribe a new drug in a health system, which is the question a reimbursement decision asks. The trial answers it well only when its cointerventions resemble those of routine care.

Costs of cointerventions in trial-based economic evaluation

In a trial-based economic evaluation, Petrou and Gray describe hospital admissions, consultations and the types and quantities of drugs being recorded for each patient over follow-up, with the perspective deciding which categories count. A health system perspective typically includes the intervention, treatment of side effects or complications and follow-up care, and may include care not directly associated with the condition.

The ISPOR task force report on cost-effectiveness analysis alongside clinical trials recommends obtaining resource use directly from participants and conducting the incremental analysis with an intention-to-treat approach. The mean cost difference then contains any difference in cointervention use, so a control arm that seeks extra care raises its own costs and narrows the health gain together.

Protocol-driven costs are different. Petrou and Gray state that protocol driven costs can and should be factored out of cost-effectiveness calculations, so monitoring that only the trial requires comes out, while a rescue drug that patients would also receive in routine care stays in. CONSORT 2025 asks reports to give the number and percentage of participants receiving each relevant type of concomitant care in each arm and, where relevant, the cumulative or average use, which is what an analyst needs to cost cointerventions and check their balance.

Background therapy and comparators in decision models

Most cost-effectiveness models draw relative effects from several trials, each with its own background therapy. The NICE manual for technology appraisal (PMG36) lists additional routine care, alongside patient characteristics, care setting and the year of the study, among the factors that can affect the estimate of relative effectiveness, and asks for such potential effect modifiers to be identified before data analysis. PMG36 also warns that trial results may not generalise because of the relevance of the comparator or of the population, setting and treatment pathway.

In a network meta-analysis the point becomes the transitivity assumption. Chapter 11 of the Cochrane Handbook states that an indirect comparison is not appropriate if the linked trials differ in their effect modifiers, and its example is a common comparator delivered weekly in one set of trials and monthly in another. Background therapy that differs between sets of trials raises the same concern and needs evidence that it does not modify the relative effect.

Background care also enters the costs when a technology extends life. PMG36 allows the committee to consider, alongside the reference case, an analysis with background care costs removed where a technology increases survival in people whose current NHS care is expensive or would not be cost effective at NICE's normal levels. The comparator in a model should carry the cointerventions that routine care supplies, not those that the trial happened to give.

Worked example with unbalanced physiotherapy in an open trial

All figures are illustrative. An open-label trial compares a new drug with standard care over 12 months in a painful musculoskeletal condition without discounting. Physiotherapy is a permitted cointervention at GBP 40 per session, and other care costs GBP 3,000 per patient in each arm. Disappointed by their allocation, control patients average 12 sessions against 5 in the drug arm. The analyst assumes, from external evidence, that the 7 extra sessions add 0.025 QALYs per control patient.

Physiotherapy costs 5 sessions at GBP 40, or GBP 200, in the drug arm and 12 sessions, or GBP 480, in the observed control arm. The observed incremental cost is 4,200 minus 3,480, or GBP 720, and the QALY gain is 0.700 minus 0.670, or 0.030, giving an ICER of GBP 24,000 per QALY gained. With physiotherapy balanced, the incremental cost is GBP 1,000 and the gain is 0.055, giving about GBP 18,182 per QALY gained.

$$ICER_{obs} = \frac{\Delta C - k}{\Delta E - g}$$

where $ICER_{obs}$ is the ICER from the observed trial, $\Delta C$ and $\Delta E$ are the incremental cost and QALYs with balanced cointervention use, $k$ is the extra cointervention cost in the control arm and $g$ is the QALYs that extra care adds there. With $k$ of GBP 280 and $g$ of 0.025, the numerator is GBP 720 and the denominator 0.030, which reproduces GBP 24,000. The observed ICER is the higher one whenever $g / \Delta E$ exceeds $k / \Delta C$, here about 0.45 against 0.28, meaning the cointervention accounts for a larger share of the true health gain than of the true incremental cost.

At an illustrative threshold of GBP 20,000 per QALY, the net monetary benefit is GBP 600 minus GBP 720, or minus GBP 120, on the observed data and GBP 1,100 minus GBP 1,000, or GBP 100, with balanced physiotherapy, so the decision reverses. If the extra sessions arose only because patients knew their allocation, the balanced estimate describes NHS care better; if NHS patients on standard care would routinely receive that physiotherapy, the observed estimate does. Since $g$ is never observed directly, a scenario analysis showing both versions is the honest presentation.

Cointervention and neighbouring concepts

Contamination is the application of one of the compared interventions in participants intended to receive the other, whereas a cointervention lies outside both. Treatment switching replaces the randomised treatment, while a cointervention adds to it, the same line ICH E9(R1) draws between additional treatments that replace and those that supplement. Performance bias is the family name, since the Handbook notes that biases from deviations from intended interventions are sometimes called performance biases, and unbalanced cointervention is one route into it alongside contamination and poor adherence. In non-randomised studies, cointerventions are assessed alongside confounding.

Common misunderstandings about cointerventions

The use of cointerventions does not by itself bias a trial. Balanced, protocol-consistent care is part of the strategies compared, and Chapter 8 of the Cochrane Handbook notes that trial protocols often intend treatment to change in response to participants' health, such as additional care for severe toxicity, so planned additions of this kind are not deviations.

Excluding patients who received a cointervention does not remove the problem. The Handbook calls naive per-protocol and as-treated analyses problematic because prognostic factors may influence whether individuals adhere to their assigned intervention.

Blinding reduces the risk without guaranteeing balance. The Handbook notes that side effects can reveal the allocation in many blinded drug trials, and care for those side effects differs between arms for legitimate reasons, so its costs belong in the analysis.

Sources

• Bührer J, Del Giovane C, Gencer B, et al. Inadequate reporting of cointerventions, other methodological factors, and treatment estimates in cardiovascular trials: a meta-epidemiological study. Mayo Clinic Proceedings: Innovations, Quality and Outcomes. 2023;7(4):231-240. https://doi.org/10.1016/j.mayocpiqo.2023.04.010
• Chaimani A, Caldwell DM, Li T, Higgins JPT, Salanti G. Chapter 11: Undertaking network meta-analyses. In: Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane; 2024. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-11
• Higgins JPT, Savović J, Page MJ, Elbers RG, Sterne JAC. Chapter 8: Assessing risk of bias in a randomized trial. In: Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane; 2024. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-08
• Hopewell S, Chan AW, Collins GS, et al. CONSORT 2025 explanation and elaboration: updated guideline for reporting randomised trials. BMJ. 2025;389:e081124. https://doi.org/10.1136/bmj-2024-081124
• International Council for Harmonisation. ICH E9(R1) addendum on estimands and sensitivity analysis in clinical trials. Step 4, 20 November 2019. https://database.ich.org/sites/default/files/E9-R1_Step4_Guideline_2019_1203.pdf
• National Institute for Health and Care Excellence. NICE technology appraisal and highly specialised technologies guidance: the manual (PMG36). London: NICE; 2022, last updated 13 October 2026. https://www.nice.org.uk/process/pmg36
• Petrou S, Gray A. Economic evaluation alongside randomised controlled trials: design, conduct, analysis, and reporting. BMJ. 2011;342:d1548. https://doi.org/10.1136/bmj.d1548
• Ramsey SD, Willke RJ, Glick H, et al. Cost-effectiveness analysis alongside clinical trials II: an ISPOR Good Research Practices Task Force report. Value in Health. 2015;18(2):161-172. https://doi.org/10.1016/j.jval.2015.02.001
• Sterne JAC, Hernán MA, McAleenan A, Reeves BC, Higgins JPT. Chapter 25: Assessing risk of bias in a non-randomized study. In: Cochrane Handbook for Systematic Reviews of Interventions version 6.5. Cochrane; 2024. https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-25
• Sterne JAC, Hernán MA, Reeves BC, et al. ROBINS-I: a tool for assessing risk of bias in non-randomised studies of interventions. BMJ. 2016;355:i4919. https://doi.org/10.1136/bmj.i4919