pith. sign in

arxiv: 2503.03049 · v3 · submitted 2025-03-04 · 📊 stat.ME · stat.AP

Estimating treatment effects with competing intercurrent events in randomized controlled trials

Pith reviewed 2026-05-23 00:49 UTC · model grok-4.3

classification 📊 stat.ME stat.AP
keywords intercurrent eventscompeting riskstreatment effect estimationrandomized controlled trialsdoubly robust estimationhypothetical strategycomposite variable strategy
0
0 comments X

The pith

A framework defines and estimates treatment effects in trials with competing intercurrent events by combining composite and hypothetical strategies.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Trials often encounter intercurrent events after treatment starts that affect how outcomes are measured or interpreted. The paper separates these into effect-informative events, addressed by assigning a failure outcome, and effect-uninformative events, addressed by assuming a world without them under conditional independence of timing and outcome. Competing events create an added layer because the first one that occurs blocks measurement of later ones. To solve this, the authors build a framework that states the target estimand, shows it can be identified from observed data without parametric assumptions, and derives weighting, outcome regression, and doubly robust estimators. They illustrate the approach on two lupus trials to recover treatment effects under these conditions.

Core claim

We propose a principled framework to formulate the estimand, establish its nonparametric identification and semiparametric estimation theory, and introduce weighting, outcome regression, and doubly robust estimators for treatment effects when competing intercurrent events of both effect-informative and effect-uninformative types are present.

What carries the argument

The estimand formulation that applies the composite variable strategy to effect-informative events and the hypothetical strategy to effect-uninformative events, with the first occurring event censoring all subsequent ones.

If this is right

  • Treatment effects remain identifiable and estimable even when multiple intercurrent events compete.
  • Doubly robust estimators protect against misspecification of either the weighting or outcome models.
  • The framework produces consistent estimates under the stated conditional independence without requiring full parametric models.
  • Application to lupus trial data yields stable results across the three estimator classes.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Trial protocols could specify handling rules for each IE category in advance to make the target estimand explicit.
  • Similar identification arguments may extend to settings with time-varying covariates or survival outcomes.
  • The methods could be adapted to observational data if treatment assignment is balanced conditional on the same covariates used for the hypothetical strategy.

Load-bearing premise

The timing of effect-uninformative intercurrent events is conditionally independent of the outcome given treatment and baseline covariates.

What would settle it

Data in which the occurrence time of an effect-uninformative event remains associated with the outcome after conditioning on treatment and baseline covariates would invalidate the hypothetical strategy.

read the original abstract

The analysis of randomized controlled trials is often complicated by intercurrent events (IEs) -- events that occur after treatment initiation and affect either the interpretation or existence of outcome measurements. Examples include treatment discontinuation or the use of additional medications. In two recent clinical trials for systemic lupus erythematosus with complications of IEs, we classify the IEs into two broad categories: effect-informative (e.g., treatment discontinuation due to adverse events or lack of efficacy) and effect-uninformative (e.g., treatment discontinuation due to external factors such as pandemics or relocation). To define a clinically meaningful estimand, we adopt tailored strategies for each category of IEs. For effect-informative IEs, which are often informative about a patient's outcome, we use the composite variable strategy that assigns an outcome value indicative of treatment failure. For effect-uninformative IEs, we apply the hypothetical strategy, assuming their timing is conditionally independent of the outcome given treatment and baseline covariates, and hypothesizing a scenario in which such events do not occur. A central yet previously overlooked challenge is the presence of competing IEs, where the first IE censors all subsequent ones. Despite its ubiquity in practice, this issue has not been explicitly recognized or addressed in previous data analyses due to the lack of rigorous statistical methodology. In this paper, we propose a principled framework to formulate the estimand, establish its nonparametric identification and semiparametric estimation theory, and introduce weighting, outcome regression, and doubly robust estimators. We apply our methods to analyze the two systemic lupus erythematosus trials, demonstrating the robustness and practical utility of the proposed framework.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 3 minor

Summary. The paper proposes a framework for defining treatment effect estimands in RCTs with competing intercurrent events (IEs), classifying IEs as effect-informative (handled by composite variable strategy assigning failure) or effect-uninformative (handled by hypothetical strategy). It claims nonparametric identification of the resulting estimand (even with competing IEs where the first censors later ones), develops weighting, outcome regression, and doubly robust estimators under semiparametric theory, and applies the methods to two systemic lupus erythematosus trials.

Significance. If the identification and estimation results hold, the work fills a gap by rigorously addressing competing IEs, which are common but previously lacked explicit methodology. Strengths include the explicit formulation of a mixed-strategy estimand, development of doubly robust estimators, and real-data application demonstrating practical utility.

major comments (2)
  1. [Identification section (around the hypothetical strategy paragraph)] The nonparametric identification result for the hypothetical strategy on effect-uninformative IEs (central to the mixed estimand) rests on the assumption that IE timing is conditionally independent of the potential outcome given treatment and baseline covariates. With competing IEs, this must hold for the observed first-event time; the paper should provide the explicit identification formula (likely in the identification section) and discuss whether the competing structure strengthens or weakens the assumption.
  2. [Estimation theory / DR estimator derivation] The doubly robust estimator inherits the conditional independence assumption without correction for its violation. Section on estimation theory should clarify whether the influence function or DR property remains valid under competing IEs, or if additional robustness is claimed.
minor comments (3)
  1. [Estimand definition] Clarify notation for the composite outcome under effect-informative IEs and how it interacts with the hypothetical counterfactual for effect-uninformative IEs in the presence of multiple possible first events.
  2. [Application to SLE trials] The application section would benefit from explicit reporting of estimated effects with and without the competing-IE adjustment to illustrate the practical impact.
  3. [Methods or application] Add a brief discussion of how baseline covariates are selected or if sensitivity to covariate choice is explored.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback. We address the major comments point-by-point below, and will incorporate clarifications in the revised manuscript.

read point-by-point responses
  1. Referee: [Identification section (around the hypothetical strategy paragraph)] The nonparametric identification result for the hypothetical strategy on effect-uninformative IEs (central to the mixed estimand) rests on the assumption that IE timing is conditionally independent of the potential outcome given treatment and baseline covariates. With competing IEs, this must hold for the observed first-event time; the paper should provide the explicit identification formula (likely in the identification section) and discuss whether the competing structure strengthens or weakens the assumption.

    Authors: We agree that explicit clarification is warranted. The conditional independence assumption is posited for the timing of the observed first IE, as subsequent IEs are censored by the first. The competing structure does not alter the core assumption but requires it to apply to the minimum observed time, which is consistent with standard competing risks frameworks. We will add the explicit identification formula for the mixed estimand under competing IEs in the identification section and include a discussion noting that the assumption is neither strengthened nor weakened but must be interpreted in the context of the first observed event. revision: yes

  2. Referee: [Estimation theory / DR estimator derivation] The doubly robust estimator inherits the conditional independence assumption without correction for its violation. Section on estimation theory should clarify whether the influence function or DR property remains valid under competing IEs, or if additional robustness is claimed.

    Authors: The influence function and doubly robust property are derived under the semiparametric model that includes the conditional independence assumption as part of the data-generating process. The DR property holds if either the outcome regression or the propensity/weighting model is correctly specified, conditional on this assumption. We do not claim robustness to violations of the independence assumption. We will revise the estimation theory section to explicitly state that the validity of the influence function and DR property relies on the maintained assumptions, including conditional independence for the first IE time. revision: yes

Circularity Check

0 steps flagged

No circularity; new identification and estimation theory is self-contained

full rationale

The paper defines a novel estimand that combines the composite-variable strategy for effect-informative IEs with the hypothetical strategy for effect-uninformative IEs under competing risks, then derives its nonparametric identification from explicit conditional-independence assumptions and constructs weighting, outcome-regression, and doubly-robust estimators. These steps introduce original semiparametric theory for a setting the abstract states had not been rigorously addressed; no equation reduces by construction to a fitted parameter, renamed known result, or load-bearing self-citation. The derivation therefore stands on its stated assumptions and new technical development rather than on circular reuse of its own inputs.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The framework rests on one key domain assumption for the hypothetical strategy; no free parameters or invented entities are introduced in the abstract.

axioms (1)
  • domain assumption The timing of effect-uninformative intercurrent events is conditionally independent of the outcome given treatment and baseline covariates.
    Invoked to justify the hypothetical strategy that assumes such events do not occur.

pith-pipeline@v0.9.0 · 5837 in / 1109 out tokens · 26336 ms · 2026-05-23T00:49:56.718940+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.