Combined-population treatment effect

Evaluates the CAMS combined-population treatment effect at the pooled information fraction from ifrac(). This is the effect corresponding to the combined population at its contribution-weighted subgroup composition \(\widetilde{\pi}\), using the linear predictor \(\alpha + \widetilde{\pi}\,(\beta + \gamma)\) (binary case; the implementation sums all non-reference IF and subgroup coefficients).

Usage

combined(
  fit,
  at = NULL,
  ifrac = NULL,
  type = c("Overall", "Prediction"),
  prob = NULL,
  max_support = NULL,
  summary = TRUE,
  contrast_sep = NULL,
  exponentiate = NULL,
  shrinkage = FALSE,
  contribution
)

Arguments

fit

A interactmeta from cams().

at

Combined-population information fraction \(\pi\); see modifiers().

ifrac

When at is NULL, passed to ifrac(): "interaction" (default) or "overall" for overall-block pooling of trial information fractions.

type

One or both of "Overall" and "Prediction" (new-trial random effect; see modifiers()).

prob, max_support, summary, contrast_sep, exponentiate, shrinkage, contribution

As in modifiers().

Value

An interactmeta_combined_effect object (list with draws, summary, overall_ifrac, overall_ifrac_source, ifrac, reference, prob). Draws are indexed by type. summary includes estimate (median), mean, sd, lci, and uci.

Details

When subgroup variances scale with sample size (or events) and allocation is similar, \(\pi\) is close to the trial's subgroup share, so this is typically a trial-mixture effect. It is still not automatically a target-population average: the pooled \(\widetilde{\pi}\) is precision-weighted across trials, and on log HR / log OR a mixture of subgroup effects is not collapsible. Trial-specific combined-population effects are the transformed quantities \(m_j = (1-\pi_j)y_{Aj}+\pi_j y_{Bj}\).

See also

modifiers() for level-specific \(\mu_A(\pi)\) and \(\mu_B(\pi)\) at the same reporting IF; interactions(), ifrac()