Assigns an evidence grade (Beleg/Hinweis/Anhaltspunkt/Kein Beleg) based on study quality, risk of bias, effect size, precision, heterogeneity, and consistency across studies. This implements the IQWiG evidence grading system as described in IQWiG Allgemeine Methoden Version 3.1.4, p.51-56.
Usage
grade_evidence(
meta_result,
rob_results = NULL,
n_studies_override = NULL,
direction = "none",
ci_level = 0.95,
publication_bias = NULL,
indirectness = NULL,
metadata = list()
)Arguments
- meta_result
A MetaResult object from meta_analysis() containing pooled effect estimate, confidence interval, p-value, and heterogeneity statistics. Alternatively, a ComparisonResult object for single-study assessments.
- rob_results
A list of RoB2Result objects (one per study), or a single RoB2Result for single-study assessment. If NULL, defaults to "Low" risk of bias for all studies.
- n_studies_override
Integer. Override the number of studies for single study assessment. Default: NULL (use meta_result@n or 1 for ComparisonResult).
- direction
Character. Direction of effect: "benefit", "harm", or "none". Default: "none" (effect direction determined from effect estimate).
- ci_level
Numeric. Confidence level for significance assessment. Default: 0.95.
- publication_bias
List or NULL. Publication bias assessment from trim_and_fill() or egger's_test(). If provided, used in grading. Expected list elements: p_value, adjusted_estimates.
- indirectness
Numeric or NULL. Indirectness score (0-1, where 1 = no indirectness). Default: NULL (assumed 1 = no indirectness).
- metadata
List. Additional metadata for the EvidenceGrade object. Default: empty list.
Examples
if (FALSE) { # \dontrun{
# Meta-analysis with 5 RCTs, all low RoB
meta_res <- MetaResult(
estimate = 0.72,
ci = c(0.62, 0.84),
p_value = 0.0001,
n = 5L,
effect_measure = "hr",
heterogeneity = list(I2 = 25, tau2 = 0.01),
method = "REML meta-analysis"
)
# Create RoB results for all studies (all low)
rob_results <- list(
assess_rob2("Study1", "Low", "Low", "Low", "Low", "Low"),
assess_rob2("Study2", "Low", "Low", "Low", "Low", "Low"),
assess_rob2("Study3", "Low", "Low", "Low", "Low", "Low"),
assess_rob2("Study4", "Low", "Low", "Low", "Low", "Low"),
assess_rob2("Study5", "Low", "Low", "Low", "Low", "Low")
)
# Grade the evidence
grade <- grade_evidence(meta_res, rob_results, direction = "benefit")
grade@grade
grade@grade_de
# Single study assessment
comp_res <- ComparisonResult(
estimate = 0.75,
ci = c(0.60, 0.93),
p_value = 0.008,
effect_measure = "hr"
)
single_grade <- grade_evidence(
comp_res,
rob_results = assess_rob2("Study1", "Low", "Low", "Low", "Low", "Low"),
n_studies_override = 1L
)
} # }