90 per cent confidence interval for the geometric mean ratio

The mean difference d between test and reference in natural logs of AUC or Cmax is exponentiated to give the geometric mean ratio, and the limits of the 90% interval are d plus or minus the one-sided 95% t quantile times the standard error, exponentiated. This interval corresponds to the two one-sided tests procedure at the 5% level. Average bioequivalence is concluded when the interval, expressed in per cent and rounded to two decimal places, is at least 80.00 and no more than 125.00.

Signature

GMR = exp(d); GMR_L = exp(d - t_crit * SE); GMR_U = exp(d + t_crit * SE)
Inputs
InputsDefinitionUnit
dEstimated difference in means of the natural log of AUC or Cmax, test minus reference, from the study modelnatural log units
t_crit95th percentile of the t distribution with the residual degrees of freedom of the model, n minus 2 for a two-period crossoverdimensionless
SEStandard error of d from the study modelnatural log units
Output
GMRPoint estimate of the ratio of geometric means of the pharmacokinetic measureratio
GMR_LLower limit of the 90% confidence interval for the ratioratio (multiply by 100 for per cent)
GMR_UUpper limit of the 90% confidence interval for the ratioratio (multiply by 100 for per cent)

Function

Average bioequivalence and generic price function

For an ANDA, bioequivalence to the reference listed drug is shown when the 90% confidence interval for the ratio of geometric means of a pharmacokinetic measure, such as AUC or Cmax, lies within 80.00% to 125.00% after rounding. The interval is computed on the log scale and exponentiated. For economic models, the approval of generic competitors is then translated into a post-entry price.

Implementations

  • Excel

    90 per cent interval for the ratio

    With the mean log difference in MeanLogDiff, its standard error in SELogDiff and the residual degrees of freedom in DF, the formulas return the ratio and its 90% limits.

    =EXP(MeanLogDiff); =EXP(MeanLogDiff-T.INV(0.95,DF)*SELogDiff); =EXP(MeanLogDiff+T.INV(0.95,DF)*SELogDiff)

Assumptions

  • Log-normal pharmacokinetic measures

    AUC and Cmax are analysed after log transformation, as the FDA guidance recommends; normality of the residuals is not tested and is not a reason to analyse on the original scale.

  • Equal-tails interval and two one-sided tests

    Each limit of the 90% interval is by itself a 95% one-sided bound, so the interval test is equivalent to two one-sided tests at the 5% level.

  • Standard limits for most drugs

    The 80.00% to 125.00% limits apply for a broad range of drugs. Narrow therapeutic index drugs, highly variable drugs and some product-specific guidances use scaled or tighter criteria.

Worked examples

  • Generic that meets the limits

    A two-period crossover in 24 subjects gives a mean log difference of minus 0.051293, a ratio of 0.95, with a standard error of 0.06. With 22 degrees of freedom the t quantile is 1.717144, so the 90% interval is 85.70% to 105.31%, inside the limits. The figures are illustrative.

    d = -0.051293; SE = 0.06; t_crit = 1.717144; GMR = 0.9500; GMR_L = 0.8570; GMR_U = 1.0531
  • Generic that fails the lower limit

    A ratio of 0.88, a mean log difference of minus 0.127833, with a standard error of 0.07 gives an interval of 78.03% to 99.24%. The point estimate is inside the limits but the lower confidence limit is not, so bioequivalence is not shown.

    d = -0.127833; SE = 0.07; t_crit = 1.717144; GMR = 0.8800; GMR_L = 0.7803; GMR_U = 0.9924

Common errors

  • Testing the point estimate instead of the interval

    A ratio of 0.88 lies within 0.80 to 1.25, but its lower confidence limit of 78.03% does not. The criterion applies to the whole 90% interval.

  • Reading the limits as allowing 20 per cent less drug

    The limits apply to the confidence interval for the ratio of exposure measures, not to the drug content of the dose. A product whose true ratio sat near 0.80 would usually fail, because its interval would extend below the limit.

  • Using a 95 per cent two-sided interval

    A 95% two-sided interval corresponds to one-sided tests at 2.5% and is wider than the interval the guidance specifies for average bioequivalence of AUC and Cmax.

Sources

  • FDA statistical guidance on average bioequivalence

    US Food and Drug Administration, Center for Drug Evaluation and Research. Statistical Approaches to Establishing Bioequivalence: Guidance for Industry. Silver Spring, MD: FDA; May 2026. Section I.B (average bioequivalence: 90% confidence interval for the ratio of population geometric means within a limit usually 80.00% to 125.00%; footnote 6, rounded interval at least 80.00% and not more than 125.00%), section II.C.1 (log transformation of AUC and Cmax), section II.D.1.a (two one-sided tests and equal-tails intervals) and Appendix B (rationale for log transformation).

    View source →

  • Regulatory definition of bioequivalence

    Code of Federal Regulations, Title 21, section 314.3. Definitions (bioequivalence). Accessed 29 September 2026. Bioequivalence: the absence of a significant difference in the rate and extent to which the active ingredient becomes available at the site of drug action when administered at the same molar dose under similar conditions in an appropriately designed study.

    View source →

Canonical Identity

Stable URI · Machine-readable · Resolvable · CC BY 4.0