Bioequivalence and the 80-125 rule

FDA scientists once pulled the results of 2,070 bioequivalence studies behind oral generic drugs the agency approved from 1996 to 2007 and compared each generic with its brand. The average difference in total drug exposure was 3.56 percent, and the average difference in peak blood level was 4.35 percent, they reported in the Annals of Pharmacotherapy in 2009¹.

Those figures sit awkwardly beside one of the most repeated claims about generics: that FDA lets them contain anywhere from 80 percent to 125 percent of the brand's active ingredient. FDA's Office of Generic Drugs addresses the claim in its training slides for health professionals² under the heading "Dispelling the 80-125% Myth." Its verdict on the idea that a generic can contain 80 percent to 125 percent of the brand's active ingredient: "This is false!"

The 80 to 125 figure is real. It belongs to a statistical test on drug levels in the blood, and how that test works explains why a generic cannot pass while sitting 20 percent below the brand. FDA rewrote its core guidance on the statistics in May 2026, which makes this a good moment to lay out what the rule does and does not allow.

The legal standard

Federal law supplies the definition, in the Food, Drug, and Cosmetic Act. As FDA's Orange Book preface³ summarizes it, a generic is bioequivalent to its reference drug when there is no significant difference in the rate and extent of absorption of the active ingredient, given at the same dose under similar conditions. FDA's regulation at 21 CFR 320.23⁴ adds that the statistical techniques used must be sensitive enough to detect such differences.

Before bioequivalence comes up at all, a generic must be pharmaceutically equivalent to the brand. FDA requires the same active ingredient, strength, dosage form and route of administration⁵. Bioequivalence is the second test, showing that the generic's formulation releases the drug into the body the way the brand's does. GxRadar's guide to how a generic drug gets approved covers the rest of the review.

How a bioequivalence study works

Most studies are crossover trials, usually in healthy adult volunteers. Each person takes the generic in one period and the brand in another, in random order, with a washout gap between them, so every volunteer serves as their own comparison. FDA recommends both fasting and fed studies for oral immediate-release products, according to an Office of Generic Drugs webinar on pharmacokinetic studies⁶, and the May 2026 statistics guidance⁷ says a study should have at least 12 evaluable subjects. The studies in the 2009 review used anywhere from 12 to 170 subjects.

Blood samples are drawn over hours or days. Two measures drive the decision. AUC, the area under the concentration-time curve, captures total exposure to the drug. Cmax, the maximum concentration, captures the peak, which reflects how fast the drug is absorbed.

What the 90 percent confidence interval has to do

For each measure, the analysis works on log-transformed values and calculates the ratio of the generic's average to the brand's (the geometric mean ratio). It then builds a 90 percent confidence interval around that ratio. To pass, the whole interval must sit within 80.00 percent to 125.00 percent, for AUC and for Cmax. The May 2026 guidance is strict on rounding: the rounded interval must be at least 80.00 percent and no more than 125.00 percent.

Those limits look lopsided but are symmetric on the log scale, since 125 percent is the inverse of 80 percent. The 2026 guidance, which replaced FDA's 2001 version⁸, describes the limits as resting on "a clinical judgment that a test product with BA measures outside this range should be denied market access."

Because the entire interval has to fit inside the limits, a generic whose true average exposure was 80 percent of the brand's would produce an interval centered near 80 percent, with roughly half of it below the line. It would fail. Passing takes an average close to 100 percent and variability low enough to keep the interval narrow. The table shows illustrative results.

Look at the third row. A product averaging 10 percent below the brand can fail even though 90 percent sits comfortably inside the 80 to 125 range, because the interval around it has to clear the boundary too.

What the approval record shows

In the 2009 review, the mean generic-to-brand ratio across all 2,070 studies was 1.00 for both AUC and Cmax, with standard deviations of 0.04 and 0.06. In nearly 98 percent of the studies, the generic's AUC differed from the brand's by less than 10 percent, the FDA authors reported¹. Their conclusion was that the criteria support FDA's aim of approving generics that are therapeutically equivalent to their brands.

Those averages include differences in either direction, so they do not show generics running consistently low or high. They show approved generics clustered tightly around the brand. That equivalence is what FDA's AB rating in the Orange Book signals, and it is why a pharmacist can usually substitute a generic such as atorvastatin for its brand under state law.

Tighter and looser limits for some drugs

Some drugs need a different approach, and the 2026 guidance sets out two main exceptions.

• Highly variable drugs. When the brand's own within-subject variability is 30 percent or more, FDA allows a replicate study design and a reference-scaled approach that widens the acceptance range in line with the brand's variability. The point estimate of the ratio must still fall within 80.00 percent to 125.00 percent, and FDA recommends at least 24 subjects.
• Narrow therapeutic index drugs. These are drugs where small differences in dose or blood level can lead to serious therapeutic failure or serious adverse reactions. FDA asks for a fully replicated crossover study and a reference-scaled test that tightens the limits, to 90.00 percent to 111.11 percent at a set level of reference variability, plus the standard 80.00 percent to 125.00 percent test. The generic's within-subject variability is compared with the brand's as well.

FDA's narrow-index method traces to a proposal agency scientists published in Clinical Pharmacology & Therapeutics in 2015⁹, which described a four-way, fully replicated crossover design that compares both the mean and the within-subject variability of the generic and the brand.

When blood levels are the wrong measure

Some products act where they are applied, among them many inhaled, topical and gastrointestinal products. Blood levels say little about how well those work, so FDA may call for in vitro tests, studies with clinical endpoints, or a combination. These are among the complex generics that often have few or no copies. FDA publishes product-specific guidances¹⁰ setting out the evidence it expects for each drug, and developers read them before designing a study (GxRadar has a separate piece on how those guidances work).

Bioequivalence is a population-level standard, and some people notice a difference when the appearance or inactive ingredients of a medicine change. GxRadar's explainer on what actually differs between generic and brand covers those inactive ingredients. Anyone with concerns about a switch, brand to generic or between generics, can raise it with a pharmacist or prescriber. FDA's May 2026 statistics guidance is now the reference point for new applications, and the agency's guidance page⁸ records its history.