`
`as
`
`Table 1. An Example of the Test V. Reference Results of a
`Highly V.~.ble Drug
`
`Number of subjects
`Tesl
`Reference
`Ratio of means (tesVreference)
`90% Confidence interval*
`Lower limit
`Upper limit
`
`AUC
`
`36
`224 ± 146
`212 ± 129
`1.05
`
`0.72
`1.07
`
`C_
`
`36
`80 ± 62
`79 ± 48
`1.01
`
`0.87
`1.21
`
`Note: A highly variable drug is defined as having a coefficient of variation
`>30%. Here fNs are >60%.
`*The 0.80-1.25 cr1ter1a Is applied to the 90% confidence Interval, not the ratio
`of means.
`
`-20%1+25% rule is only satisfied by statistical criteria
`which show that the 90% confidence interval around the
`ratio of measured parameters will fall within the accepted
`0.8 to 1.25 range. That is, in practice, two one-sided
`statistical tests are carried out using log-transformed data
`from the bioequivalence study to show that the 90%
`confidence interval for the ratio of the generic to the
`innovator for Aue and the ratio for c..~ are within the
`limits of 0.8 and 1.25 (ie, -20%1+25%). This requirement
`that the 90% confidence interval for the generic product
`relative to the innovator fall within -20%1+25% is very
`diJIerent than the implied criteria that the ratio of the
`average values fall within this interval.
`For a highly variable drug (coefficients of variation
`greater than 30%), even when the ratio of means (test!
`reference) is very close to 1.0, the product can still fail the
`FDA bioequivalence requirements (Table 1). In the exam(cid:173)
`ple shown in Table 1, the ratio of means for AUe was 1.05
`and the ratio of means for ~ was 1.01. However, the test
`product is not bioequivalent to the reference product since
`the lower limit of the 90% confidence interval for AUe (ie,
`0.72) is outside of the 0.8 to 1.25 acceptable range. Note
`that even for ~ where the ratio is 1.01, the upper limit of
`the 90% confidence interval for c..~ (ie, 1.21) approaches
`the maximum acceptable 1.25 value. Thus, cliuicians shonld
`have a much greater confidence in the equivalence of
`approved generic products when they realize that the
`statistical criteria are based on the 90% confidence interval,
`not the means for AUe and c,.,...
`In 1987, after only 2 years of experience with the new law
`for approval of generic drugs, Nightingale and Morrison'
`reviewed the data for 224 generic drug products approved
`up to that time. They noted that the mean diJIerence in
`AUe between approved generic products vs the innovator
`product was 3.5% and that the percentage of products
`within ±5% of innovator AUe was about 80%. Only 1 of
`224 approved products had an AUe greater than 15%
`diJIerent from the innovator. It shonld be noted that the
`products reviewed by Nightingale and Morrison were ac(cid:173)
`cepted on the basis of a diJIerent statistical criteria than the
`90% confidence interval now required. Most likely the
`product with a 15% diJIerence in AUe wonld not pass the
`
`BENET
`
`Table 2. Bioequivalenca Studies for New Molecular Entities
`Approved by the FDA From January 1, 1981 through
`December 31, 1990
`
`Total approved
`For oral administration
`Bioavailability requirements waived (no
`or minimal absorption)
`Unavailable or insulTicient data to judge
`Final marketed formulation same as
`clinical trial
`Final marketed formulation di1rerent from
`that in clinical trial tror 50 new
`molecular entities bioequivalence
`tested in vivo)
`
`220
`97
`3(3.3%)'
`
`7
`34 (37.8%)
`
`53 (56.9%)
`
`"Percentage falling into each category where sull'icient data are available.
`(Data from Benet and Goyan.~
`
`present more restrictive statistical requirements. It has been
`reported that the FDA is currently reanalyzing over 2000
`generic drug products that have been approved since the
`1987 evaluation. Although the resnlts are not final, the early
`calcnlations suggest that diJIerences between approved ge(cid:173)
`neric products and innovator products will be minimal,
`certainly on average less than 5%.
`One of the major issues often raised by those opposing a
`particnlar generic product is that the studies required for
`bioequivalence do not include clinical tests within the
`intended patient population. Many clinicians and patients
`wrongly presume that new drug products coming to the
`market have been evaluated in patient popnlations. Benet
`and Goyan4 have pointed out that the community of
`concerned health professionals is apparently unaware that
`the majority of new oral drug products on the market have
`been approved based on a bioequivalence study in healthy
`volunteers. That is, the dosage form that is finally approved
`by the regulatory agencies is usually not the one used in
`clinical efficacy studies. In such cases, the innovator is
`required only to prove to the regulatory agencies that the
`final marketed form is bioequivalent to that formnlation
`used in the efficacy studies. In November 1990, in testimony
`before the Edward's Committee recommending changes in
`the FDA, 6 Benet pointed out that approval of the majority
`of new orally administered drug products was based on the
`same bioequivalency criteria that are used for the approval
`of generic drugs. At his request, the FDA reviewed the new
`molecnlar entities approved during the decade of the 1980s
`with the resnlts presented in Table 2. As suspected, only
`38% of those products approved for oral administration
`used the original market formnlation in the pivotal clinical
`trials. In contrast, 59% used diJIerent experimental formu(cid:173)
`lations, and therefore were required to prove bioequiva(cid:173)
`lence of the formnlation to be marketed with that used in
`the trials. Thus, for the majority of new drugs administered
`orally during the past decade, clinical studies were not
`carried out on the marketed formulation, just as is the case
`with generic drugs.
`Recently, there has been increased interest in bioequiva-
`
`p. 2
`
`
`
`UNDERSTANDING BIOEQUIVALENCE TESTING
`
`9S
`
`lency issues with the contention being raised that the
`present requirements are not adequate for narrow thera(cid:173)
`peutic index (NTI) drugs. Officials at the FDA have vigor(cid:173)
`ously denied that the present regulations are not appropri(cid:173)
`ate. However, clinicians may hear a great deal of discussion
`concerning "population" vs "individual" bioequivalence.
`These terms are translated into issues related to "prescrib(cid:173)
`ability" and "switchability." It has been proposed that the
`present bioequivalence criteria based on a single dose
`crossover of the test and reference products will assure only
`that a particular generic product will be equivalent to the
`innovator for the population as a whole, that is, it is
`"prescribable." In contrast, of greater importance is the
`assurance of bioequivalence in a patient already stabilized
`on the innovator product when the patient is switched from
`the innovator drug product to the generic. Theoretical
`proposals requiring replicate design studies, ie, studies
`where each subject in a bioequivalence study receives the
`generic formulation twice and the innovator formulation
`twice, thereby providing measures of intrasubject variabil(cid:173)
`ity, are needed to assure switchability. In my opinion,
`individual bioequivalence is a promising, clinically relevant
`method which should theoretically provide further confi(cid:173)
`dence to clinicians and patients that generic drug products
`are, indeed, equivalent in an individual patient. However, as
`of this time, little prospective data exist to validate this
`theoretical approach and to provide confidence to the
`scientific and clinical commuuity that the methodology
`required and the expense entailed would be justified.
`Currently, individual bioequivalence is a theoretical solu(cid:173)
`tion to solve a theoretical clinical problem. There is no
`evidence of a clinical problem, either a safety or an efficacy
`issue. Furthermore, there is no evidence that individual
`
`bioequivalence would solve the problem if it did exist. What
`is needed is the generation of a large database which would
`provide the FDA and drug company scientists with the
`necessary information to make a reasoned consensus judge(cid:173)
`ment as to the appropriate criteria for individual bioequiva(cid:173)
`lence.
`In conclusion, this report has provided background infor(cid:173)
`mation on how bioequivalence studies are carried out at
`present, the limits of differences allowed for acceptable
`products, and some history concerning actual differences
`between generic and innovator products as provided by the
`FDA. Additionally, it has been shown that innovator man(cid:173)
`ufacturers often use the same test required for approval of
`a generic product to receive market approval for the
`innovator product. Finally, the controversial issues relating
`to individual bioequivalence were reviewed, particularly
`with reference to NTI drugs, and as have been raised with
`respect to generic cycIosporine. The following presentations
`in this symposium address specific issues related to inunu(cid:173)
`nosuppressive drugs.
`
`REFERENCES
`1. Federal Register: 42:1648, 1977
`2. Approved Drug Products with Therapeutic Equivalence Eval(cid:173)
`uations. 17th Ed. DHHS/Food and Drug Administration, Center
`for Drug Evaluation and Research, Rockville, MD, 1997
`3. Sheiner LB: Slat Med l1:1m, 1992
`4. Benet lZ, Goyan JE: Pharmacotherapy 15:433, 1995
`5. Nightingale SL, Morrison JC: JAmMed Assoc 258:1200,
`1987
`6. U.S. Department of Health and Human Services, Advisory
`Committee of the Food and Drug Administration. Final report.
`Washington, DC, 1991
`
`p. 3
`
`