CARDIOVASCULAR JOURNAL OF AFRICA • Vol 21, No 2, March/April 2010
70
AFRICA
mezzo but it proved to be useful after all. Critical and clinically
relevant questions were asked about the traditional concept of
average bioequivalence, and this concept emerged strongly. In
particular, the crucial question, namely whether bioequivalence
implied therapeutic equivalence, was answered in the affirma-
tive. In a highly competitive market and litigious society such
as the United States one would expect cases of therapeutic
inequivalence of bioequivalent products to be publicised quickly
and widely. Nevertheless, Rheinstein
15
could state in 1990 that
‘To date, there is no evidence of therapeutic inequivalence in a
properly manufactured generic drug which has been approved as
bioequivalent by the FDA’.
A literature search conducted by Gould
11
10 years later to
answer the question whether average bioequivalence implied
switchability of drug products in practice came to the same
conclusion: essentially no evidence of therapeutic failure of
bioequivalent products could be found. Chow and Liu
2
reported
how in the United States innovator companies file citizen peti-
tions in order to convince the regulatory agency (FDA) that a
generic copy of a brand-name drug will not achieve therapeutic
equivalence even if the generic has been shown to be bioequiva-
lent to the brand name drug. Those authors cite no case of a
generic that has been approved as bioequivalent by the FDA
but has been shown to be therapeutically inequivalent to the
innovator. [However, it should be noted that there are classes of
drugs whose safety are particularly sensitive to the conditions
of administration. For example, responses to immunosuppres-
sants, and the contrasts between different drug products can vary
with time after transplantation, the target organ, ethnicity and
concomitant disease conditions (e.g. diabetes) of the patients.
16
]
In summary, the fundamental bioequivalence assumption,
namely that bioequivalent products are therapeutically equivalent
and can be used interchangeably, has survived strong scrutiny;
scrutiny that was
inter alia
a by-product of the discussion of and
research on the individual bioequivalence concept. The conven-
tional concept of average bioequivalence seems to have served
the consumers of drugs rather well.
Highly variable drugs and widening the
bioequivalence acceptance range, or scaling
What about the concerns that the producers of drugs might
have with the bioequivalence concept? It is well known that
the conventional approach of average bioequivalence, with an
80–125% acceptance range, can make it very difficult to show
bioequivalence for drug products with highly variable bioavail-
ability (so-called ‘highly variable drugs’ or drug products). For
such drugs, sample sizes of 100 subjects and higher can be
required to demonstrate bioequivalence. ‘A feature of the diffi-
culties involving the determination of bioequivalence of highly
variable drugs is that, under typical conditions, a drug product
may not be found bioequivalent to itself.’
17
This is clearly unsat-
isfactory, in particular to producers of drug products who face
inordinate costs when conducting bioequivalence studies for
highly variable drugs.
A potential solution to the problem of highly variable drugs
is suggested by the observation that most highly variable drugs
have a wide therapeutic index. If such a drug indeed has a wide
therapeutic index, it should be clinically acceptable to widen
the bioequivalence acceptance range for it. Various ways of
appropriately widening the acceptance range for highly variable/
wide therapeutic-index drugs have recently been discussed and
investigated.
17,18
The approach that currently seems to be favoured
by FDA scientists and researchers in the field is that of scaled
average bioequivalence.
17,18
Without going into the methodological and statistical details,
the scaled average bioequivalence concept effectively widens
the conventional acceptance range for bioequivalence, namely
80–125%, proportionally to the within-subject standard devia-
tion of the bioavailability of the reference product. Therefore, the
more variable the bioavailability of the reference drug product,
the wider the effective acceptance range for bioequivalence.
Interestingly, the basic concept of scaling the bioequivalence
criterion had already been proposed early in the development
of characteristics for individual bioequivalence,
8,19,20
so that the
scaled average bioequivalence concept can be viewed as another
by-product of the research into individual bioequivalence. While
there still are some problems with the scaled average bioequiva-
lence concept,
17
at present it seems to be the most promising and
practical approach for handling the problem of highly variable
drugs in bioequivalence.
Narrow therapeutic index drugs and narrow-
ing the bioequivalence acceptance range
The mirror image of highly variable drugs with wide therapeutic
index is narrow therapeutic-index drugs whose variability typi-
cally is low. If it is reasonable to widen the bioequivalence accept-
ance range for highly variable drugs with wide therapeutic index,
it seems equally reasonable to narrow the bioequivalence accept-
ance range for drugs with low variability and narrow therapeutic
index. Such a narrowing of the bioequivalence acceptance range
for narrow therapeutic-index drugs could increase assurance,
particularly of the safety of generics in this drug class, without
imposing an undue burden, financial or otherwise, on the spon-
sors of bioequivalence studies for such drugs. Indeed, the new
European bioequivalence guideline
21
envisages that ‘in specific
cases of products with narrow therapeutic index, the acceptance
interval for AUC should be tightened to 90–111.11%’.
Conclusion
The standard approach to bioequivalence assessment, namely
conventional average bioequivalence, has proven itself under
strict scrutiny over more than 20 years. Drug products that under a
proper regulatory regime have been approved as bioequivalent to
a reference product can generally be assumed to be therapeutical-
ly equivalent to that reference product. The bioequivalence limits
could be widened relative to the conventional acceptance range
of 80–125% for handling the problem of highly variable drugs,
and could be narrowed for narrow therapeutic-index drugs. For
highly variable drugs, scaled average bioequivalence provides an
alternative, effective approach to the comparison of drug products.
ROBERT SCHALL,
Department of Mathematical Statistics and Actuarial Science,
University of the Free State, Bloemfontein, South Africa
LASZLO ENDRENYI
Department of Pharmacology, University of Toronto, Ontario,
Canada
