E-drug: Re: Antibiotic prescribing (1)
----------------------------------------------------------
The moderator's comment captures some of the key points. With respect to
the concept of "therapeutic coverage", which I defined back in 1991 in a
poster for a joint meeting of the Society for Clinical Trials and the Int'l
Soc for Clin Biostat in Brussels(ref at end), the key variable is the
duration of the drug's therapeutic action, not the duration of drug
concentration in plasma being maintained above some arbitrary level. Of
course drug action is much more difficult to measure in many instances than
drug concentration in plasma, but there are numerous instances in which drug
action persists long after drug concentrations in plasma have fallen to
zero. Penicillin provided the first example, giving rise to the term
"post-antibiotic effect", defined by the late Harry Eagle, one of the
pioneers of research on the actions of penicillin in animal models of
infectious diseases. The term referred to the persistence of penicillin's
cidal effect on various micro-organisms for appreciable periods of
time after all measurable penicillin had vanished from plasma. Of course
the methods for analyzing drug concentrations in plasma were very crude at
that time (the late 1940's), but penicillin's rapid excretion via renal
tubular secretion meant that it was gone within several hours of its
administration. In any case, Eagle's term stuck, but is rarely used today.
A modern example of stark discrepancy between drug concentration in plasma
and drug action is provided by the largest selling product in the history of
the pharmaceutical industry, omeprazole, which has a plasma half-life of ca.
30 min and a duration of action of 3-5 days. It has a once-daily dosing
regimen, which could be said to be excessive, but perhaps not if you take
into account the fact that omeprazole use has a success rate in excess of
90%, which is
probably contributed to in substantial measure by its exceptionally high
degree of forgiveness (post-dose duration of action minus recommended
inter-dose interval), making its acid-suppressing action largely independent
of all but the most egregious errors in compliance.
A pertinent example of radical down-sizing in anti-infective drug regimens
is provided by the regimen currently in use in directly-observed therapy for
tuberculosis by the New York City Public Health Department. That is
described in the following paper:
Weis SE, Slocum PC, Blais FX, King B, Nunn M, Matney GB, Gomez E,
Foresman BH. The effect of directly observed therapy on the rates of drug
resistance and relapse in tuberculosis. N Engl J Med 330: 1179-84, 1994
There you see the superior effectiveness of drugs administered just 4 times
a week in the DOT program. These are drugs that originally were specified
for administration 2-3 times a day in self-managed care. The DOT program's
findings indicate that the original regimens were substantially
over-specified. It may be that the regimens used in DOT are also suboptimal
and could be improved upon, for the decision on the regimen was made out of
urgent necessity in the face of rising prevalence of multi-drug resistant
TB, not on the basis of experiment.
Choosing a drug regimen is a difficult task, made all the more so by strong
pressures to pick "the" regimen early so that clinical development can begin
using the regimen intended for recommendation in product labeling. There is
also a strongly felt need to give enough to be sure that efficacy is
demonstrable. Both factors predispose to excess, and the documented
examples of later-revised regimens run strongly toward reduction in dose
and/or frequency rather than the reverse. (Ibuprofen is a conspicuous
exception.) Bob Temple wrote a nice review of overestimated cardiovascular
drug regimens a decade ago:
Temple R. Dose-response and registration of new drugs. In: Dose-
response relationships in clinical pharmacology. Eds: Lasagna L, Erill S,
Naranjo CA. Amsterdam: Elsevier, 1989, pp 145-167.
The experiment you have run is a very revealing one, which takes advantage
of what Don Rubin has called the "natural experiment in dose ranging
imbedded in every randomized controlled trial", created by patients'
variable compliance with the prescribed drug regimen. Having seen how far
patients can stray, with errors of omission, from the recommended regimen
and still end up with a good outcome of treatment, you can then reconsider
the question of what might be considered the optimal regimen. "Optimal" is
not necessarily the regimen with the most slender margin between
not-quite-enough and just-enough, because of the prevalence of partial
compliance and delayed/omitted doses. Drug holidays -- multiday lapses in
dosing -- afford an opportunity in anti-microbial treatment regimens for the
emergence of drug resistant micro-organisms, the rising occurrence of which
was one of the main stimuli for the DOT program in NY City. It has also
been an issue in AIDS treatment with the protease inhibitors, where there is
a body of observational evidence that multiday lapses in dosing (drug
holidays) are associated with emergent drug-resistance.
One additional point to note is that the regimen for erythromycin
administration was set long before chlamydia trachomatis was defined as a
pathogen in ocular and genital disease. An optimal regimen for C.
trachomatis may be suboptimal for S. aureus, N. gonorrheae, T. pallidum, or
others.
But setting those perhaps subtleties aside, it is an interesting judgment
call as to how much of what engineers call "redundancy" one would build into
a regimen in order to forgive the more common lapses in dosing. In the
historical dichotomy between "roundheads" and "cavaliers", to use the
parlance of the Cromwellian era in England, roundheads will opt for little
forgiveness,
and cavaliers for lots.
An interesting case study is provided by the contraceptive steroid,
norgestrel. As a once-daily oral product, nicknamed the "minipill",
norgestrel has a post-dose duration of action reckoned to be 27 hrs, so it
has a forgiveness of only three hours. (The combined progestin-estrogen
contraceptives in widest use today have a forgiveness of 12-24 hours,
depending on whether you read the UK or the US labeling.) Not surprisingly,
the minipill has the poorest contraceptive
efficacy in trials of any of the steroidal contraceptives. The same agent,
formulated as a 5-yr implant (NORPLANT), has the best contraceptive efficacy
of any of the steroidal contraceptives. The implant's automatic delivery
removes virtually all issues of compliance, allowing one to see the drug's
potential utility when administered uniformly. There are, of course, a lot
of other
issues related to the acceptance of NORPLANT besides efficacy, but looking
solely at efficacy one sees in the contrast between the once-daily oral form
and the once-a-half-decade implant form the way in which erratic compliance
can undermine efficacy of an unforgiving pharmaceutical. This and related
points, plus the reference to the short duration of action of norgestrel,
are discussed in a paper I wrote about a year ago:
Urquhart J. Can drug delivery systems deliver value in the new
pharmaceutical marketplace? Brit J Clin Pharmacol 44: 413-419, 1997.
In summary, one can say that faith in plasma levels as the guide to regimen
optimization is often misplaced, and that regimens set long ago were/are
often set at doses that are too high, taken more often than strictly
necessary. The use of the natural experiment in dose ranging created by
spontaneous variations in compliance, when monitored with reliable means
(i.e. with electronic monitoring, not pill counts because they grossly
exaggerate compliance), can be a way to estimate the lower bounds of drug
exposure consistent with good outcome. The question of how much forgiveness
is enough is a matter on which people are bound to disagree, but it comes
down to playing the odds of overdose-related toxicity and its consequences
vs the odds of various patterns of noncompliant dosing and their
consequences vs the odds of therapeutic failure and its consequences. With
on-patent drugs, there is also the cost issue, for the ability to pare down
the dosing regimen usually means a reduction in treatment cost, but if you
go too far and end up with a regimen that only a minority of exceptionally
punctual compliers can execute, then there will be many poor outcomes, with
their associated costs.
Here's the reference for therapeutic coverage:
Urquhart J. Therapeutic coverage: a parameter for analyzing the
pharmacodynamic impact of partial patient compliance. Program and
Abstracts, Society for Clinical Trials/International Society for Clinical
Biostatistics, Joint Meeting, Brussels, 1991, P12.
I can send you the abstract for this if you'd like. The algorithm for
computing it is included in the CSS software of AARDEX Ltd., which is the
leading supplier of electronic monitors and software for analysis of
compliance data. (www.scopus.ch/aardex)
I hope this is helpful.
John U.