E-drug: RFI: Scientific basis for defining non-adherence (2)
------------------------------------------------------------
[Thanks John for this very full response. Does anyone want to add to or
challenge these ideas? TD]
In reply to Andy Gray's inquiry about definitions of non-adherence, I
can suggest the following publications that address this problem:
Urquhart J. Ascertaining how much compliance is enough with
Outpatient antibiotic regimens. Postgrad Med J 68 (Suppl 3), S49-
S59, 1993.
Urquhart J. Variable patient compliance in ambulatory drug trials:
nuisance, threat, opportunity. J Antimicrobial Chemotherapy 32:
643-9, 1993.
Urquhart J. The electronic medication event monitor - lessons for
pharmacotherapy. Clin Pharmacokinet 32: 345-356, 1997.
and a good overall review article is:
Kastrissios H, Blaschke TF. Medication compliance as a feature in
drug development. Ann Rev Pharmacol Toxicol 37: 451-75, 1997.
A now widely-recognized definition of 'patient compliance with a
prescribed drug regimen' is 'the extent to which the patient's dosing
history corresponds to the prescribed drug regimen'. That definition
presumes a reliable compilation of the patient's dosing history. The
virtual gold standard is electronic monitoring, short of which, one
would rely on careful interviews or questionnaires, which of course,
tend to overestimate the extent of drug exposure. Two recent
studies of the relative reliability of electronic monitoring and self-
report are:
Liu H, Golin CE, Miller LG, et al. A Comparison Study of Multiple
Measures of Adherence to HIV Protease Inhibitors. Ann Intern Med
134:968-77, 2001.
Arnsten JH, Demas PA, Faradegan H, Antiretroviral Therapy
Adherence and Viral Suppression in HIV-Infected Drug Users:
Comparison of Self-Report and Electronic Monitoring. Clinical
Infectious Diseases 33:1417-23, 2001.
An obvious problem arises when the prescribed regimen is, or is
suspected of being, suboptimal. One of the ways in which one
deduces suboptimality is by looking carefully at the clinical
correlates of reliably compiled dosing histories of individual patients
In general, patients' dosing histories are skewed downwards from
the prescribed regimen, because most of the errors in dosing (other
than with drugs of abuse) are errors of omission -- delayed or
omitted doses, sometimes with sequential doses omitted (called
drug 'holidays'). Drug holidays are, of course, widely recognized as
breeding ground for drug resistant micro-organisms.
A point that seems not to be widely recognized about the generally
successful DOT program with TB patients in New York City (Weis
SE, Slocum PC, Blais FX, et al. The effect of directly observed
therapy on the rates of drug resistance and relapse in tuberculosis.
N Engl J Med 330: 1179-84, 1994) is that the regimens adopted for
DOT have a much less frequent dosing than the originally
recommended regimens for the drugs in question. The NY DOT
regimen administers drugs on each of 4 days of the week. It is my
understanding that this regimen came to be what it is by economic
and logistical necessity, not any pharmaco-dynamic considerations.
I believe that the drugs in question (isoniazid, streptomycin, PAS,
rifampicin, and others) were all originally recommended for 3 or 4
times daily administration. But, as most were developed long ago,
at least some of those regimens were defined more by custom than
by science.
Yet the science of setting regimens is not ideal, either. Carl Peck
and his colleagues and Georgetown U in Washington DC, and Bert
Leufkens, Rob Heerdink and I in the Netherlands have
independently reviewed changes in dosing regimens in drugs
approved since 1980. Peck et al did it by reviewing FDA approved
changes in label-recommended dosing regimens, and we did it by
reviewing the WHO data on `defined daily doses', as compiled by
the WHO's statistical research centre in Oslo. We both have found
essentially the same thing, namely that, with literally one
exception, the changes in doses of drugs outside the anti-infectives
arena were all decreases of a factor of 2 or more, amounting to 1 in
about 4.5 - 5 drugs since 1980 undergoing such decreases in
recommended dose. In contrast, the changes that occurred among
anti-infective agents were all in the upward direction, presumably
reflecting changes in resistance, but that is only an assumption, and
indeed the matter needs careful study. The study by Peck et al is
so far published only in abstract form (Cross JT, Lee HK, Nelson
JS, Grudzinskas CV, Peck CC. One in Five Marketed Drugs
Undergoes a Dosage Change: 1980-1999. Clin Pharmacol Ther
69(2): 63, 2001.); our paper is in editorial review for publication. A
striking finding in our data is that the problem is getting worse, not
better, with the passage of time. Whether that change is due to
more errors being made in drug development, or better methods for
detecting suboptimal regimens, remains to be seen. A case can be
made for each.
Finally there is a question of taxonomy -- compliance, adherence,
concordance, cooperation, fidelity, l'observance (French),
therapietrouw (Dutch). The proliferation of terms arises in part be-
cause there are two very different research perspectives. One is a
behavioural perspective, which asks why people do not follow
professional advice as to what drug(s) to take, in what dose(s), and
when. The other is the pharmacologic perspective, which asks what
the clinical and economic consequences are of the deviations that
occur between prescribed regimens of drug administration and
patients' actual dosing histories. As Don Rubin pointed out a
decade ago, the spontaneously occurring variations in patients'
execution of prescribed drug regimens constitutes a rich natural
experiment in variable dosing and its clinical consequences. This
point would not be especially meaningful were it not for the fact that
defining the proper dosing regimen for a new drug is one of the
most difficult, if not the most difficult, problem in new drug
development.
Underlying the behavioural and pharmacologic perspectives is a
certain discontent with the sometimes authoritarian relationship
between doctor and patient. But before discarding the baby with the
bath water, it is well to recognize that also underlying these different
perspectives is the scientific authority of drug regimens that are
well-validated not only in the amount and timing of doses, but also
in the safe margins for deviation from amount and timing of doses.
The US and UK labelling for the combined estrogen/progestin oral
contraceptives provide so far the best validated regimens in both
respects, and are thus a model for all other prescription medicines
to emulate, if and when they are studied so as to adduce sound
evidence for such information. Such information sets clear limits on
how far one can deviate from the prescribed regimen without
incurring heightened risk that the hormonal blockade of
ovulation will cease, allowing breakthrough ovulation and the risk of
conception. Neither wishful thinking nor empathy for a patient's
quite reasonable wishes for less stringent requirements on the need
for punctuality in daily dosing will change these limits. Similarly tight
limits on punctual dosing appear to apply to the current protease
inhibitors for antiretroviral treatment of HIV infections (see Paterson
et al. Adherence to protease inhibitor therapy and outcomes in
patients with HIV infection. Ann Int Med 133: 21-30, 2000.)
The authors of the 'concordance' idea in the UK have given less
than adequate recognition to the fact that some regimens have
scientifically well-grounded, strict limits on how far one can deviate
from the recommended regimen without compromising
effectiveness or incurring hazard. That, plus the fact that there is no
verb form of 'concordance' ("I concord"?, no; "I am in concordance
with..."?, semantically OK, but too many words.) make
'concordance' a non-starter.
As we can now compile patients' dosing histories with a high degree
Of reliability (which was not possible until about a decade ago), we
need a taxonomy for deviations from scientifically well-supported
drug regimens to facilitate sound, quantitative analysis of the clinical
impact of the commonly occurring patterns of those deviations. To
that end, Bernard Vrijens and I made the following suggestion at the
Conference on Causal Inference held this past August, sponsored
by Emory University and the US National Cancer Institute. Taking a
leaf from Aristotle's analysis of the drama, we noted that ambulatory
pharmacotherapy has 3 parts, a beginning, a middle, and an end.
(Aristotle got much mileage from this stunningly simple but apt
formulation, which shows the value of being first.) In any case, the
'beginning' is the patient's acceptance, or not, of the prescribed drug
regimen -- a more or less dichotomous process, allowing for the
possibility of some ambivalent wobble for a short while. The 'middle'
is the patient's execution of the prescribed regimen -- a continuous
process, which can and, in some patients, often does vary widely
from one week to the next, as seen when one compiles patients'
dosing histories electronically. The 'end' occurs when, for whatever
reason (sometimes decided by the physician, sometimes by the
patient) the treatment is discontinued. The 'end' is also subject at
times to ambivalent wobble, but is inherently dichotomous.
The taxonomic problem posed by having one continuous, and two
dichotomous, processes is that a quantitatively useful parameter
cannot cover all three, because of their fundamentally different
dynamics. One can have a literary parameter, e.g., adherence, for
all three, but if one says that 'adherence is poor', there is no way,
short of separately examining the three parts, to know whether it
was non-acceptance, poor execution, or early discontinuation, or a
combination of the latter two. Accordingly, Vrijens and I suggested
using the term 'compliance' for the quality of execution of the
prescribed regimen, i.e., the extent to which the patient's dosing
history conforms to the prescribed regimen. That is the outcome of
comparing two time series (the dosing history and the regimen), and
which opens into a variety of different ways to express the data, the
choice among which is best guided by understanding of the drug's
underlying pharmacodynamics, which is not always well-enough
understood to be up to the task. Here again, the example of what
we know about the dynamics of hormonal blockade of ovulation,
and what kinds of experiments had to be performed to provide
sound dosing instructions, including what to do when one misses
doses, is a model for other chronic-use medicines.
A further, useful term is 'persistence', defined as the time between
the start of treatment and its end. Persistence is measured in units
of time. A patient can, in principle, comply well but have short
persistence, or comply only partially and have long persistence.
It is well to note, however, that persistence increases one dose at
a time, so that there probably is a fairly strong but not absolute
link between compliance and persistence, as defined. ('Persistence'
has recently begun to be recognized as a major determinant of
revenues from chronic-use prescription pharmaceuticals, leading
some to coin the term 'persistency', but as we don't say 'compliancy'
or 'resistancy', why pick on 'persistence' for useless embroidery?)
A key point about the term 'compliance': it is the term by which
papers are indexed by the National Library of Medicine, Index
Medicus, so whether authors choose other terms, e.g. adherence or
concordance, they are indexed under 'compliance'. That long-
standing practice stands as a form of force majeur in the field. The
American Heart Association has also adopted the term
"compliance", as reflected in the title of the excellent book which is
the most recent in their monograph series: "Compliance in
Healthcare and Research" Eds Burke LE, Ockene IS. Armonk
(NY): Futura Publishing Co., 2001 (American Heart Association
Monograph Series).
Among the lessons to be drawn from all of this, the foremost is the
value of reliable measurements. Research on patient compliance
was basically at an impasse as long as one had to rely on means of
assessing drug exposure in ambulatory pharmacotherapy that allow
patients easily to censor evidence for delayed or omitted doses.
Those means include returned tablet counts, questionnaires,
interviews, diaries, and, through the effects of 'white-coat
compliance', spot checks of drug levels in plasma. (Many usually
poorly compliant patients shift into correct dosing during the several
days prior to a scheduled visit, and about 80% of drugs in common
use have plasma half-lives of 12 hrs or less, which means that
correct dosing during the last 48 hrs prior to the time of blood
sampling will result in a drug concentration in the therapeutic range
so the seemingly objective measures of drug concentration in
plasma are subject, de facto, to patient censoring of evidence for
omitted doses.) Obviously, these various, easily patient-censored
methods can indicate poor or totally non-compliance, e.g. if the drug
level is zero, if all the pills are returned untaken, or if the patient
says he/she took few or no doses. But as the above cited papers
indicate, the odds are unfavourable for getting the proper mix of
recall, cooperation, and candour needed for reliability.
A penultimate point is that the whole subject is probably best set in
terms of measuring ambulatory patients' exposure to prescribed
drugs. If satisfactorily measured, exposure is what it is. When
exposure is compared to what it 'should be', you encounter the
problems created by sub-optimal regimens, of which history
suggests there are many still awaiting discovery. If the regimen is
prescribed on a 'take as needed' (prn) basis, then the concept of
what it 'should be' goes out the window, leaving one with the
measure of exposure as a quantitative indicator for how often the
patient perceived the need to take the medicine.
The ultimate point is that reliable measurements open a wholly new
Way to improve the quality of patients' execution of prescribed drug
regimens, viz. their compliance, as defined. Their key addition is
information on dose-timing and daily dosing patterns, which gives
the caregiver a unique handle on linking the dosing to robust
routines in patients' daily lives (see the papers by Cramer &
Rosenheck J Nerv Mental Dis 187: 53-4, 1999, and by Burnier et al.
J Hypertens 2001; 19:335-41). These papers show that about
3/4ths of otherwise poorly compliant patients can, with very little
investment of time on the part of the caregiver(s), achieve levels of
compliance good enough for all but the least forgiving medicines to
act effectively. These studies need to be repeated, as there are
many unanswered questions, but it is a promising new departure in
a field that many had written off as impractical or academic.
John Urquhart, MD, FRCP(Edin)
Professor of Pharmaco-epidemiology, Maastricht University, Maastricht NL
Chief Scientist, AARDEX Ltd/APREX Corp, Zug CH, Maastricht, Union City
CA
Professor of Biopharmaceutical Sciences, UCSF, San Francisco
urquhart@ix.netcom.com
--
To send a message to E-Drug, write to: e-drug@usa.healthnet.org
To subscribe or unsubscribe, write to: majordomo@usa.healthnet.org
in the body of the message type: subscribe e-drug OR unsubscribe e-drug
To contact a person, send a message to: owner-e-drug@usa.healthnet.org
Information and archives: http://www.healthnet.org/programs/edrug.html