Content uploaded by Allen Lebovits
Author content
All content in this area was uploaded by Allen Lebovits on Jan 29, 2018
Content may be subject to copyright.
Predictors
of
Physician Nonadherence
to Chemotherapy Regimens
Steven
J.
Schleifer,
MD,*
Sushi1 Bhardwaj, MD,t Allen Lebovits, PhD,$
J.
S.
Tanaka, PhD,§ Madelyn Messe, PhD, and James
J.
Strain, MDII
Physician nonadherence to cancer treatment regimens may diminish treatment
efficacy and compromise clinical research. The influence of clinical, demographic,
and psychosocial patient characteristics on physician adherence to breast cancer
chemotherapy was investigated, as was the role of the clinician’s attitudes
concerning the chemotherapy. One hundred seven women recently diagnosed with
breast cancer were followed for
26
weeks of treatment. Fifty-six(SZ%) of the
patients experienced unjustified modification for at least one chemotherapeutic
agent. Stepwise multiple regression revealed independent contributions of increased
patient age, treatment setting [clinic
versus
academic
or
community private
practice), and stage of disease to physician nonadherence. Regimen complexity,
delay in seeking treatment, and presence of psychiatric disorder did not contribute,
in general, to unjustified regimen modifications. Patient psychological and
psychiatric factors, however, did influence prescribing behavior for vincristine.
Physician awareness of factors contributing to unnecessary treatment modifications
may reduce the frequency of such behaviors.
Cancer
67:945-951,1991.
EDICAL NONCOMPLIANCE
is a prevalent but not
M
fully appreciated phenomenon. Twenty-five per-
cent to
50%
of patients fail to comply with medical rec-
ommendations including taking prescribed medications,
keeping appointments, and altering personal habits.’-5 In
addition to noncompliance by the patient, successful
medical management may be compromised by the phy-
sician’s nonadherence to defined treatment regimens for
reasons not justified by clinical indications. The extent to
From the Mount Sinai School of Medicine, New York, New York.
*
Associate Professor
of
Psychiatry, Director of Consultation/Liaison
Psychiatry, UMDNJ-New Jersey Medical School, Newark, New Jersey.
t
Associate Professor
of
Neoplastic Disease, The Mount Sinai Medical
Center, New York, New York.
4
Clinical Associate Professor, Departments of Anesthesiology and
Psychiatry, SUNY Health Science Center, Brooklyn, New York.
4
Associate Professor, Department of Psychology, New York Univer-
sity, New York, New York.
11
Professor/Director, Division of Behavioral Medicine
&
Consultation
Psychiatry,
The
Mount Sinai Medical Center, New York, New York.
Supported by National Cancer Institute (NCI) grant CA32960.
The authors thank Drs. James Holland, Roy Jones, Larry Norton,
and Steven Malamud for their help: and Jaimie Schwarzfeld, Harleen
Ruthen, and Daniel Remine for data collection.
Address for reprints: Steven
J.
Schleifer, MD. Department of Psy-
chiatry, UMDNJ-New Jersey Medical School, 185 South Orange Avenue,
Newark, NJ
07103.
Accepted for Publication April
27,
1990.
which such physician behaviors occur has not been sys-
tematically assessed.
The cancer chemotherapy setting provides a clinically
relevant model to assess patient and physician behavior
in response to factors that may serve to encourage devia-
tion from established protocols and which can also have
substantial consequences in relation to patient s~rvival.~~’
Bonadonna and Valagussa* found that breast carcinoma
patients receiving less than
85%
of their total chemother-
apy regimen had a poorer clinical course than those re-
ceiving more complete treatment. Only
17%
of the pa-
tients actually received criterion levels of treatment. For
62%
of those not receiving adequate treatment, subopti-
mal regimens appeared to result from factors other than
toxicity such as physician reluctance to risk adverse effects
and imprecise dosage calculations. Thus, although the
impact of patient noncompliance may be attenuated by
the increasing utilization
of
intramuscular and intrave-
nous routes
of
chemotherapy administration, effects as-
sociated with physician nonadherence may increase in
importance.
We studied adherence to breast cancer chemotherapy
protocols by both patient and physician in three practice
settings: academic-private office, community-private of-
fice, and university hospital clinic. Forty-three percent
of
945
946
CANCER
February
15
199
1
Vol.
61
patients showed evidence of noncompliance with orally
administered medications over a 6-month period8 and
52% of patients experienced nonadherent prescribing be-
havior by their physicians.’ Understanding factors that
contribute to physician nonadherent behavior may be of
crucial importance in the implementation of clinical and
research protocols and in the training of future oncologists.
This report describes the contribution of patient demo-
graphic, psychological, social, and medical factors, and of
physician treatment attitudes to physician nonadherent
behavior.
Subjects
and
Methods
To
reduce subject heterogeneity, a single disease (breast
cancer) and treatment modality (chemotherapy) were
studied. Physician adherence to the treatment regimen
was assessed for each prescribing opportunity over a 6-
month period. Patient compliance as well as psychosocial
characteristics were studied concurrently and the impact
of these factors on physician adherence ascertained. Pa-
tients were enrolled in a number of chemotherapy regi-
mens with defined criteria of drug dosage that permitted
identification of justified and unjustified deviations. The
variety of chemotherapy programs enabled the evaluation
of factors such as drug side effects, route of administration,
and protocol complexity.
Si
ib
jects
One hundred seven women recently diagnosed with
breast cancer, presenting to three treatment settings affil-
iated with the Mount Sinai Medical Center (New York,
NY), and signing consent to participate in the study, were
entered into the project. Accrual occurred by consecutive
referral by physicians who were asked
to
refer all patients
with breast cancer about to undergo their first course of
chemotherapy. Approximately one third of the potentially
eligible patients were not studied because of failure of
staff at doctors’ offices to refer to the study. Observation
of office practice patterns suggested that most nonreferrals
were random and related to oversight by busy staff.
Seitings
Three treatment settings were evaluated: an academic
hospital-based private office where treatment is rendered
by full-time attending faculty physicians (six physicians,
treating 28 patients): a clinic setting, where patients are
treated by the Department’s Fellows (nine physicians,
3
1
patients) along with clinic nurses: and community-based
private practice settings (ten physicians, 48 patients) out-
side the hospital where patients are treated by private at-
tending physicians and their staff.
Instrumenis
Assessments were obtained for physician’s prescribing
behavior for each patient based upon chart review for
each of the first 26 weeks of treatment. Patient compliance
behavior and psychological and other characteristics as
well as physician attitude data were obtained by interview
at regular intervals: at initiation of treatment and at
2,
4,
13,
and
26
weeks.
Research oncologist
’s
assessment
of
adherence:
The de-
termination of the prevalence of operationally defined
unjustified modifications among physicians using formal
treatment and research regimens was accomplished by
newly developed chart review methodologies, as follows.
An attending research oncologist
(S.
B.)
who was familiar
with the treatment and research protocols determined
nonadherence based on the physicians’ weekly written
chart recordings. Justified and unjustified modifications
were determined using the defined criteria for the regimen
employed by the clinician. For each patient visit, the
amount of each intravenously administered drug, the
prescribed oral medication for the following week, and
the reasons for deviation from the predetermined dosing
and schedule criteria by the treating oncologist were ob-
tained from the clinical oncology flow sheets and from
medical chart notes.
The research oncologist evaluated the physician’s pre-
scribing behavior for each of
26
weeks for each patient in
the study using the following information:
(1)
the actual
dosage administered for each drug; and (2) the expected
dosage for each drug based on regimen specifications
(e.g,
body surface area, stage of disease, hematologic and lab-
oratory parameters, and symptoms).
If
actual dosage dif-
fered from expected dosage, the treating oncologist’s rea-
son for modification, as noted in the chart, was assessed
by the research oncologist (e.g., blood count, rounding of
dose, dose decrease on schedule, mucositis, infection,
neuropathy, nonscheduled dose change approved by pro-
tocol chairperson, error in date of administration). Each
modification was judged as justified (within the accepted
guidelines of the regimen) or unjustified (a deviation from
the regimen). The research oncologist had access to the
age, height, weight, surface area, stage of disease, complete
blood counts, liver chemistries, and clinical symptoms
and signs known to the physician at the time he wrote
the chemotherapy order. All reviews were deferred until
the 26-week observation period had passed to avoid bias-
ing the physician’s prescribing behavior by an apparent
weekly audit.
Classijcation
of
regimen complexity:
To evaluate the
impact of the regimen on compliance, the research on-
cologist classified all chemotherapy regimens into simple
and complex, based on scheduling and number of drugs
administered.
No.
4
PHYSICIAN
NONADHERENCE
TO
CHEMOTHERAPY
-
Schleifer
et
a/.
947
Physician
s
semistrztctiired interview:
Physicians were
questioned by a trained research assistant using a struc-
tured four-point scale concerning expectations of patient
compliance, the physician’s attitudes concerning the pa-
tient’s regimen and anticipated success of her treatment,
and any instructions given to the patient (used to assess
patient compliance). These questions were administered
at the initiation of patient treatment and at
2,
4,
13, and
26 weeks after beginning chemotherapy.
Patient sernistrzictured interview:
Patients were inter-
viewed at the initiation of treatment and at 2, 4, 13, and
26 weeks. The interviews utilized variables found to be
important in prior compliance studies, using questions
derived from standardized instruments when available.
Demographic and clinical data were collected
on
age, stage
of
disease, race, marital status, income, treatment location,
perceived social support, level
of
education, perceived
fi-
nancial “comfort,” the prescribed chemotherapy regimen,
delay from onset of symptoms to physician contact and
diagnosis, and patient attitudes toward recovery from
cancer in general, breast cancer in particular, and her own
likelihood of recovery. Psychiatric and psychological
measures (initial, 13, 26 weeks) included a Research Di-
agnostic Criteria (RDC)” checklist, the Mini Mental State
Examination,” and SCL-90 Analogue.12 The Karnofsky
Patient Performance Rating13 was assessed at the same
intervals and the Dosage Record and Treatment Emergent
Scale (DOTES)I4 was administered at each follow-up in-
terview to quantify treatment side effects. This instrument,
originally designed to assess the effects of psychotropic
medications, has many items relevant to cancer patients
and was adapted to include side effects specifically appli-
cable to cancer chemotherapy
(e.g.,
hair loss).
Patients were interviewed concerning their actual med-
ication use during the preceding week at
2,
4, 13, and 26
weeks. This was compared with the physician’s prescrip-
tion to determine patient compliance.
Definitions
yf
Physician Nonadherence
Because of the lack of standard definitions, three indices
of
physician nonadherence were developed.
Behavioral nonadherence:
Behavioral nonadherence
(BNA)
was defined as the percentage
of
prescribing oc-
casions (behavioral events) when either less than 90%
or
more than
1
10% of the expected milligram amount of a
specific drug was prescribed by the physician
on
a non-
justified basis.
This index provides a profile of the physician’s pre-
scribing behavior and
is
sensitive to the frequency of de-
viant decisions during a course of treatment. It does not
reflect the actual amount
of
drug that was overprescribed
or
underprescribed.
Total chemical nonadherence:
Total chemical non-
adherence (TCN) was defined as the total milligram dose
(nonjustified) deviance as a percentage
of
the expected
milligrams
of
drug to be prescribed according to regimen
for the observation period
(e.g.,
26 weeks).
As
such, the
total chemotherapy received over the entire treatment pe-
riod may be predictive of the potential medical conse-
quences of physician behavior.
Total net dosage nonadherence:
Net dosage nonadher-
ence (NDN) represents the dosage consequences of each
deviant prescribing decision over the period of treatment.
The absolute milligrams of deviance from the regimen-
defined expected dose to be administered for each week
(i.e.,
overprescribing and underprescribing) was deter-
mined and the percentage deviation for each visit calcu-
lated. The mean
of
the percentage deviations for all
of
the patient visits was calculated.
This decision-based quantification was designed to
measure the (absolute) magnitude
of
physician deviation,
without allowing overprescribing and underprescribing
occasions to cancel each other. This measure may be es-
pecially useful as an estimate of the net magnitude effect
of behavioral deviations and of the potential consequences
of such deviations, especially for treatment regimens in
which the sequence
of
medication administration is cru-
cial.
Patient Noncompliance
Two measures of noncompliance were developed:
“dosage” noncompliance, defined as positive when
10%
or
more of the total prescribed dosage of the drug was
missed over the
four
observation periods: and “behav-
ioral” noncompliance, defined as positive when there were
one
or
more occurrences during the
four
assessed pre-
scription intervals in which the criterion dosage of a drug
was not taken. The criterion for behavioral deviation was
10%
or
more of the prescribed dosage.
Data Analvsis
Student’s
t
test and analysis of variance techniques for
continuous variables and chi-square for dichotomous
variables were employed in preliminary bivariate tests.
To assess the contribution of patient and physician char-
acteristics to physician nonadherence, a series of partially
hierarchical, stepwise multiple regression analyses were
undertaken on the sample
of
82
patients
for
whom com-
plete data was available. The primary analyses utilized,
as the independent variables, a predetermined set
of
vari-
ables representative of the sociodernographic, behavioral/
psychological, and medical domains considered likely,
a
priori.
to contribute
to
physician behavior. These included
patient’s age, stage
of
disease (dichotomized as 1-11
versus
Ill-IV),
complexity of chemotherapy regimen, treatment
948 CANCER
February
15
1991
Vol.
61
setting, presence of psychiatric disorder at any point dur-
ing the course of treatment, and patient delay in seeking
treatment.
Another set of primary variables relating to physician
attitudes concerning anticipated success of treatment was
available
for
most of the subjects (n
=
63); a second set
of
analyses were undertaken including these attitudinal
variables along with the other primary independent vari-
ables. For all analyses, age was forced into the model, as
a control as well as a predictor variable, with the other
variables permitted to enter stepwise. Following these
analyses, tertiary analyses were undertaken to determine
whether entry of additional variables belonging to the
several domains
of
interest clarified relationships obtained
in the primary analyses
or
suggested other important pre-
dictors of nonadherence.
The regression analyses
for
the three measures of phy-
sician adherence behavior (BNA, TCN,
NDN)
were first
undertaken to determine the predictors
of
physician non-
adherence for any of the drugs in each patient’s treatment
regimen. Further analyses was then undertaken to begin
to investigate predictors
of
nonadherence
for
specific
agents. Due to the potentially large number of statistical
tests and the small number
of
patients for some individual
drugs, a more cautious analytic approach was used: anal-
yses with the “primary” variables were undertaken as de-
scribed above.
For
the secondary independent variables,
analyses were considered only to clarify effects obtained
consistently
(ie.,
for at least two definitions
of
nona-
dherence) with the primary independent variables.
Results
Of the 29 physicians
in
the three treatment locations,
48%
(N
=
14) prescribed unjustified modifications for their
patients some time during the 26-week observation period;
52Y0 of the 107 patients
(N
=
56)
experienced at least one
occurrence of unjustified modification for at least one of
their chemotherapy medications. Bivariate comparisons
revealed significant differences across locations in the
prevalence of nonadherent behavior, with the academic
private practice setting having the least nonjustified de-
viations and the clinic setting the most.’
Physician Nonadherence jor
Any
Drug
in
the
Regimen
Regression analyses were performed to determine first
the relative contribution of the main demographic, psy-
chological, and medical variables to the three measured
dimensions
of
nonadherent behavior (see “Methods”).
These included the following: patient age, stage of disease,
location, complexity of treatment regimen, delay in seek-
ing initial evaluation and treatment for the cancer, and
the presence
of
any RDC psychiatric disorder.
For
BNA, the stepwise regression analysis
on
the main
independent variables
(n
=
82) revealed a number of fac-
tors to be independently predictive
of
physician nona-
dherence. These included increased patient age
(F
7.8
1;
degrees of freedom [dfl 1,80;
P
<
0.007), location of treat-
ment
(F
5.26; df
1,80;
P
<
0.02), and higher stage ofdisease
(F 4.08; df 1,80;
P
<
0.05).
There was
no
significant effect
of regimen complexity, delay in seeking treatment, or
presence of psychiatric disorder. The location of treatment
effect indicated higher levels of nonadherence in the clinic
compared with the private community and academic set-
tings.
Our
prior observation in bivariate analyses of greater
nonadherence in the clinic compared with other settings’
was thus confirmed when controlling for the other de-
mographic and clinical characteristics. The analysis on
the subset of patients for whom attitudinal data was avail-
able (physician’s reservation about giving the patient her
regimen, physician’s confidence in the success of the
treatment, and patient optimism concerning success of
cancer treatment [n
=
63]), showed no significant con-
tribution of these variables to physician behavior.
Analyses of the effects of the main independent vari-
ables on the two dose-related measures of adherence
(n
=
82) also revealed a significant contribution
of
increased
patient age to physician nonadherent behavior
(TCN:
F
6.78; df
1,80,
P
<
0.01;
NDN:
F
5.49; df
1,80,
P
<
0.02)
with no contribution
of
the other variables to these mea-
sures.
Similar to the findings for
BNA,
the attitudinal
variables were not significant predictors of physician non-
adherent behavior.
To further explore the several domains of potential
predictors of nonadherence, the secondary analyses were
undertaken. Analyses of the sociodemographic variables
(N
=
56)
revealed that income and racial background,
which were each highly correlated with treatment location
(P
<
0.001), entered the regression models along with age
and in preference to location. Lower income (F
=
7.7
1
;
df 1,52;
P
<
0.008)
and the presence of a larger social
network (F 2.95; df 1,52;
P
=
0.09) were associated with
BNA
whereas non-white
versiis
white racial background
was associated with TCN
(F
5.19; df
I
,52,
P
<
0.03) and
NDN
(F
12.03, df 1,52;
P
<
0.001)
physician nonadher-
ence. Sample size precluded the determination of whether
any of these intercorrelated variables were significantly
more important predictors than location itself.
Secondary analyses assessing the various components
of the psychological domain, including RDC diagnosis
and psychological symptoms as measured by the SCL-90
analogue (depression, anxiety, somatization, hostility, and
the global index), revealed no evidence for a contribution
of any of the psychological symptom dimensions to the
three measures of physician nonadherence
(P
>
0.1);
age
and location entered into the model as described.
Secondary analyses
of
an expanded list of medical
symptomatology variables provided no evidence for a
No.
4
PHYSICIAN NONADHERENCE
TO
CHEMOTHERAPY
Scklezfer ct
al. 949
contribution of patient functional health status, total drug
side effects,
or
number of hospitalizations to any of the
three measures of physician nonadherence (P
>
0.1
).
Once
again, age and location
of
treatment entered into the
model as described above.
Finally, in order to determine if those patients who
themselves tended to be less compliant with orally ad-
ministered medications experienced more physician non-
adherence, regression analyses were performed including,
as independent variables, age, stage of disease, and treat-
ment location, as well as patient behavioral and dosage
noncompliance (n
=
51). There was no evidence that pa-
tient compliance behaviors were associated with greater
physician nonadherence for any of the three measures of
physician behavior
(P
>
0.1).
Phjsiciari Nonadherence
to
Specific Chemotherapeutic
Agents
To investigate whether the observed predictors of phy-
sician nonadherence to chemotherapy regimens were as-
sociated with specific chemotherapeutic agents, the pri-
mary regression analyses were repeated for patients re-
ceiving those specific agents that were commonly
prescribed. Since each drug was prescribed for a differing
subset
of
patients, the data for single drugs is not directly
comparable either with that obtained across all drugs or
for other single agents. In addition, the reduced sample
for some drugs and the multiple statistical tests dic-
tated more cautious analyses and interpretation (see
“Methods”).
Cvto.uan
(cyclophosplzainide):
More advanced age was
associated with increases in each ofthe three types of phy-
sician nonadherence (BNA:
F
3.47; df 1,76,
P
<
0.07;
TCN: F 8.27; df 1,76,
P
<
0.005; NDN:
F
3.61; df 1,76,
P
<
0.06). Location also predicted BNA
(F
4.48; df 1,76,
P
<
0.04), with nonadherent behavior highest in the clinic
setting. Location did not predict dosage-related (TCN,
NDN) nonadherence. No other variables were predictive
of physician nonadherence with Cytoxan (n
=
78).
Methotrexate:
Here also, increased age predicted each
type of physician nonadherence (BNA:
F
6.23, df 1,7
1.
P
<
0.02;
TCN:
F
5.8
1,
df 1,7
1,
P
<
0.02; NDN:
F
4.13, df
1,71,
P
<
0.05) and location predicted BNA
(F
4.0, df
1,7
1,
P
<
0.05) but not TCN
or
NDN.
For
methotrexate,
more advanced stage of illness was also related to BNA
(F
4.6
1,
df 1,7
1,
P
<
0.03). No other variables were pre-
dictive (n
=
73).
5-Fluorouracil:
As with the other agents often used
concurrently (Cytoxan, methotrexate), increased age pre-
dicted physician nonadherence (BNA:
F
4.63, df 1,76,
P
<
0.04; TCN: F 9.20, df 1,76,
P
<
0.003; NDN:
F
5.06,
df 1,76,
P
<
0.03). More serious stage of illness, as with
methotrexate, predicted BNA
(F
9.2; df 1,76,
P
<
0.003).
None of the other primary variables were predictive of
physician behaviors. An association was suggested be-
tween the attitudinal variables and 5-FU prescribing be-
havior, although effects were not consistent across mea-
sures: physician’s reservations about giving the patient
the specific regimen appeared to be associated with BNA
(F 2.85, df 1,50,
P
<
0.1) and lower patient optimism
concerning treatment with TCN
(F
2.82, df 1,53,
P
<
0.1)
(n
=
78).
Vincristine:
There was no contribution of age or loca-
tion to physician behavior for this drug; however, the
presence of an RDC diagnosis appeared to be associated
with nonadherence to vincristine: TCN
(F
3.2, df 1,39,
P
<
0.08); NDN
(F
2.98, df 1,39,
P
<
0.09).
Secondary
analyses with the expanded set of psychological variables
(n
=
45) revealed that increased patient (global) psycho-
pathology significantly predicted both TCN
(F
5.44, df
1,36,
P
<
0.03) and NDN
(F
5.51, df 1,40,
P
<
0.02)
nonadherence by their physicians. Higher patient anxiety
was associated with greater BNA
(F
5.82, df 1,38,
P
<
0.02).
In contrast, fewer somatization symptoms were
associated with nonadherence (BNA:
F
10.73, df
1,38,
P
<
0.002; TCN
F
6.98, df 1,36,
P
<
0.01; NDN
F
9.46, df
1,40,
P
<
0.004) as were lower levels of hostility (BNA:
F4.34,df1,38,
P<0.04;NDN:F5.5l,df1,40,P<0.02)
(n
=
42).
Prednisone:
Location alone predicted BNA
(F
5.13, df
1,36,
P
<
0.03); complexity of chemotherapeutic regimen
appeared to predict TCN (F 3.09, df 1,35,
P
<
0.09), and
none of the variables predicted NDN deviation (n
=
39).
Discussion
Physician nonadherence to treatment regimens and re-
search protocols for breast cancer may contribute to pa-
tient morbidity‘ as well as compromise accumulation of
clinical research data.’ Investigators of compliance have
noted that noncompliant behavior is determined by mul-
tiple The current study focused primarily on
patient attributes that may contribute to physician be-
havior. The most salient observation was that patient de-
mographic and clinical characteristics were
of
greatest
importance in predicting physician adherence to regimens
whereas patient psychological characteristics did not, in
general, appear to influence physician’s prescribing be-
havior. The latter, however, appeared to influence deci-
sions concerning treatment with vincristine.
Physician behavior was predicted independently by
several patient characteristics: age, treatment location site,
and stage of disease. The factors of age and advanced
disease in the patient may elicit concerns in the physician
that result in excessive caution in the treatment process.
That such responses may not have been apparent to the
physician is suggested by the failure of attitudinal variables
(e.g..
reported reservations about giving the patient che-
950
CANCER
February
15
199
1
Vol.
67
motherapy) to predict physician adherent behavior. It is
worth noticing that although the prevalence of nonad-
herent behavior was high
(52%
of patients), physicians
reported no reservations about giving chemotherapy for
68%
of their patients.
In contrast to factors which are themselves overt such
as age and stage of illness, patient characteristics that may
be less apparent did not appear to influence physician
behavior. These included the presence of psychiatric syn-
dromes (which had a prevalence of
20%
in the current
sample) and psychological symptomatology, the patient’s
attitudes concerning treatment, and the patient’s com-
pliance behavior. These observations are consistent with
other studies showing that general physicians often fail to
detect the presence of psychiatric
disorder^."^'^
The pa-
tient’s functional status and drug side effects profile also
did not appear to be associated with unjustified prescribing
behavior. Studies are indicated to determine the extent
to which physicians are aware of these specific patient
factors as they prescribe chemotherapy.
Consistent with the hypothesis that physician’s pre-
scribing behavior is associated with awareness of patient
factors
perceived
as relevant, we found that patient psy-
chological state, although not predicting nonadherence in
general, was associated with prescribing behavior for vin-
cristine, a drug with known neurologic and psychological
effects. Physicians may focus on central nervous system
(CNS)-related symptoms in patients treated with this agent
and be especially sensitive to psychological symptoms in
this group. Patient knowledge and verbalization concern-
ing the neurologic and psychological effects of drugs such
as vincristine may also influence physician behavior.
However, prescribing behavior for prednisone, another
agent with important neuropsychiatric effects, was not
significantly related to patient psychological symptoms,
indicating a need for further study of this dimension.
Treatment location was a major predictor of physician
nonadherence, with nonadherent behavior significantly
greater in the clinic than in the other settings. This may
relate to both patient and physician characteristics. The
clinic population had significantly lower income and so-
cioeconomic status and included more non-whites than
the other settings’; lower income and racial background
may in fact be more powerful predictors of physician
nonadherence than location itself. It may be that these
patient characteristics elicit subtle expectations and con-
cerns about patient behavior that result in more cautious
prescribing behavior. In addition to these possible patient-
related factors, physicians in academic clinic settings tend
to be less experienced than those in other settings. They
may be more cautious and possibly less comfortable with
lower socioeconomic status patients. It is also possible
that some clinic physicians were less sufficiently informed
of the precise constraints of the diverse treatment regi-
mens.
In addition to assessing predictors of overall compliance
with treatment regimens, the predictors of nonadherence
for each individual drug were investigated. Two salient
observations emerged from these analyses. First is the
suggestion that the contribution of patient age and location
to physician behavior is related to treatment with Cytoxan,
methotrexate, and 5-fluorouraci1, drugs commonly pre-
scribed for breast cancer patients and often prescribed
together (CMF). Second, as noted, is that physician’s pre-
scribing behavior for vincristine is related to very different
factors. Although preliminary in nature, the data specif-
ically suggest greater vincristine nonadherence in patients
manifesting more anxiety and global psychological symp-
toms, but fewer somatic preoccupations and less expressed
hostility. Physicians concerned about vincristine’s CNS
toxicity may be especially cautious in treating the anxious
but passive patient who may actually underreport toxic-
ities.
It is worth noticing that of the three measures of non-
adherence used, the behavioral measure
(BNA)
appeared
to be most sensitive to the predictors. This is not unex-
pected since this behavior-based measure is most closely
linked to the decision making of the physician. It may
therefore also be the measure of greatest value in devel-
oping and monitoring interventions to reduce physician
nonadherent behavior. The TCN and NDN dose-related
measures,
in
contrast, may be most useful in assessing
the potential clinical consequences of physician nonad-
herent behaviors. Total chemical nonadherence
or
NDN
may be preferable depending upon whether the relevant
clinical factor is the total dose delivered
or
the week by
week dosage deviations.
Physician nonadherence to cancer chemotherapy reg-
imens requires further evaluation. The current study was
limited in the scope of its assessment of physician char-
acteristics and was not designed to sample large numbers
of physicians representative of the range of backgrounds,
practice conditions, and styles. The number of patients
studied with the extensive psychological measures was
also
moderate and focused on those with a single diagnostic
entity treated with a select group of chemotherapeutic
agents. Although
it
is likely that some aspects of physician
nonadherent behavior will be similar with other neoplastic
diseases and treatments, factors unique to specific disor-
ders require investigation
(e.g.,
physician gender, personal
and family cancer history, gynecologic
versus
other neo-
plasias). Studies are required utilizing samples sufficient
to investigate physician characteristics such as age, level
of training, years in practice, attitudes toward specific reg-
imens and disease states, as well as interactions of specific
physician characteristics with those
of
the patient.
No.
4
PHYSICIAN
NONADHERENCE
TO
CHEMOTHERAPY
-
Schleifer
et
af.
95
1
Whereas elucidation
of
the determinants
of
physician
nonadherent behavior in the cancer setting awaits further
research, education
of
oncologists-in-training concerning
this issue appears to be indicated. Ongoing education and
support for
all
physicians participating in the testing and
implementation of chemotherapy regimens may reduce
the frequency and magnitude of avoidable treatment
modifications.
REFERENCES
1.
Haynes RB. Introduction. In: Haynes RB, Wayne T, Sackett DL,
eds. Compliance in Health Care. Baltimore: Johns Hopkins,
1979; 1-7.
2.
Bonadonna G, Valagussa P. Dose response effect of adjuvant che-
motherapy in breast cancer.
N Engl JMed
1981;
304:lO-15.
3.
Feinstein AR. Clinical biostatistics: Biostatistical problems in
“compliance bias.”
Clin Pharmacol Ther
1977; 169464357.
4.
Blackwell B. Medical intelligence: Drug therapy and patient com-
pliance.
N
Engl
JMed
1973; 289:249-252.
5.
Marston MV. Compliance with medical regimens: A review of the
literature.
Nursing Res
1970; 19:312-323.
6.
Gastorf JW, Galanos AN. Patient compliance and physicians’ at-
titude.
Fam
Pruct Res
J
1983; 2:190-198.
7.
Rothert MG. The likelihood of patient adherence to a medical
regimen: Comparison of patients’ and physicians’ judgments.
Dissert
Abstr
Intl
1981;
41(9-A):3895.
8.
Lebovits AL, Strain JJ, Schleifer
SJ,
Tanaka JS, Bhardwaj
S,
Messe
MR. Patient noncompliance with self-administered chemotherapy.
Cancer
1990; 65:17-22.
9.
Strain
et
a/.
Submitted for publication.
10.
Spitzer RL, Endicott J, Robin E. Research diagnostic criteria:
Rationale and reliability.
Arch Gen Psychiatry
1978; 35:773-782.
11.
Folstein MF, Folstein SE, McHugh PR. Mini-Mental State.
J
Psychiatr Res
1975; 12:189-198.
12.
Derogatis LR, Lippman
RS,
Covi L. The
SCL-90:
An outpatient
psychiatric rating scale-preliminary report.
Psychopharm
Bull
1973; 9:
13.
Karnofsky DA, Burchenal JH. The clinical evaluation of che-
motherapeutic agents in cancer. In: MacLead CM, ed. Evaluation of
Chemotherapeutic Agents. New
York
Columbia University Press,
1949;
14.
Dosage record and treatment: Emergent symptom scales (DOTES).
In: Guy
W,
ed. ECOEU Assessment Manual for Psychopharmacology.
Rockville, MD:
1976;
DHEW publication no.
76-338. 223-244.
15.
McDonald
M,
Grimm RH Jr. Compliance with hypertension
treatment: Strategies for improving patient cooperation.
Postgrad Med
16.
Kane JM. Compliance issues in outpatient treatment.
J
Clin
Psy-
chopharm
1985; 5:22S-27S.
17.
Marks J, Goldberg D, Hillier V. Determinants of the ability of
general practitioners to detect psychiatric illness.
Psycho1 Med
1979; 9:
337-353.
18.
Kessler L, Cleary P, Burke J. Psychiatric disorders in primary
care: Results of a follow up study.
Arch Gen Psychiatry
1985; 42:583-
587.
13-27.
I9 1-205.
1985; 77:233-236, 241-242,
