San Diego State University Solution to The Opioid Crisis Articles Analysis

Journal of Substance Abuse Treatment 44 (2013) 515?521
Contents lists available at SciVerse ScienceDirect
Journal of Substance Abuse Treatment
Effectiveness of drug tests in outpatients starting opioid substitution therapy
Julie Dupouy, M.D. a, b, c,?, Lise Dassieu, M.Sc. d, Robert Bourrel, M.D. e, Jean-Christophe Poutrain, M.D. a,
Serge Bismuth, M.D. a, St?phane Oustric, M.D. a, Maryse Lapeyre-Mestre, M.D., Ph.D. b, c
a
D?partement Universitaire de M?decine G?n?rale, Universit? de Toulouse, Facult? de M?decine, 133 route de Narbonne, 31063 Toulouse, France
Inserm UMR1027, Universit? de Toulouse III, Facult? de M?decine, 37 all?es Jules Guesde, 31000 Toulouse, France
CEIP-Addictovigilance, Service de Pharmacologie Clinique, Facult? de M?decine, 37 all?es Jules Guesde, 31000 Toulouse, France
d
Laboratoire Interdisciplinaire Solidarit?s Soci?t? Territoires ? Centre d’Etude des Rationalit?s et du Savoir UMR 5193, Universit? de Toulouse II, France
e
Echelon R?gional du Service M?dical Midi-Pyr?n?es, Caisse Nationale d’Assurance Maladie des Travailleurs Salari?s Cnam-TS, Toulouse, France
b
c
a r t i c l e
i n f o
Article history:
Received 9 July 2012
Received in revised form 20 November 2012
Accepted 20 November 2012
Keywords:
Opioid-related disorders
Substance abuse detection
Opiate substitution treatment
Ambulatory care
Cohort studies
a b s t r a c t
We aimed to assess the effectiveness of drug tests for treatment retention in outpatients starting opioid
substitution therapy. A retrospective cohort was created from the data of the French health insurance system
database for the Midi-Pyrenees region. Patients starting opioid substitution treatment (OST) were included
and followed for 18 to 30 months. Two groups of patients were de?ned: the drug test group (at least one
drug test reimbursement) and a control group (no drug test reimbursement). The cohort included 1507
patients. During follow-up, 39 subjects (2.6%) had at least one drug test reimbursement. Mean treatment
retention was 207 days in the control group and 411 days in the drug test group (p b 0.001). With a
multivariate Cox model, drug tests were associated with treatment retention: hazard ratio 0.55 (95% CI: 0.38?
0.80). Use of a drug test in follow-up of opioid substitution treatment, although rarely prescribed,
signi?cantly improved treatment retention.
? 2013 Elsevier Inc. All rights reserved.
1. Introduction
In 2009, nearly 140,000 patients were treated in France with an
opioid substitute (OS) (Commission Nationale des Stup??ants et des
Psychotropes, 2010). Methadone and buprenorphine received marketing authorization in 1995 and 1996, respectively, for substitution
treatment for major opiate dependence as part of overall medical,
social and psychological therapeutic management. These two drugs
are prescribed and delivered according to two different, very strict
guidelines. Methadone, a pure ? agonist, is a listed narcotic and can
be prescribed for a maximum of 14 days and delivered for a
maximum of 7 days. Primary prescription of methadone is restricted
to physicians in specialized units and to hospital physicians. When
the patient is stabilized, treatment may be continued in an outpatient
setting and followed up by any physician, whether specialist or
primary care physician (PCP). Buprenorphine, a partial ? agonist, is a
class I psychotropic drug (Schedule III of the 1988 Convention) and
can be prescribed for maximum periods of 28 days and delivered for
periods of 7 days. It is widely used in France as it can be prescribed by
any physician, whether specialist or PCP, and can be delivered in any
? Corresponding author. Inserm UMR1027, Universit? de Toulouse III, Facult? de
M?decine, 37 all?es Jules Guesde, 31000 Toulouse, France. Tel.: +33 561145918; fax:
+33 561145928.
E-mail address: [email protected] (J. Dupouy).
0740-5472/$ ? see front matter ? 2013 Elsevier Inc. All rights reserved.
http://dx.doi.org/10.1016/j.jsat.2012.11.006
local pharmacy, so in this country more than three-quarters of
patients treated for opioid addiction receive buprenorphine (Auriacombe, Fats?as, Dubernet, Daulou?de, & Tignol, 2004; Fatseas &
Auriacombe, 2007). In summary, in France, methadone must be
started in a specialized unit or a hospital and can be continued, after
the patient is stabilized, in an outpatient setting, whereas buprenorphine is easily accessible as it can be started in a specialized unit, a
hospital or an outpatient setting. The physician can prescribe
methadone for a maximum of 14 days and buprenorphine for
28 days, but the pharmacist must deliver the medication only for
7-day periods unless otherwise indicated by the prescriber. The
systematic review by Mattick, Kimber, Breen, and Davoli (2008)
demonstrated the ef?cacy of buprenorphine maintenance treatment,
with a lower retention rate than methadone but giving a similar
decrease in opiate consumption. Maintenance on OS is requisite for
successful treatment.
Methadone prescription guidelines detail the recommended urine
tests: a ?rst, obligatory test before starting methadone treatment and
later control tests. The ?rst urine test con?rms current drug
consumption and the absence of methadone intake. Tests are
subsequently done once or twice a week during the ?rst 3 months
of treatment, then twice monthly. When the patient has transferred to
an outpatient setting, tests can be done if the physician considers it
necessary. Tests are not obligatory for buprenorphine. In 2004,
updated French guidelines on optimal opiate addict care reinforced
these recommendations and advised a standardized screening test
516
J. Dupouy et al. / Journal of Substance Abuse Treatment 44 (2013) 515?521
schedule in the initiation and follow-up of methadone and buprenorphine treatment (Haute Autorit? de Sant?, 2004).
The drug tests can be carried out by immunochemical methods,
either by automated analyzers in the biology laboratory or by drug
screening kits. Screening can be done during the patient’s visit, in
either a specialized addiction center or the physician’s of?ce. These
tests (whether on a laboratory automat or using a commercial kit) are
qualitative and have de?ned thresholds. False negatives exist for
cocaine and benzodiazepines in particular, but test quality is
intrinsically better for opiate detection (1.9% false negatives) (Pesce
et al., 2010). Results can be con?rmed by the reference method, liquid
or gas phase chromatography with mass spectrometry, which gives
quantitative measurements (Bag?ien, Morken, Zahlsen, Aamo, &
Spigset, 2009). Laboratory tests, regardless of the method or the
biological medium, are reimbursed by the French health insurance
system with no limit on their number or time period. A recent study
has shown that PCPs rarely used these tests and that most had limited
knowledge of the subject (Dupouy, Bismuth, Oustric, & LapeyreMestre, 2012).
The value of drug tests for the diagnosis of substance abuse has
been demonstrated: many studies agreed that drug tests were more
sensitive than self-reports (Galletly, Field, & Prior, 1993; Kilpatrick,
Howlett, Sedgwick, & Ghodse, 2000; Lundy et al., 1997; Olshaker,
Browne, Jerrard, Prendergast, & Stair, 1997; Perrone, De Roos,
Jayaraman, & Hollander, 2001) or physician clinical evaluation
(Mordal, Holm, M?rland, & Bramness, 2010). The value of drug
testing for clinical management and patient outcomes has been
further assessed in several settings: in school programmes (Roche,
Bywood, Pidd, Freeman, & Steenson, 2009) and in occupational
drivers (Cashman, Ruotsalainen, Greiner, Beirne, & Verbeek, 2009),
drug tests are known to be effective. In emergency settings, a recent
review found that drug tests had no in?uence on therapeutic
management (Tenenbein, 2009). In chronic non-cancer pain patients,
the effectiveness of urine testing to reduce opioid misuse is still
debated (Starrels et al., 2010).
As far as we are aware, no study has yet assessed the effectiveness
of these tests in ambulatory care in opioid-dependent patients,
although they are recommended and they could help to improve
therapeutic management when opiate substitutes are used. Our aim
was to assess the effectiveness of drug tests for treatment retention in
outpatients starting opioid substitution therapy.
2. Materials and methods
Data were extracted from the database of the main French health
insurance system (Extraction, Recherches, Analyses pour un Suivi
M?dico-Economique, ERASME) for the Midi-Pyrenees region from
January 1, 2009 to December 31, 2011 in order to build a cohort of
patients who were starting opioid substitution therapy (Appendix Fig.
1). This health insurance system covers 87% of the French general
population (other systems cover speci?c populations such as farmers,
soldiers and railway workers) (Martin-Latry & B?gaud, 2010).
Patients with universal coverage (coverage for the unemployed and
low income insurees) are automatically registered in this system.
Inclusion criteria for the study were absence of reimbursement for OS
between January 1, 2009 and June 30, 2009 and at least one recorded
reimbursement for OS between July 1, 2009 and June 30, 2010. These
patients had therefore started an OS between July 1, 2009 and June 30,
2010 and were followed for 18 to 30 months. The reference date was
de?ned as the ?rst OS delivery. The period of 30 days before the ?rst
delivery of the ?rst OS prescribed to the end of opioid substitution
treatment (OST) was considered as the addiction treatment period.
The study was approved by the French Data Protection Authority
(Commission Nationale Informatique et Libert?s, CNIL), authorization
for evaluation of health care practices n?1516745.
2.1. De?nition of exposure
The exposure studied was prescription of a drug test during the
addiction treatment period. Exposure was considered as a timedependent variable with a single change, and the period of exposure to
a drug test was de?ned as the period extending 30 days before the ?rst
test was carried out until occurrence of the event or censoring (Appendix
Fig. 1). Laboratory tests were identi?ed and described by the laboratory
classi?cation codes used for reimbursement by the health insurance
system (Appendix Table 1). Tests for analgesics, narcotics and psychotropic drugs were taken into account, whatever the biological medium.
2.2. De?nition of the event
The primary outcome measure was OST retention, de?ned as
regular delivery of an OS (delivery every 35 days maximum for
buprenorphine and 18 days for methadone). A period of more than
35 days between two deliveries of buprenorphine (or more than
18 days for methadone) was considered as treatment interruption
and has been validated as such in other studies (Pradel et al., 2004,
2009). This de?nition of OST retention takes into account the
particular conditions of prescription and delivery of these drugs. A
patient who interrupted treatment was considered as having
discontinued treatment (any later reinitiation was considered a new
treatment cycle and was not analyzed here). Data were rightcensored at the end of the study period (administrative censoring at
December 31, 2011), at the patient’s death or loss to follow-up. For
patients who presented the event ?treatment discontinuation? before
death, the event was taken into account if it occurred more than
30 days before death. If the date of the last reimbursed act or the last
reimbursed drug was prior to the date of treatment discontinuation by
the patient, he/she was considered as lost to follow-up.
2.3. De?nition of co-variables
The doctor-shopping indicator, proposed by Pradel et al. (2004,
2009), was used to take into account simultaneous use of several
physicians by a patient in order to obtain prescriptions. To determine
the doses of methadone and buprenorphine delivered, we used
de?ned daily doses according to the World Health Organization
guidelines (WHO Collaborating Centre for Drug Statistics Methodology). The dose delivered corresponded to the daily dose delivered to
the patient. The dose prescribed corresponded to the daily dose that
would have been delivered to the patient if he/she had only one
physician. The doctor-shopping dose corresponded to the daily dose
obtained by the prescriptions of several physicians in a same period of
time. The doctor-shopping indicator was obtained by dividing the
doctor-shopping dose by the dose delivered. If the doctor-shopping
indicator was greater than zero, we considered that the patient
exhibited doctor-shopping behavior (binary variable).
We used other reimbursed drugs as indicators of associated
comorbidities. These drugs were identi?ed by their anatomical
therapeutic chemical classi?cation code (ATC code) throughout the
study period (Appendix Table 2). The variables taken into account
during the addiction treatment period were hospital admission, status
of the bene?ciary (insured person or their dependent), health insurance
coverage for a chronic disease (100% reimbursement of health care for
the disease in question), complementary private insurance, universal
coverage (coverage for the unemployed and low income insurees), state
medical aid (coverage for foreigners not legally resident in France), and
pregnancy (identi?ed through maternity bene?t).
2.4. Statistical analysis
Qualitative variables were expressed in numbers and percentages
and compared between the drug test group (patients prescribed at
J. Dupouy et al. / Journal of Substance Abuse Treatment 44 (2013) 515?521
Fig. 1. Flow diagram of patient inclusion.
least one drug test) and the control group (patients with no drug test
prescription) using the chi-square test or Fisher test. Quantitative
variables were expressed as means and standard deviations, and were
compared between the two groups using the Wilcoxon rank test. The
primary outcome measure in both groups was survival analysis by
Kaplan-Meier curves. A Cox proportional hazards model was
constructed to assess treatment retention in the two groups. The
Cox model tested the different variables and adjusted for potential
confounding factors. The alpha risk threshold was set at 0.20 for
selection of variables entered in the multivariate model. The
de?nition of time-dependent exposure was tested using the Cox
model in univariate and multivariate analysis (backward procedure,
a = 5%). Data analysis was carried out using SAS 9.2 software (SAS
Inst., Cary, North Carolina, USA).
3. Results
The data collected concerned 1554 patients. Thirty patients were
excluded because of missing demographic data (Fig. 1). Following a
check by the medical of?cer of the health insurance system, 14
patients aged less than 16 years were excluded due to mistaken
attribution as their parents were the consumers. Lastly, three patients
did not meet the inclusion criteria as they had been reimbursed for an
OS before July 1, 2009. Finally, 1507 patients were analyzed.
In the cohort of 1507 patients, 39 (2.6%) had at least one
reimbursement for a drug test during the addiction treatment
period (mean 1.2, SD 0.4). These tests (68 tests) consisted of
measurement of analgesics or narcotics in a biological ?uid other
than blood (38 tests, 55.9%), testing for benzodiazepines in a
biological ?uid other than blood (12 tests, 17.6%), testing and
measurement in a biological ?uid other than blood of a psychotro-
517
pic agent not otherwise categorized (8 tests, 11.8%), measurement
of analgesics or narcotics in blood (7 tests, 10.3%), testing and
measurement of a psychotropic drug in blood (2 tests, 2.9%) and
testing for benzodiazepines in blood (1 test, 1.5%).
Regarding the physicians of the patients in the drug test group, 137
had prescribed an OS and 41 of these physicians had started OST in
these patients. Thirty physicians (25 PCPs, 5 psychiatrists) had
prescribed drug tests.
Regarding PCPs of the Midi-Pyrenees region, 841 had prescribed
OS. A total of 2620 PCPs were shown in our database to have
prescribed a drug or carried out a medical act.
The social and demographic characteristics of the patients of the
whole cohort, the drug test group and the control group are given in
Table 1. Of the 384 women of the cohort, 14 (3.7%) were pregnant during
follow-up (all in the control group, p=1.00). During follow-up, 37 (2.5%)
subjects died (none in the drug test group, p=0.62). Two hundred forty
?ve subjects were considered as lost to follow-up: 234 (15.9%) in the
control group and 11 (28.2%) in the drug test group (p=0.041).
Type of OS, durations of treatment retention, retention rates,
characteristics of primary prescriber of an OS and doctor-shopping
indicators are given in Table 2. Three hundred twenty-four patients
(21.5%) exhibited doctor-shopping behavior: 19 (48.7%) in the drug
test group and 305 (20.8%) in the control group (p b 0.001).
Reimbursements of medications according to class are given in
Table 3. Few hospital admissions were observed during the addiction
treatment period (mean 2.1, SD 9.3) with no difference between the
two groups (p = 0.25).
Overall, 1252 (83.1%) patients discontinued OST: 1224 (83.4%) in
the control group and 28 (71.8%) in the drug test group. Kaplan?Meier
curves revealed differences in OST retention between the two groups
(Fig. 2). This difference was signi?cant in the log rank test (p b 0.001).
In univariate analysis, the variables associated with treatment
retention included age, doctor-shopping behavior, drug tests, one or
more hospital admissions, zolpidem reimbursements, morphine
sulfate reimbursements, antidepressants reimbursements, alcohol
abstinence drugs reimbursements, the type of OS prescribed (buprenorphine or methadone), the primary prescriber speciality and
primary prescriber previously known to patient. Table 4 shows the
results of multivariate analysis in a Cox model. In the multivariate Cox
model, a drug test was independently associated with OST retention
with an HR of 0.55 (95% CI, 0.38?0.80) (p = 0.002).
4. Discussion
Of the 1507 patients who had started OST, only 39 had been
reimbursed for a drug screening test during the medical management
Table 1
Demographic characteristics and type of health insurance cover in the whole cohort, the drug test group and the control group.
All patients
(N = 1507)
Gender n (%)
Male
Female
Mean age (SD) in years
Status of bene?ciary (dependent versus insured) n (%)
Health insurance coverage for a chronic disease n (%)
Complementary insurance n (%)
No
Yes
Universal coverage
Not speci?ed
State medical aid n (%)
No
Yes
Not speci?ed
Drug test group
(n = 39)
Control group
(n = 1468)
p value
1123
384
33.2
73
282
(74.5)
(25.5)
(9.3)
(4.8)
(18.7)
29
10
31.5
0
5
(74.4)
(25.6)
(8.9)
(0.0)
(12.8)
1094
374
33.3
73
277
(74.5)
(25.5)
(9.3)
(5.0)
(18.9)
0.98
43
364
605
495
(2.9)
(24.2)
(40.1)
(32.8)
0
9
14
16
(0.0)
(23.1)
(35.9)
(41.0)
43
355
591
479
(2.9)
(24.2)
(40.3)
(32.6)
0.68
1460
18
29
(96.9)
(1.2)
(1.9)
39
0
0
(100.0)
(0.0)
(0.0)
1421
18
29
(96.8)
(1.2)
(2.0)
1.00
0.182
0.26
0.34
518
J. Dupouy et al. / Journal of Substance Abuse Treatment 44 (2013) 515?521
Table 2
Characteristics of opioid substitution treatment (OST) in the whole cohort, the drug test group and the control group.
All patients
(N = 1507)
OST n (%)
Buprenorphine
Methadone
Buprenorphine and/or methadone
Mean duration of treatment retention (SD)
Retention rate at 6 months n (%)
Retention rate at 12 months n (%)
Retention rate at 18 months n (%)
Primary prescriber of an OS
Specialty n (%)
Primary care physician
Specialist
Not known
Hospital physician n (%)
Physician previously known to patient n (%)
No
Yes
Doctor-shopping indicators
Mean dose delivered (SD)a
Mean dose prescribed (SD)a
Mean doctor-shopping dose (SD)a
Mean doctor-shopping indicator (SD)
Drug test group
(n = 39)
Control group
(n = 1468)
p value
1053
344
110
212
547
321
224
(69.9)
(22.8)
(7.3)
(241)
(36.3)
(21.3)
(14.9)
23
10
6
411
27
21
16
(59.0)
(25.6)
(15.4)
(267)
(69.2)
(53.9)
(41.0)
1030
334
104
207
520
300
208
(70.2)
(22.8)
(7.1)
(238)
(35.4)
(20.4)
(14.2)
1422
75
10
285
(94.3)
(5.0)
(0.7)
(18.9)
38
1
0
7
(97.4)
(2.6)
(0.0)
(18.0)
1384
74
10
278
(94.3)
(5.0)
(0.7)
(18.9)
1190
317
(79.0)
(21.0)
30
9
(76.9)
(23.1)
1160
308
(79.0)
(21.0)
0.75
(1.9)
(1.8)
(0.3)
(0.09)
b 0.001
b 0.001
b 0.001
b 0.001
1.6
1.5
0.1
0.04
(1.8)
(1.8)
(0.3)
(0.09)
2.0
1.8
0.1
0.05
(1.0)
(0.9)
(0.3)
(0.12)
1.5
1.5
0.1
0.03
0.089
b 0.001
b 0.001
b 0.001
b 0.001
0.78
0.88
a
In de?ned daily dose (DDD), dose delivered = daily dose delivered to the patient, dose prescribed = daily dose that would have been delivered to the patient if he/she had only
one physician; doctor-shopping dose = daily dose obtained by the prescriptions of several physicians in a same period of time, doctor-shopping indicator = doctor-shopping dose/
dose delivered.
interruption of treatment so that the patient can again be considered
as starting treatment. Nevertheless, this raises the problem of the ?rst
delivery of methadone, which must be prescribed in a specialized
center or in a hospital setting. Data concerning drugs delivered in
hospital settings are not available in the database. For some subjects
receiving methadone, the duration of treatment retention has in fact
probably been underestimated.
We studied the in?uence on therapeutic management of the use of
drug tests, whether in blood or urine, in the physician’s of?ce or in the
laboratory. These tests can be used in various clinical situations, but in
our selected population of patients starting OST and during the
de?ned period of medical addiction treatment, we may assume that
they were prescribed in connection with the management of opiate
addiction. As PCPs seldom carry out such tests (Dupouy et al., 2012),
we may consider that differences in practice do not limit the validity
of their addiction. Their treatment retention was signi?cantly longer.
The association between drug tests and OST retention was con?rmed
by multivariate analysis.
This observational study is based on the data of the French health
insurance system. Use of such databases has become generalized in
France for more than 10 years, particularly in the ?eld of addiction
(Lapeyre-Mestre et al., 2003; Thirion et al., 2002), and the validity of the
ERASME database of the health insurance system has been con?rmed
(Latry, Molimard, B?gaud, & Martin-Latry, 2010). It has been shown in
the Three-Cities (3C) cohort that the reimbursement data of the health
insurance system agreed with consumption data (Noize et al., 2009).
Incident status was de?ned as non-reimbursement for an OS
during the ?rst 6 months of the study period. It can be assumed that
the absence of a prescription for an OS over a 6-month period
indicates that there was no previous treatment, or at least a prolonged
Table 3
Drugs reimbursed for patients in the whole cohort, the drug test group and the control group.
All patients
(N = 1507)
Class of drugs reimbursed
Cardiovascular system n (%)
Platelet antiaggregants n (%)
Lipid-lowering agents n (%)
Statins n (%)
Hepatitis C treatment n (%)
HIV treatment n (%)
Alcohol abstinence drugs n (%)
Psychotropic agents n (%)
Antipsychotics n (%)
Benzodiazepines n (%)
Antidepressants n (%)
Potential drugs of abuse
Flunitrazepam n (%)
Clonazepam n (%)
Diazepam n (%)
Bromazepam n (%)
Oxazepam n (%)
Alprazolam n (%)
Zolpidem n (%)
Methylphenidate n (%)
Morphine sulphate n (%)
Drug test group
(n = 39)
Control group
(n = 1468)
p value
232
37
45
36
30
23
112
1058
391
1031
536
(15.4)
(2.5)
(3.0)
(2.4)
(2.0)
(1.5)
(7.4)
(70.2)
(26.0)
(68.4)
(35.6)
8
1
1
1
0
0
5
29
18
29
23
(20.5)
(2.6)
(2.6)
(2.6)
(0.0)
(0.0)
(12.8)
(74.4)
(46.2)
(74.4)
(59.0)
224
36
44
35
30
23
107
1029
373
1002
513
(15.3)
(2.5)
(3.0)
(2.6)
(2.0)
(1.6)
(7.3)
(70.1)
(25.4)
(68.3)
(35.0)
0.37
1.00
1.00
0.62
1.00
1.00
0.21
0.57
0.004
0.42
0.002
23
145
236
305
319
215
362
1
80
(1.5)
(9.6)
(15.7)
(20.2)
(21.2)
(14.3)
(24.0)
(0.1)
(5.3)
1
8
9
6
15
5
13
0
2
(2.6)
(20.5)
(23.1)
(15.4)
(38.5)
(12.8)
(33.3)
(0.0)
(5.1)
22
137
227
299
304
210
349
1
78
(1.5)
(9.3)
(15.5)
(20.4)
(20.7)
(14.3)
(23.8)
(0.1)
(5.1)
0.46
0.047
0.20
0.45
0.007
0.79
0.17
1.00
1.00
J. Dupouy et al. / Journal of Substance Abuse Treatment 44 (2013) 515?521
Patients at risk, n
Drug test group
Control group
39
36
28
25
22
18
17
12
6
3
1
1,468
738
520
390
301
242
209
142
97
38
3
Fig. 2. Kaplan?Meier curves comparing opioid substitution treatment (OST) retention
in the two groups (drug test group and control group), log rank test p b 0.001.
of the drug test codes selected. This was con?rmed by our ?ndings,
since only six laboratory test codes were in fact recorded out of the
eight codes initially selected. Also, the majority of codes corresponded
to tests in a biological ?uid other than blood, probably urine tests.
Adding 30 days before the ?rst delivery of the ?rst OS to de?ne the
addiction treatment period was justi?ed because a drug test could be
prescribed earlier and could, therefore, in?uence patient behavior.
This study is subject to several sources of bias. Af?liation to the
main French health insurance system is dependent on occupation;
part of the population, with a different social and economic pro?le, is
not included in our analysis. This leads to a selection bias. This bias
was limited by the fact that we were able to access the data of student
health insurance organizations and by the fact that persons in
dif?culty were included.
Some patients possibly had drug tests in a specialized center or a
hospital. Nevertheless, data concerning tests performed in hospital
settings are not available in the database. Furthermore, some patients
possibly had urine tests with commercial strips or kits. As these tests
are not covered by the health insurance system, we had no trace of
them and some patients were probably wrongly included in the
control group, leading to underestimation of the hazard risk of drug
screening. Use of a time-dependent variable to study exposure to a
drug screening test avoided immortal time bias.
Only a small number of subjects were tested by physicians. This
highlights that physicians seldom carry out such tests, which is in
Table 4
Analysis of opioid substitution treatment retention using the multivariate Cox model.a
Adjusted hazard ratio (95% CI) p value
Drug test
0.55 (0.38?0.80)
Doctor-shopping behavior
0.36 (0.31?0.41)
Primary prescriber specialist versus PCP 0.69 (0.53?0.90)
Age (years)
25?30 (versus b 25)
0.88 (0.75?1.05)
30?40 (versus b 25)
0.78 (0.67?0.91)
N 40 (versus b 25)
0.75 (0.63?0.89)
1 or more hospital admission
1.14 (1.01?1.28)
Morphine sulfate
1.35 (1.06?1.71)
Alcohol abstinence drugs
1.38 (1.11?1.70)
0.002
b 0.001
0.007
0.153
0.002
b 0.001
0.036
0.014
0.004
PCP = primary care physician.
a
Variables initially included in the model were age, gender, status of bene?ciary,
health insurance coverage for a chronic disease, universal coverage, complementary
insurance, state medical aid, pregnancy, being hospitalized at least once, OST, specialty
of primary prescriber, primary prescriber being an hospital physician, primary
prescriber previously known to patient, dose delivered, doctor-shopping behavior,
other drugs reimbursed and potential drugs of abuse reimbursed.
519
agreement with declarative data (Dupouy et al., 2012) but raises
concerns about selection of patients tested. Two different hypotheses
with opposite effects can be imagined. First, patients tested may be
more compliant patients and may be self-selected on their motivation.
However, this hypothesis is not in agreement with exposure of tested
patients to certain drugs of abuse (benzodiazepines in particular) and
doctor-shopping behavior which was more frequent in tested
patients. The second hypothesis is that tested patients may be more
heavily addicted patients, as physicians need to test them to assess
their consumption.
Lastly, residual confounding factors are another bias in this work.
Personal history, addiction severity, parallel drug consumption
(Duburcq, Charpak, Blin, & Madec, 2000), injection pro?le, family
support and the occupational (Stein, Cioe, & Friedmann, 2005), and
social context (Batel et al., 2004) are variables that have a strong
impact on OST retention in these patients, but these factors are not
available in the database. The population that underwent drug tests
was possibly more severely addicted, as suggested by exposure to
certain drugs of abuse.
Demographic data of included patients were similar to demographic data of a French cohort in the same area (Lapeyre-Mestre
et al., 2003). The primary care physicians who had issued a
prescription or performed any medical act in this patient population
represented 68.7% of PCPs in regular practice in the Midi-Pyrenees
region on January 1, 2010 according to the data of the national order of
physicians, the Ordre National des M?decins (Conseil National de
l’Ordre des M?decins, 2010). This high proportion suggests that our
data are exhaustive. Of these PCPs, 32.1% prescribed OST. This is in
agreement with the declarative data we collected during a previous
survey in the same area (Dupouy et al., 2012). Of the physicians
treating patients in the drug test group, 30 had prescribed a test. This
indicates that the differences between the two groups were not due
only to differences in practice or training of some physicians.
Drug tests appeared to be associated with better OST retention.
This can be explained by better assessment of drug consumption and
easier dialogue between patient and practitioner. Drug tests are
known to be effective in school programmes (Roche et al., 2009) and
in screening of occupational drivers (Cashman et al., 2009). In an
emergency setting, a recent review found no in?uence of drug tests on
therapeutic management (Tenenbein, 2009). The effectiveness of
urine testing to reduce opioid misuse in chronic non-cancer pain
patients is still debated (Starrels et al., 2010).
Doctor-shopping appeared to be independently associated with
better OST retention. This may simply be explained by the fact that
doctor-shopping results in bias in the measurement of treatment
retention time. Patients who see several prescribers are more likely to
have a shorter time between two prescription deliveries and so to remain
in treatment longer, leading to a non-differential information bias.
In conclusion, our study increased our knowledge of the value of
drug tests in treatment of opioid addiction in an outpatient setting. In
a cohort of 1507 patients starting OS treatment and followed for 18 to
30 months, only 39 (2.6%) had at least one reimbursement for a drug
test. Treatment retention was longer in these patients, after taking the
available confounding factors into account. These ?ndings deserve to
be con?rmed by more detailed study.
Acknowledgments
The authors would like to thank the Acad?mie Nationale de
M?decine (French National Academy of Medicine) for funding the
second year of a postgraduate master’s degree by the ?rst author
(J.D.). No other funding (direct or indirect) was received. The authors
would also like to thank the French Health Insurance system for
enabling them to access data, and Nina Crowte for her help in
correcting the English text.
520
J. Dupouy et al. / Journal of Substance Abuse Treatment 44 (2013) 515?521
Appendix Table 1
References
Medical laboratory classi?cation codes used to identify drug screening tests in the
French health insurance system database.
Medical
laboratory code
Label of corresponding analysis
1659
Measurement in blood of analgesics or narcotics not otherwise
categorized
Measurement in a biological ?uid other than blood