Available via license: CC BY-NC-ND 4.0
Content may be subject to copyright.
Contents lists available at ScienceDirect
International Journal of Drug Policy
journal homepage: www.elsevier.com/locate/drugpo
Research Paper
An assessment of the limits of detection, sensitivity and specificity of three
devices for public health-based drug checking of fentanyl in street-acquired
samples
Traci C. Green
a,b,⁎
, Ju Nyeong Park
c
, Michael Gilbert
d
, Michelle McKenzie
e
, Eric Struth
f
,
Rachel Lucas
g
, William Clarke
h
, Susan G. Sherman
c
a
Department of Emergency Medicine, The Warren Alpert School of Medicine of Brown University, Rhode Island Hospital, USA
b
Department of Epidemiology, Brown University School of Public Health, 55 Claverick St., 2nd floor, Providence, RI 02903, USA
c
Department of Health, Behavior, and Society, Johns Hopkins Bloomberg School of Public Health, Baltimore, MD, USA
d
Independent, Portland, OR, USA
e
Department of Immunology, The Miriam Hospital, Center for Prisoner Health and Human Rights, Providence, RI, USA
f
Department of Emergency Medicine, Rhode Island Hospital, 55 Claverick St., 2nd floor, Providence, RI 02903 USA
g
Baltimore Police Department Forensic Laboratory, Baltimore, MD, USA
h
Department of Pathology, Johns Hopkins Bloomberg School of Medicine, Baltimore, MD, USA
ARTICLE INFO
Keywords:
Fentanyl
Drug checking
Overdose
Public health
Surveillance
Harm reduction
ABSTRACT
Background: Fentanyl has caused rapid increases in US and Canadian overdose deaths, yet its presence in illicit
drugs is often unknown to consumers. This study examined the validity in identifying the presence of fentanyl of
three portable devices that could be used in providing drug checking services and drug supply surveillance:
fentanyl test strips, a hand-held Raman Spectrometer, and a desktop Fourier-Transform Infrared Spectrometer.
Methods: In Fall 2017, we first undertook an assessment of the limits of detection for fentanyl, then tested the
three devices’sensitivity and specificity in distinguishing fentanyl in street-acquired drug samples. Utilizing test
replicates of standard fentanyl reference material over a range of increasingly lower concentrations, we de-
termined the lowest concentration reliably detected. To establish the sensitivity and specificity for fentanyl, 210
samples (106 fentanyl-positive, 104 fentanyl-negative) previously submitted by law enforcement entities to
forensic laboratories in Baltimore, Maryland, and Providence, Rhode Island, were tested using the devices. All
sample testing followed parallel and standardized protocols in the two labs.
Results: The lowest limit of detection (0.100 mcg/mL), false negative (3.7%), and false positive rate (9.6%) was
found for fentanyl test strips, which also correctly detected two fentanyl analogs (acetyl fentanyl and furanyl
fentanyl) alone or in the presence of another drug, in both powder and pill forms. While less sensitive and
specific for fentanyl, the other devices conveyed additional relevant information including the percentage of
fentanyl and presence of cutting agents and other drugs.
Conclusion: Devices for fentanyl drug checking are available and valid. Drug checking services and drug supply
surveillance should be considered and researched as part of public health responses to the opioid overdose crisis.
Introduction
Fentanyl-related overdoses have rapidly become a major public
health crisis in many communities in the United States (U.S.)
(O'Donnell, Halpin, Mattson, Goldberger & Gladden, 2017;Rudd, Seth,
David & Scholl, 2016) and Canada (Fischer, Murphy, Rudzinski &
MacPherson, 2016). Fentanyl is a synthetic opioid that is estimated to
be 100 times more potent than morphine (Centers for Disease Control &
Prevention, 2017). It is pharmaceutically manufactured for clinical use
in anesthesia and pain management, but can also be produced using
illegal clandestine methods (U.S. Drug Enforcement Agency, 2018). The
challenge includes not only fentanyl per se, but also many analogs (e.g.,
furanyl fentanyl, carfentanil) and fentanyl-like chemicals (e.g.,
U47700) linked to overdose outbreaks (O'Donnell et al., 2017).
https://doi.org/10.1016/j.drugpo.2020.102661
Abbreviations: PWUD, People who use drugs; FTIR, Fourier-transform infrared; FTS, fentanyl test strips; GC/MS, gas chromatography and mass spectrometry
⁎
Corresponding author at: Department of Epidemiology, Brown University School of Public Health, 55 Claverick St., 2nd floor, Providence, RI 02903, USA.
E-mail addresses: traci.c.green@brown.edu (T.C. Green), ju.park@jhu.edu (J.N. Park), mmckenzie@lifespan.org (M. McKenzie),
rachel.lucas@baltimorepolice.org (R. Lucas), wclarke@jhmi.edu (W. Clarke), ssherman@jhu.edu (S.G. Sherman).
International Journal of Drug Policy 77 (2020) 102661
0955-3959/ © 2020 Published by Elsevier B.V.
T
Every year, thousands of Americans die from fentanyl-related
overdose, with the highest rates concentrated in the eastern U.S.
(Sanger-Katz, 2018). In 2017, 1594 overdose deaths in the state of
Maryland involved illegally manufactured fentanyl, compared with 413
deaths involving prescription opioids, and in Baltimore City, fentanyl-
related overdose deaths increased by 37% from 2016 to 2017
(Maryland Department of Health, 2018). Fentanyl was detected in 64%
of all overdose deaths in Rhode Island (RI) in 2017 and in 85% of opioid
deaths in Massachusetts (Marshall et al., 2017;
Massachusetts Department of Public Health, 2018;Rhode Island
Department of Health, 2018). The dose and type of fentanyl in street
heroin and other drugs is often unknown, placing individuals at a high
risk of unintentional overdose. The U.S. Drug Enforcement Agency as-
serts that fentanyl doses as low as two milligrams can be lethal, and also
report that seized fentanyl samples under one kilogram were less than
three percent pure (U.S. Drug Enforcement Administration, 2019).
While efforts are underway to address the fentanyl crisis with evidence-
based approaches like expanding access to naloxone and medications
for opioid use disorder, other preventative measures are needed to
minimize harm among people who use drugs (PWUD) at risk of fentanyl
exposure.
Drug checking is a public health intervention that introduces the
concept of product safety into the unregulated illicit drug supply. Drug
checking services permit people to submit drug samples for chemical
analysis. The results of these analyses are traditionally shared with
submitters prior to use, facilitating informed decisions about drug use
behaviors (Brunt et al., 2016;Harper, Powell & Pijl, 2017). While this
strategy is an established harm reduction tool in many European
countries (Brunt et al., 2016), it is a new endeavor in the U.S. and
Canada. In the U.S., drug checking has largely been limited to on-site,
reagent-based, colorimetric testing at cultural events and off-site, mail-
based gas chromatography and mass spectrometry (GC/MS) analyses,
with testing focusing on putative MDMA products.
Recently, harm reduction programs in the U.S. and Canada have
begun distributing immunoassay fentanyl test strips (FTS), with the
intent that participants will use them prior to drug use (Amlani et al.,
2015;Tupper, McCrae, Garber, Lysyshyn & Wood, 2018). Specifically,
objective determination of whether drug samples contain fentanyl or
fentanyl analogs can help mitigate consumers’risk of overdose and
extend safety interventions (Barratt, Bruno, Ezard & Ritter, 2018;
Goldman et al., 2019;Hungerbuehler, Buecheli & Schaub, 2011;
Krieger et al., 2018;Sande & Sabic, 2018;Sherman et al., 2019). A
recent study in North Carolina found substantial changes in overdose
safety and drug use behaviors following FTS utilization (Peiper et al.,
2019). A Canadian study found that people selling and supplying drugs
were more likely to use drug checking services (Kennedy et al., 2018),
but structurally vulnerable populations may be less likely to partake in
more extensive drug checking services, even when offered at low-bar-
rier service sites like a safe consumption site (Bardwell, Boyd, Tupper &
Kerr, 2019). Data from the FORECAST Study, which surveyed PWUD in
three U.S. cities about reactions to and experiences with fentanyl in the
drug supply, found that many do not prefer drugs containing fentanyl
(Sherman et al., 2019) and 39% employ a number of practices to reduce
overdose risk in a context of unknown drug purity and content
(Rouhani, Park, Morales, Green & Sherman, 2019), suggesting pro-
spects for disseminating drug checking results and harm reducing
messages. Regardless of the device used, drug checking alters the in-
formation asymmetry of a risky and unpredictable drug market by
equipping PWUD with actionable information to protect themselves
and others around them from drug supply related harm.
Furthermore, data gleaned from drug checking can provide in-
formation about the illicit supply and drug market trends, including
purity, contaminants, and cutting agents, as demonstrated by the 6-
country Trans European Drug Information project (Brunt et al., 2016).
Aggregated data about the local drug supply can be used to better in-
form public health responses and warning campaigns (Laing, Tupper &
Fairbairn, 2018). However, data on the accuracy of field-testing devices
for fentanyl, especially when applied to “real world”street-acquired
drug samples in quantities packaged for individual consumption, are
lacking, hindering broader scale-up of field test application for public
health benefit and overdose prevention.
The goal of the current study was to test the validity of three por-
table drug checking devices in detecting fentanyl: FTS; a hand-held
Raman spectrometer; and a portable Fourier-transform infrared (FTIR)
spectrometer.Specifically, we aimed to estimate limits of detection,
sensitivity (the device's ability to detect true positives), and specificity
(the device's ability to detect true negatives) for fentanyl, compared to
an established GC/MS confirmatory method.
Methods
We first undertook an assessment of limits of detection for fentanyl,
then tested the three devices’ability to detect fentanyl in street drug
samples. The first phase entailed using chemical standard reference
material and the second phase required collaboration with police de-
partments to use seized drug samples that were packaged for end-user
consumption, confiscated during arrests, and previously submitted for
forensic testing. Testing took place in September and October 2017 at
the Baltimore Police Department (BPD) Forensic Laboratory in
Baltimore, Maryland, and the Drug Chemistry Laboratory at the RI
Department of Health State Health Laboratories in Providence, RI.
Forensic scientists from the BPD laboratory ran all tests per protocol in
Baltimore; in Providence, all testing was completed by contracted
chemists. The team held teleconferences on a weekly basis for con-
sistency checks and to clarify any instrument or protocol questions that
arose (e.g., library or instrument updates, deciphering error messages).
Devices tested
There were three devices tested, none of which were designed for
public health applications but all of which have potential for mobile or
point-of-care drug checking Harper et al., 2017). The first device tested
was a fentanyl test strip (Rapid Response™, BTNX Corporation, Canada)
which uses an immunoassay on a paper strip to provide a rapid (within
five minutes) indication of the absence or presence of fentanyl and
several analogs: norfentanyl, acetyl fentanyl, carfentanyl, furanyl fen-
tanyl, butyryl fentanyl, valeryl fentanyl, ocfentanil, 3-methyl fentanyl,
remifentanil, sufentanil, and p-fluoro fentanyl
(BTNX corporation, 2019). The FTS do not measure fentanyl con-
centration and do not provide results that differentiate between or
among fentanyl and any of the analogs. The device only measures
fentanyl presence and is approved in Canada, but not in the U.S., for in
vitro diagnostic testing in urine. Use of the urine strips for drug
checking is considered off-label, as the FTS are designed and approved
for in vitro use only. This product was used in several pilot drug
checking projects (Krieger et al., 2018;Peiper et al., 2019;Sande &
Sabic, 2018;Tupper et al., 2018). The size and portability of FTS sup-
ports home use and broad distribution.
The procedure for fentanyl testing involves creating a drug solution
by dissolving a small amount of drug in water. Since drugs may be used
by multiple routes of administration, we employed this simple testing
procedure, though testing of drug remnants in once-used drug pre-
paration materials (i.e., cookers, cottons used in injection preparation)
could proceed in similar fashion. The package insert-reported sensi-
tivity and specificity for detecting fentanyl in urine are 96.8% and
100%, respectively. These single-use strips are low-cost and easy to use
Harper et al., 2017).
The second device was a Raman spectrometer (TruNarc™,
ThermoFisher Scientific, Waltham, MA), a hand-held instrument used
by some law enforcement agencies for presumptive field testing of
narcotics (Fig. 1). For lower concentration samples and for suspected
heroin and fentanyl samples, a two-step process is encouraged. Step one
T.C. Green, et al. International Journal of Drug Policy 77 (2020) 102661
2
involves a non-contact, direct scan (“Point-and-Shoot”) of the drug
sample and step two involves dissolving and drying drug samples on a
specialized stick (Surface Enhanced Raman Spectroscopy or “SERS
kit”), then scanning the stick. The device can simultaneously detect up
to two substances, drawing on a library of about 300 drugs. It provides
only presence of a substance in a sample. The time to detect the sub-
stance ranges from 30 s to five minutes, depending on the number of
steps above applied. The lowest limit of detection for any substance, as
reported by the manufacturer to our research team, is 2% weight. To
our knowledge, this device is not currently in use by community, public
health, or harm reduction programs.
The third device tested was a Fourier-transform infrared
(FTIR) spectrometer device (Bruker Alpha™, Bruker Optics, Billerica,
Massachusetts), which uses infrared light to scan test samples and ob-
serve chemical properties (Fig. 1). The FTIR can provide information on
both the presence and amount of a substance in a sample within sec-
onds but requires some handling to ready a powder or pill for scanning.
Post-scan, results can be analyzed for main components and mixtures,
providing more information about the substance. Reference libraries
can be created and numerous others are available for subscription
against which samples could be tested. For the current study, 11 li-
braries specific to DEA controlled substances and cutting agents were
used in this analysis. The Bruker Alpha™is commonly used in phar-
maceutical supply chain management and material commodities certi-
fication, and is being used for drug checking in the United Kingdom by
the drug checking project ‘The Loop’, by the Drug Information and
Monitoring System (DIMS) in the Netherlands, as well as by British
Columbia's drug checking program (Bardwell & Kerr, 2018;Brunt et al.,
2016;Tupper et al., 2018;Yau et al., 2019). This device was tested only
at the Providence, Rhode Island site. According to the manufacturer
(personal communication, Bruker Optics, Billerica, Massachusetts), the
exact limit of detection values depends on the analyte or mixture of
interest, but it has capacity to detect and quantify materials at the pi-
cogram level. This FTIR is comparable in price to the Raman spectro-
meter (~$20,000) and requires a computer to analyze and interpret
scanned results.
Limits of detection procedures
Powder fentanyl standards were obtained from DEA registered
vendors (Lipomed, Inc., Cambridge, Massachusetts; Cerilliant
Corporation, Round Rock, Texas), and prepared in solutions of ethanol,
as is commonly done for known and questioned specimens
(Clarke's Analytic Forensic Toxicology, Second Edition, 2013). Solu-
tions of varying concentrations were prepared, based on the range of
known or estimated detection of the devices.
Starting at concentrations approximately 10 times that of the ven-
dor's estimated limit of detection, we sequentially tested samples of
known concentration of fentanyl in triplicate for detection by each
device and decreasing the value of the known concentration by serial
dilution. If the device produced a response at a given concentration, the
value was recorded. The concentration was incrementally reduced,
until the compound was undetectable. Then, we increased the con-
centration to the last detectable concentration and reduced the con-
centration by 25%. Once we found the estimated limit of detection for
each device, we tested several additional samples (n= 7) at this con-
centration to ensure its replicability within ± 25%. For the purposes of
analysis of the limit of detection, an ‘unclear-indeterminate’observa-
tion was considered ‘not-detected’. The actual detection limit was
considered the lowest concentration at which fentanyl was detected.
Each device was run according to the manufacturer's instructions
and the respective company trainer's guidance. For the GC/MS devices,
we ensured that all protocols, method parameters, and the minimum
criterion for acceptability were harmonized so that the results were
comparable and able to optimize similar determinations (i.e., absence,
presence) of the drug samples.
Sensitivity and specificity procedures
We also tested actual drug samples from Baltimore and Providence,
to determine the sensitivity and specificity of each testing approach in
detecting fentanyl from drug samples obtained by police pursuant to
criminal investigations. Individuals who were not involved with the
laboratory protocol execution identified 106 fentanyl-positive samples
and 104 fentanyl-negative samples (as previously determined by GC/
MS), resulting in N= 210 samples overall. The selected fentanyl po-
sitive samples were either fentanyl only or contained fentanyl in com-
bination with other drugs. Samples eligible for testing included powder
and pill samples submitted by the study partner police departments i.e.,
BPD and Providence Police Department, from cases representing per-
sonal quantities. This included exhibits from cases with ≤1 g of drug,
small-scale (<50 bags) distribution, or, if charged with manufacturing
with intent to distribute and there were over 50 bags prepared for sale,
then only one sample from any exhibit could be considered for inclu-
sion in the testing protocol. Samples always remained securely stored at
the forensic labs.
Testing of the samples occurred using each device, according to the
package inserts or manufacturer's training, by chemists blinded to the
sample's true content. Both the FTIR and Raman spectrometer trainings
were conducted in person by the company's respective trainers to en-
sure consistency and lasted about 1.5 h each. A 30-min training for the
FTS was conducted by the study team. Drug samples used for fentanyl
strip testing were dissolved in tap water in Baltimore and in deionized
water in Providence.
Scanning with the Raman spectrometer was completed by the lab
staff. Prior to scanning the sample, a machine self-check was performed
and passed. On all samples, results of up to three “Point-and-Shoot”
scans were recorded, followed by use of the SERS kit with three scans
performed and recorded. Approximately 30-s per sample scans of each
sample with the FTIR were completed in the Providence lab; all blinded
Rapid Response™ fentanyl
testing strips (FTS), BTNX
Laboratories, Canada
TruNarc™ Raman
spectrometer, Thermofisher
Scientific, Waltham, MA
USA
Bruker Alpha™ Fourier-
Transform Infrared (FTIR),
Bruker Optics, Billerica, MA
USA
Fig. 1. Devices examined for limit of detection,
sensitivity, and specificity of fentanyl, Baltimore
Police Department (BPD) Forensic Laboratory and
Drug Chemistry Laboratory at the Rhode Island
Department of Health State Health Laboratories,
September to October 2017
FTS=fentanyl testing strips
FTIR= Fourier-Transform Infrared.
T.C. Green, et al. International Journal of Drug Policy 77 (2020) 102661
3
analyses were performed by TG, who received training from the man-
ufacturer's analytic team. The steps for identification of fentanyl's dis-
tinct patterns were applied to all analyses.
Results were entered into a secure computer database. GC/MS re-
sults for the samples that had already been run were extracted from the
respective laboratory databases. Sensitivity and specificity were calcu-
lated for each of the field test devices, using GC/MS results as the es-
tablished confirmatory method comparator. Since both of the spectro-
meters have the capacity to detect other controlled substances and
cutting agents, sensitivity and specificity for all substances including
fentanyl and for non-fentanyl containing samples were also calculated.
An exploratory analysis examined the ability of the device to detect
fentanyl analogs. Samples (n= 4) from Baltimore and Providence
Police Departments containing any type of fentanyl analog alone or in
combination with drugs other than fentanyl were eligible for inclusion
in the sample subset.
Results
Limit of detection
Table 1 shows findings on the limits of detection. The GC/MS limit
of detection was instrument and method specific: 3.1 mcg/mL in Pro-
vidence and 100 mcg/mL in Baltimore. The FTS had the lowest limit of
detection of the devices tested 0.100 mcg/mL), with both sites gen-
erating similar results. The Raman spectrometer's limit of detection for
fentanyl was 25 mcg/mL, or 250 times higher than the FTS. The FTIR
detected samples according to weight, indicating that, for a given
powder or pill tested containing fentanyl, the smallest detectable
amount for fentanyl is 3–4% weight.
Characteristics of seized drug samples
Table 2 presents the composition of the included seized drug sam-
ples. Cutting agents that were not classified as controlled substances by
each laboratory are not reported. In general, the Baltimore samples
contained more heroin and fewer fentanyl-only samples compared to
Providence samples. Fentanyl and cocaine mixture samples were only
found in Providence, as were samples with the fentanyl precursor
chemical 4-ANPP. Regarding non-fentanyl samples, the majority of
samples for both labs were cocaine, heroin, benzodiazepines, or me-
thamphetamine.
Sensitivity and specificity
As can be seen in Table 3, the performance of each technology
compared to GC/MS in detecting fentanyl differed. The FTS had the
highest sensitivity compared to the other devices (100% in Baltimore,
96% in Providence). The two fentanyl samples that failed to be detected
(sensitivity) contained only fentanyl (no other drugs present). The five
samples that falsely detected fentanyl actually contained, according to
GC/MS results, buprenorphine only (n= 1), heroin only (n= 2),
methamphetamine only (n= 1), and noscapine only (n= 1).
Table 1
Limits of detection for fentanyl by method and study forensic laboratory.
Laboratory instrument Baltimore laboratory Rhode Island laboratory
GC/MS 100 mcg/mL 3.1 mcg/mL
Fentanyl test strips (FTS) 0.150 mcg/mL 0.100 mcg/mL
Raman spectrometer*–25 mcg/mL
FTIR spectrometer** N/A 3–4% weight
⁎
Results from the two-step process. The Raman spectrometer was tested in
both labs however due to changes in the software library, results were deemed
reliable only in the Rhode Island data.
Conversions: 1 mg = 1000 mcg = 1000,000 ng.
⁎⁎
Bruker Alpha was tested only in Rhode Island. The limit of detection for
fentanyl powder standard was deduced digitally by the authors and with the
assistance of the Bruker Optics analytic team.
GC/MS= gas chromatograph/mass spectrometer.
FTS=fentanyl test strips.
FTIR= Fourier-Transform Infrared.
Table 2
Number and types of street-acquired drug samples from Baltimore, Maryland
and Providence, Rhode Island, selected for sensitivity and specificity testing for
fentanyl presence.
Baltimore drug
sample
Providence drug
sample
NN
Fentanyl positive result by GC/MS 52 54
Fentanyl 29 35
Fentanyl, Heroin 21 12
Fentanyl, Acetyl Fentanyl 0 1
Fentanyl, Furanyl Fentanyl 1 0
Fentanyl, Furanyl Fentanyl,
Heroin
10
Fentanyl, Cocaine 0 3
Fentanyl, Heroin, Cocaine 0 1
Fentanyl, 4-ANPP 0 2
Fentanyl negative result by GC/MS 52 52
Cocaine 22 25
Heroin 15 9
Oxycodone 6 3
Clonidine 3 0
Alprazolam 2 2
5-Fluro-ADB 1 0
Promethazine 1 0
Suboxone/buprenorphine 1 1
Tramadol 1 0
4-Chloro-Alpha-PPP 0 1
Clonazepam 0 2
Amphetamine 0 1
Dibutylone 0 1
Ephylone 0 2
Methamphetamine 0 3
Noscapine 0 1
Tramadol, haloperidol,
chlorphenamine
01
GC/MS: Gas Chromatograph/Mass Spectrometer.
Table 3
Performance of each device against original gas chromatograph/mass spectro-
meter-reported forensic lab result for detection of fentanyl.
Baltimore
laboratory
Rhode Island
laboratory
N= 104 N= 106
Fentanyl test strips
Sensitivity 100.0% 96.3%
Specificity 98.1% 90.4%
Raman spectrometer –point and
Shoot#
Sensitivity 3.8% 3.7%
Specificity*98.1% 100.0%
Raman spectrometer –SERS kit#
Sensitivity 38.5% 61.1%
Specificity 92.3% 91.5%
FTIR spectrometer
Sensitivity N/A 83.3%
Specificity N/A 90.2%
#
At least one of three readings displaying “fentanyl”,“fentanyl/metham-
phetamine”or “heroin/fentanyl/methamphetamine”were included as posi-
tives. Three inconclusive results in a row coded as fails.
⁎
Note that the FTIR testing was only conducted in the Rhode Island la-
boratory.
SERS= Surface Enhanced Raman Spectroscopy
FTIR= Fourier-Transform Infrared
T.C. Green, et al. International Journal of Drug Policy 77 (2020) 102661
4
The Raman spectrometer “Point-and-Shoot”mode returned the
highest specificity at both labs, but the sensitivity did not reach com-
parable values in either the “Point-and-Shoot”mode or with the SERS
kit. There was also considerable variability between the two labs ob-
served on the SERS kit findings, with 38.5% sensitivity in Baltimore and
61.1% sensitivity in Providence. The sensitivity of detecting all present
substances including fentanyl with the Raman spectrometer in the
“Point-and-Shoot”mode was 25.7%, with a 48.1% sensitivity for non-
fentanyl containing samples. Using the SERS kit, the sensitivity of de-
tecting all present substances including fentanyl improved to 53.8%,
with a 57.7% sensitivity for non-fentanyl containing samples.
The FTIR generated high detection rates of fentanyl and non-fen-
tanyl samples (83.3% sensitivity, 90.2% specificity) for the Providence,
Rhode Island samples. The sensitivity of detecting all present sub-
stances including fentanyl with the FTIR was 81.9%, with an 80.4%
sensitivity for identifying non-fentanyl containing samples.
Exploratory analyses
Four samples, containing acetyl fentanyl alone (n= 2), furanyl
fentanyl alone (n= 1), and furanyl fentanyl in combination with heroin
(n= 1), were eligible for testing. On all samples, the FTS detected the
presence of fentanyl. While the Raman spectrometer “Point-and-Shoot”
returned inconclusive results for all scan sets, scans of the SERS kit
detected fentanyl in three of the four samples (75% sensitivity). For the
FTIR, the libraries available did not initially identify the fentanyl
analog, however, the exploratory sample infrared scans were visually
similar to those of the fentanyl standard scan, allowing detection of
fentanyl in three of the four samples (75% sensitivity). A counterfeit pill
containing furanyl fentanyl failed to be detected with either the Raman
or FTIR spectrometer.
Discussion
This study reports performance-based characteristics of three de-
vices with potential for use in a public health-led drug checking
strategy. It is the first study to examine three viable forensic advances
for detecting fentanyl in samples intended for street-sale and personal
consumption and has direct relevance to public health decision makers.
Findings indicate that, for public health applications aiming to detect
fentanyl in real-world samples, the FTS had high sensitivity and spe-
cificity, and could detect the presence of at least two additional analogs
alone or in the presence of another drug, in both powder and pill forms.
Furthermore, the limit of detection for the FTS was substantially lower
(better) than all other testing modalities, including the GC/MS.
However, the FTS do not indicate any other information on quantity,
purity, presence of adulterants or fill, which may limit their singular
utility in fentanyl-saturated drug markets or when more information is
desired.
The hand-held Raman spectrometer, even when used with a SERS
kit, returned a low sensitivity. These findings are likely based on the
low purity and volume in which fentanyl was found in the selected
seized samples. The much greater utility of the Raman spectrometer
appears to be for field testing samples with higher purity, such as might
be found during trafficking and transport investigations, and more
generally in narcotics investigation work. In our “real-world”drug
sample testing, the Raman spectrometer ruled out fentanyl well in
samples that were shown by the GC/MS to not contain fentanyl (i.e.,
low false positive rate) but missed a substantial portion of fentanyl-
containing drug samples (i.e., high false negative rate). In the absence
of a more sensitive testing step, these characteristics pose serious
challenges to adoption of the Raman spectrometer alone for public
health purposes that propose personal-quantity drug sample testing.
The FTIR demonstrated moderate sensitivity and high specificity,
detecting instances of fentanyl presence despite the low purity of fen-
tanyl contained in the samples tested. The false positive rate was low
relative to the Raman spectrometer, at 9.8%, and a false negative rate of
16.7%. In addition, for comprehensively identifying drug mixtures and
understanding more about the components of street drug samples
containing fentanyl, data gleaned from the FTIR could be useful, and
has potential for public health and drug surveillance applications.
Indeed, a recent study in Vancouver by Tupper et al. (2018) found that
drug samples scanned by FTIR could describe the drug mixtures as well
as adulterants and contaminants that could be harmful to humans.
While the device does not require much training to perform the drug
scan, it does require trained personnel to interpret the scans and an up-
to-date and curated library for substance identification, which may
limit wide-scale uptake. Advantages of both the Raman and FTIR
spectrometers were the speed at which results were received and the
potential amount of information that could be gathered, all qualities
reviewed by Harper et al. (2017)). A 30 to 60 s scan on the FTIR could
reliably depict the spectrum of the substance and could be re-run with
ease. Both machines also required little handling of the sample, thereby
reducing exposure risks. Finally, both machines generate scanned re-
sults that could be catalogued, crowdsourced, and matched to local or
regional libraries for identification, thereby improving their utility in
real-time. Drugs tested on the FTIR did not need to be destroyed during
testing.
Difficulties in the identification of fentanyl (and with other drugs)
occurred in two key areas: multiple controlled substances (i.e., heavily
cut drugs) such as heroin, cocaine, and fentanyl mixtures together, and
when the drug was in pill form, which also renders the percent active
pharmaceutical ingredient low. The chemists and analyst were blinded
to the original form (i.e., powder, pill, since all tested samples were
prepared as powder) of the sample, which may have underestimated
the specificity of the device. This is because information that could be
gleaned at the time of drug checking from the sample (i.e., shape, form,
indicia) and/or from the individual requesting the test could help in-
form the prior probability of identification. The range of possibilities for
a library match and interpretation would therefore expect to be im-
proved for both the Raman and FTIR spectrometers.
We observed between-lab differences on the Raman spectrometer
and, to a smaller degree, the FTS. One hypothesis for this discrepancy is
that the Baltimore samples contained more mixtures of heroin and
fentanyl and heroin-only control samples compared with Providence,
which may have influenced the detection ability of the “Point-and-
Shoot”mode and SERS kit. In contrast, the Raman spectrometer de-
tected cocaine and other non-opioid narcotics well, which were more
numerous in the Providence samples and found more easily in the
“Point-and-Shoot”mode. These differences underscore the importance
of conducting a multi-site study and serve as a reminder of the power of
geography in drug supply and risk. The relatively small differences
between labs on the sensitivity and specificity of the FTS may be due to
differences in GC/MS limits of detection, the presence of undetected
fentanyl analogs, or other factors, which will be subject to further
analysis.
Taken together, our results suggest that, if detecting any presence of
fentanyl alone is of interest, the FTS would be the device of choice
provided that, at most, a 9.6% false negative and, at most, a 3.7% false
positive rate were acceptable. Even in locations where fentanyl is per-
vasive, FTS programs, particularly those coupled with tailored educa-
tion and naloxone distribution, may confer benefits beyond increasing
community awareness of fentanyl. If more information about the drug
sample beyond fentanyl's presence is of interest to PWUD and to the
public's health, and if higher false negative and positive rates were
tolerable, then a desktop FTIR device is a good option. Finally, if both
high sensitivity and specificity as well as more information about the
drug components are of interest, a two-step drug checking could be
considered. This could be achieved using the FTS in combination with
one of the other two devices, or conducting additional, though time-
delayed, confirmatory testing with GC/MS, for instance. In the U.S.,
powder street sample purities hover around three percent
T.C. Green, et al. International Journal of Drug Policy 77 (2020) 102661
5
(U.S. Drug Enforcement Administration, 2019) (and thus may be missed
by reliance on FTIR procedures alone) and 14% of seized counterfeit
pills contained lethal doses of fentanyl or analogs, supporting a two-
step drug checking process. Critically important and practical con-
siderations are the need for training and access to up-to-date sample
libraries for both the Raman and FTIR spectrometers and, for all drug
checking approaches, the potential need for state legislative change,
program cost, and ease of reaching high-risk PWUD with information
about the drug supply (Glick et al., 2019). As with all harm reduction
interventions, it is essential to involve PWUD in the design and pro-
gramming of drug checking services, with early evaluation and feed-
back from participants and nonparticipants (Bardwell et al., 2019).
There are numerous practical implications suggested by this re-
search. For instance, drug checking devices could be purchased by
public health agencies to be situated in easily accessible fixed or mobile
spaces, like a safe consumption site or syringe services program, for the
purposes of providing ongoing, free drug checking and harm reduction
education and resources prior to or proximal to drug use. Services could
be provided for on-site checking as well as FTS distribution for take-
home use. Legal provisions that permit operation of safe consumption
sites, that explicitly permit drug checking, and that define drug para-
phernalia and simple possession vary by location and may need to be
revisited to provide legal protection for participants and staff
(Glick et al., 2019;Park et al., 2019).
Communities lacking safe consumption sites or that are concerned
that drug checking violates drug possession laws could still improve
drug supply knowledge by testing the residue found in street drug
packaging with the devices. In the U.S., drug litter is regularly collected
by community organizations, gathered by first responders (e.g., fire-
fighters, law enforcement, ambulance staff) and municipalities, or
could be donated by or purchased from people who use drugs. Testing
actual drug or remnants of the drug supply could create an aggregated
and real-time drug trend dataset comparable to other international
surveillance systems, to better tailor responses, identify novel sub-
stances or harmful contaminants, and reduce the burden of overdose
(Brunt et al., 2016;Gine et al., 2017;Hondebrink et al., 2019). A data
source that provides a more balanced view of the street-available drug
supply being consumed could also help inform safer drug laws and
policies. Specifically, this study and the current body of knowledge
highlight the uncertainties associated with detecting fentanyl in the
illicit drug supply and harms associated with punitive supply-side ap-
proaches (e.g., Beletsky & Davis, 2017;Csete et al., 2016;Werb et al.,
2011) that support the need for criminal justice reform. Finally, once
established, future studies should explore how drug checking strategies
could influence a safer supply, for example, by reaching dealers or
higher level distributors (Bardwell, Boyd, Arredondo, McNeil & Kerr,
2019) as well as by engaging PWUD who may have never sought harm
reduction services (e.g., PWUD in more remote communities), and re-
engaging those lost to care or under-involved in harm reduction ser-
vices.
Conclusion
Findings from this study suggest there are valid and easy-to-use
devices which can provide important insights about the changing
nature of the illicit drug supply meaningful to both people who use
drugs and public health decision makers. As a public health response to
the opioid overdose crisis in North America, drug checking should be
considered and researched for its surveillance, monitoring, interven-
tion, and harm reduction applications.
CRediT authorship contribution statement
Traci C. Green: Conceptualization, Methodology, Formal analysis,
Writing - original draft, Funding acquisition. Ju Nyeong Park: Project
administration, Supervision, Validation, Formal analysis, Writing -
review & editing. Michael Gilbert: Conceptualization, Methodology,
Writing - review & editing. Michelle McKenzie: Project administration,
Supervision, Data curation, Writing - review & editing. Eric Struth:
Data curation, Investigation, Writing - review & editing. Rachel Lucas:
Data curation, Project administration, Supervision, Validation, Writing
- review & editing. William Clarke: Methodology, Writing - review &
editing. Susan G. Sherman: Conceptualization, Supervision, Writing -
review & editing, Funding acquisition.
Declaration of Competing Interest
The authors report no conflict of interest to declare.
Acknowledgments
The authors wish to thank The Providence and Baltimore Police
Departments for their collaboration and cooperation in this research.
We thank Joseph Muy for his laboratory guidance and protocol ex-
ecution. We are grateful to the leadership and staffof the Rhode Island
Department of Health State Health Laboratories and the Drug
Chemistry Laboratory for supplying space, instruments, and samples
used in this study. We thank Kimberly Pognon for her assistance in
extracting cases. Thanks to Thermofisher Scientific and Bruker Optics
for permitting the use of their devices for the duration of the study. The
companies had no role in the analysis or writing of this manuscript. The
authors report no conflict of interest to declare.
Funding: This work was supported by the Bloomberg American
Health Initiative, Baltimore, Maryland. The funding organization had
no role in the design and conduct of the study; in the collection, ana-
lysis, and interpretation of the data; and in the preparation, review, or
approval of the manuscript.
References
Clarke's analytic forensic toxicology (Second ed.). London, UK: Pharmaceutical Press.
Amlani, A., McKee, G., Khamis, N., Raghukumar, G., Tsang, E., & Buxton, J. A. (2015).
Why the fuss (Fentanyl urine screen study)? A cross-sectional survey to characterize
an emerging threat to people who use drugs in British Columbia, Canada. Harm
Reduction Journal, 12, 54. https://doi.org/10.1186/s12954-015-0088-4 10.1186/
s12954-015-0088-4. Retrieved from http://www.ncbi.nlm.nih.gov/entrez/query.
fcgi?cmd=Retrieve&db=PubMed&dopt=Citation&list_uids=26577516 pii .
Bardwell, G., Boyd, J., Arredondo, J., McNeil, R., & Kerr, T. (2019). Trusting the source:
The potential role of drug dealers in reducing drug-related harms via drug checking.
Drug and Alcohol Dependence, 198,1–6. https://doi.org/10.1016/j.drugalcdep.2019.
01.035. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/30856370.
Bardwell, G., Boyd, J., Tupper, K. W., & Kerr, T. (2019). "We don't got that kind of time,
man. We're trying to get high!": Exploring potential use of drug checking technologies
among structurally vulnerable people who use drugs. International Journal of Drug
Policy, 71, 125–132. https://doi.org/10.1016/j.drugpo.2019.06.018. Retrieved from
https://www.ncbi.nlm.nih.gov/pubmed/31336258.
Bardwell, G., & Kerr, T. (2018). Drug checking: A potential solution to the opioid over-
dose epidemic. Substance Abuse Treatment, Prevention, and Policy, 13(1), 20. https://
doi.org/10.1186/s13011-018-0156-3. Retrieved from https://www.ncbi.nlm.nih.
gov/pubmed/29801458.
Barratt, M. J., Bruno, R., Ezard, N., & Ritter, A. (2018). Pill testing or drug checking in
Australia: Acceptability of service design features. Drug and Alcohol Review, 37(2),
226–236. https://doi.org/10.1111/dar.12576. Retrieved from https://www.ncbi.
nlm.nih.gov/pubmed/28635057.
Beletsky, L., & Davis, C. S. (2017). Today's fentanyl crisis: Prohibition's iron law, revisited.
International Journal of Drug Policy, 46, 156–159. https://doi.org/10.1016/j.drugpo.
2017.05.050. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/28735773.
Brunt, T. M., Nagy, C., Bucheli, A., Martins, D., Ugarte, M., Beduwe, C., et al. (2016). Drug
testing in Europe: Monitoring results of the Trans European Drug Information (TEDI)
project. Drug Testing and Analysis.https://doi.org/10.1002/dta.1954. Retrieved from
http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?cmd=Retrieve&db=PubMed&
dopt=Citation&list_uids=26888408.
BTNX corporation. (2019). Harm reduction. Retrieved from https://www.btnx.com/
HarmReduction.
Centers for Disease Control and Prevention. (2017). Fentanyl. Retrieved from https://
www.cdc.gov/drugoverdose/opioids/fentanyl.html.
Csete, J., Kamarulzaman, A., Kazatchkine, M., Altice, F., Balicki, M., Buxton, J., et al.
(2016). Public health and international drug policy. Lancet (London, England),
387(10026), 1427–1480. https://doi.org/10.1016/S0140-6736(16)00619-X.
Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/27021149.
Fischer, B., Murphy, Y., Rudzinski, K., & MacPherson, D. (2016). Illicit drug use and
T.C. Green, et al. International Journal of Drug Policy 77 (2020) 102661
6
harms, and related interventions and policy in Canada: A narrative review of select
key indicators and developments since 2000. International Journal of Drug Policy, 27,
23–35. https://doi.org/10.1016/j.drugpo.2015.08.007. Retrieved from https://
www.ncbi.nlm.nih.gov/pubmed/26359046.
Gine, C. V., Vilamala, M. V., Measham, F., Brunt, T. M., Bucheli, A., Paulos, C., et al.
(2017). The utility of drug checking services as monitoring tools and more: A re-
sponse to Pirona et al. International Journal of Drug Policy, 45,46–47. https://doi.org/
10.1016/j.drugpo.2017.05.018. Retrieved from https://www.ncbi.nlm.nih.gov/
pubmed/28582668.
Glick, J. L., Christensen, T., Park, J. N., McKenzie, M., Green, T. C., & Sherman, S. G.
(2019). Stakeholder perspectives on implementing fentanyl drug checking: Results
from a multi-site study. Drug and Alcohol Dependence, 194, 527–532. https://doi.org/
10.1016/j.drugalcdep.2018.10.017. Retrieved from https://www.ncbi.nlm.nih.gov/
pubmed/30551090.
Goldman, J. E., Krieger, M. S., Buxton, J. A., Lysyshyn, M., Sherman, S. G., & Green, T. C.
(2019). Suspected involvement of fentanyl in prior overdoses and engagement in
harm reduction practices among young adults who use drugs. Substance Abuse, 1–8.
https://doi.org/10.1080/08897077.2019.1616245. Retrieved from https://www.
ncbi.nlm.nih.gov/pubmed/31206354.
Harper, L., Powell, J., & Pijl, E. M. (2017). An overview of forensic drug testing methods
and their suitability for harm reduction point-of-care services. Harm Reduction
Journal, 14(1), 52. https://doi.org/10.1186/s12954-017-0179-5. Retrieved from
https://www.ncbi.nlm.nih.gov/pubmed/28760153.
Hondebrink, L., Nugteren-van Lonkhuyzen, J. J., Hunault, C. C., van den Berg, J., van der
Gouwe, D., & van Riel, A. (2019). New psychoactive substances (NPS) in the
Netherlands: Occurrence in forensic drug samples, consumer drug samples and poi-
sons center exposures between 2013 and 2017. Addiction (Abingdon, England).
https://doi.org/10.1111/add.14868. Retrieved from https://www.ncbi.nlm.nih.gov/
pubmed/31656050.
Hungerbuehler, I., Buecheli, A., & Schaub, M. (2011). Drug checking: A prevention
measure for a heterogeneous group with high consumption frequency and polydrug
use –evaluation of Zurich's drug checking services. Harm Reduction Journal, 8, 16.
https://doi.org/10.1186/1477-7517-8-16. Retrieved from https://www.ncbi.nlm.
nih.gov/pubmed/21663618.
Kennedy, M. C., Scheim, A., Rachlis, B., Mitra, S., Bardwell, G., Rourke, S., et al. (2018).
Willingness to use drug checking within future supervised injection services among
people who inject drugs in a mid-sized Canadian city. Drug and Alcohol Dependence,
185, 248–252. https://doi.org/10.1016/j.drugalcdep.2017.12.026. Retrieved from
https://www.ncbi.nlm.nih.gov/pubmed/29475198.
Krieger, M. S., Yedinak, J. L., Buxton, J. A., Lysyshyn, M., Bernstein, E., & Rich, J. D.
(2018). High willingness to use rapid fentanyl test strips among young adults who use
drugs. Harm Reduction Journal, 15(1), 7. https://doi.org/10.1186/s12954-018-0213-
2. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/29422052.
Laing, M. K., Tupper, K. W., & Fairbairn, N. (2018). Drug checking as a potential strategic
overdose response in the fentanyl era. International Journal of Drug Policy, 62,59–66.
https://doi.org/10.1016/j.drugpo.2018.10.001. Retrieved from https://www.ncbi.
nlm.nih.gov/pubmed/30359874.
Marshall, B. D. L., Krieger, M. S., Yedinak, J. L., Ogera, P., Banerjee, P., Alexander-Scott,
N. E., et al. (2017). Epidemiology of fentanyl-involved drug overdose deaths: A
geospatial retrospective study in Rhode Island, USA. International Journal of Drug
Policy, 46, 130–135. https://doi.org/10.1016/j.drugpo.2017.05.029. Retrieved from
https://www.ncbi.nlm.nih.gov/pubmed/28601512.
Maryland Department of Health. (2018). Unintentional Drug- and Alcohol-Related in-
toxication deaths in Maryland, 2017. Retrieved from Baltimore, MD Https://bha.
health.maryland.gov/OVERDOSE_PREVENTION/Documents/Drug_Intox_Report_
2017.pdf.
Massachusetts Department of Public Health. (2018). Data brief: Opioid-related overdose
deaths among Massachusetts residents, May 2018. Boston, MA: Commonwealth of
Massachusetts, Department of Public Health.
O'Donnell, J. K., Halpin, J., Mattson, C. L., Goldberger, B. A., & Gladden, R. M. (2017).
Deaths involving fentanyl, fentanyl analogs, and U-47700 –10 States, July–December
2016. Morbidity and Mortality Weekly Report, 66(43), 1197–1202. https://doi.org/10.
15585/mmwr.mm6643e1. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/
29095804.
Park, J. N., Sherman, S. G., Rouhani, S., Morales, K. B., McKenzie, M., Allen, S. T., et al.
(2019). Willingness to use safe consumption spaces among opioid users at high risk of
fentanyl overdose in Baltimore, providence, and Boston. Journal of Urban Health:
Bulletin of the New York Academy of Medicine, 96(3), 353–366. https://doi.org/10.
1007/s11524-019-00365-1. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/
31168735.
Peiper, N. C., Clarke, S. D., Vincent, L. B., Ciccarone, D., Kral, A. H., & Zibbell, J. E.
(2019). Fentanyl test strips as an opioid overdose prevention strategy: Findings from
a syringe services program in the southeastern United States. International Journal of
Drug Policy, 63(Jan), 122–128. https://doi.org/10.1016/j.drugpo.2018.08.007
Retrieved from.
Rhode Island Department of Health. (2018). Overdose death data. Retrieved from
Preventoverdoseri.org.
Rouhani, S., Park, J. N., Morales, K. B., Green, T. C., & Sherman, S. G. (2019). Harm
reduction measures employed by people using opioids with suspected fentanyl ex-
posure in Boston, Baltimore, and providence. Harm Reduction Journal, 16(1), 39.
https://doi.org/10.1186/s12954-019-0311-9. Retrieved from https://www.ncbi.
nlm.nih.gov/pubmed/31234942.
Rudd, R. A., Seth, P., David, F., & Scholl, L. (2016). Increases in drug and opioid-involved
overdose deaths –United States, 2010–2015. Morbidity and Mortality Weekly Report,
65(5051), 1445–1452. https://doi.org/10.15585/mmwr.mm655051e1.Retrieved
from https://www.ncbi.nlm.nih.gov/pubmed/28033313.
Sande, M., & Sabic, S. (2018). The importance of drug checking outside the context of
nightlife in Slovenia. Harm Reduction Journal, 15(1), 2. https://doi.org/10.1186/
s12954-018-0208-z. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/
29329544.
Sanger-Katz, M. Bleak new estimates in drug epidemic: A record 72,000 overdose deaths
in 2017. The New York Times. Retrieved from https://www.nytimes.com/2018/08/
15/upshot/opioids-overdose-deaths-rising-fentanyl.html.
Sherman, S. G., Morales, K. B., Park, J. N., McKenzie, M., Marshall, B. D. L., & Green, T. C.
(2019). Acceptability of implementing community-based drug checking services for
people who use drugs in three United States cities: Baltimore, Boston and providence.
International Journal of Drug Policy, 68,46–53. https://doi.org/10.1016/j.drugpo.
2019.03.003. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/30991301.
Tupper, K. W., McCrae, K., Garber, I., Lysyshyn, M., & Wood, E. (2018). Initial results of a
drug checking pilot program to detect fentanyl adulteration in a Canadian setting.
Drug and Alcohol Dependence, 190, 242–245. https://doi.org/10.1016/j.drugalcdep.
2018.06.020. Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/30064061.
U.S. Drug Enforcement Administration, S. T. a. R. L. (2019). Fentanyl signature profiling
program report. Retrieved from.
U.S. Drug Enforcement Agency. (2018). 2017 National drug threat assessment. Retrieved
from Washington, DC https://www.dea.gov/docs/DIR-040-17_2017-NDTA.pdf.
Werb, D., Rowell, G., Guyatt, G., Kerr, T., Montaner, J., & Wood, E. (2011). Effect of drug
law enforcement on drug market violence: A systematic review. International Journal
of Drug Policy, 22(2), 87–94. https://doi.org/10.1016/j.drugpo.2011.02.002.
Retrieved from https://www.ncbi.nlm.nih.gov/pubmed/21392957.
Yau, B., Arredondo, J., Tobias, S., Grant, C., Choi, J. C., Lysyshyn, M., et al. (2019). Afield
assessment of Fourier-transform infrared spectroscopy (FTIR) and fentanyl immunoassay
strips as point-of-care drug checking technologies. Retrieved from Vancouver, British
Columbia.
T.C. Green, et al. International Journal of Drug Policy 77 (2020) 102661
7
