Background
Medical complications related to drugs account for a significant fraction of patient visits to the emergency department (ED). These visits may be a result of illicit drug abuse, intentional or inadvertent overdose of prescription or over-the-counter medications, or drug-drug interactions [
1‐
3]. There is increasing concern about the danger posed by misuse of prescription medications, particularly those with high potential abuse liability (e.g., opioids), especially when used in combination with ethanol or street drugs [
4]. In some patients, such as those with altered mental status, a medical history may be unclear at the time of presentation to the ED. To aid in the diagnosis and management of drug-related complications, laboratory tests to screen for the presence of drugs and drug metabolites are widely used in emergency medicine [
3,
5]. We will refer to these tests as 'drug of abuse/toxicology (DOA/Tox) screening tests'.
Over the last four decades, a number of methods have been used for DOA/Tox screening including antibody-based assays (immunoassays) [
6,
7]. DOA/Tox immunoassay screens for amphetamines, barbiturates, benzodiazepines, cannabinoids, methadone, opiates, and tricyclic antidepressants (TCAs) were first introduced into clinical practice in the United States in the 1970s, initially as radioimmunoassays and later as non-radioactive immunoassays [
8,
9]. Immunoassays have steadily displaced other DOA/Tox screening methods such as thin-layer chromatography or colorimetric assays [
7]. Currently, the most common methods used in the United States for DOA/Tox screening are homogeneous immunoassays that can be performed rapidly on a variety of different instruments, ranging from small devices that can be located within or near the ED to large, high-throughput analyzers found in hospital clinical laboratories or off-site reference laboratories [
6,
7]. Screening assays are different from confirmatory tests such as gas chromatography/mass spectrometry (GC/MS) that can provide definitive identification of individual drugs and their metabolites [
7]. Confirmatory tests are often more labor-intensive, technically demanding, and expensive compared with screening tests. For many EDs, confirmatory tests are available only by referral of patient samples to an off-site reference laboratory, such that turnaround time for results is often not fast enough to aid in real-time patient management.
In the United States, there are currently marketed DOA/Tox screening immunoassays for 18 targets (i.e., single drugs or drug classes) including: amphetamines, barbiturates, benzodiazepines, cocaine metabolite/benzoylecgonine, buprenorphine, cannabinoids, heroin metabolite/6-acetylmorphine (6-AM), lysergic acid diethylamide (LSD), MDMA/Ecstasy (3,4-methylenedioxymethamphetamine), methadone, methadone metabolite/EDDP (2-ethylidine-1,5-dimethyl-3,3-diphenylpyrrolidine), methaqualone, nicotine metabolite/cotinine, opiates, oxycodone, phencyclidine (PCP), propoxyphene, and TCAs. For some drugs or metabolites (e.g., buprenorphine, heroin metabolite/6-AM), there may be only one or two manufacturers marketing an assay whereas for more common tests (e.g., amphetamines, benzodiazepines, opiates), there are many different marketed assays. Different assays for the same analyte may vary in terms of analytical sensitivity and specificity, leading to potential difficulties in clinical interpretation. DOA/Tox screening test is most often performed on urine but, in some cases, serum/plasma or saliva may be used [
5,
7,
10].
DOA/Tox screening immunoassays may be designed by raising antibodies against a single drug or drug metabolite ('target compound'). Alternatively, multiple target compounds may be used to achieve broader detection of a class of drugs. There is a general trend towards use of monoclonal antibodies in marketed assays, but assays using polyclonal antibodies are still used widely in some cases [
6,
7]. Theoretically, use of monoclonal antibodies provides more consistent performance over polyclonal antibodies. DOA/Tox screening assays may be directed at classes of drugs such as amphetamines, barbiturates, benzodiazepines, cannabinoids, and opiates [
7,
10]. In these 'broad specificity' DOA/Tox assays, ideally the specificity of the assay is broad enough to detect a range of 'within-class' compounds but not too non-specific to cross-react with 'out-of-class' compounds that may have similar chemical structures. Other DOA/Tox screening assays are directed towards detection of a single target compound (drug or drug metabolite) without cross-reactivity with other similar structures. Examples of 'single target' DOA/Tox assays include those for buprenorphine, methadone, and propoxyphene.
DOA/Tox screening immunoassays have two main limitations. First, false positives may occur when an 'out-of-class' compound with structural similarity to the target compound(s) causes a positive screening result [
3,
5,
6,
10]. Such cross-reactive molecules can be structurally related drugs, drug metabolites, or endogenous compounds [
7,
11]. Manufacturers of DOA/Tox screening immunoassays typically test commonly used drugs for cross-reactivity including over-the-counter and prescription medications likely to be taken concomitantly with the target drug, as well as various other compounds [
12]. Information on assay sensitivity and cross-reactivity is normally reported in the package insert of the assay or the website of the manufacturer. In other cases, cross-reacting compounds for DOA/Tox screening assays are not reported by the assay manufacturer in the package insert but instead are first described in the medical literature. Examples of such published reports of DOA/Tox assay cross-reactivity include fluoroquinolone antibiotic cross-reactivity with opiate assays [
13], venlafaxine cross-reactivity with PCP immunoassays [
14‐
16], and quetiapine cross-reactivity with TCA assays [
17‐
19]. The second main limitation of DOA/Tox screening immunaossays is failure to detect some drugs within a class, resulting in false negatives [
3,
5,
6,
10]. Examples of false negatives would be inability to detect clonazepam in a benzodiazepines assay or oxycodone in an opiates assay. Some examples of drugs that can cause false negatives and false positives in DOA/Tox immunoassays are listed in Tables
1 and
2.
Table 1
Drugs or drug metabolites that can produce false negatives on DOA/Tox screening immunoassays
MDMA | AMPH | 0.361 | 1,300 | 2,500 | 2,000 | 1,300 | 697,000 | 34,300 |
Alprazolam | BENZO | 0.610 | 113 | 300 | 450 | 25 | 219 | 65 |
Clonazepam | BENZO | 0.656 | 214 | 300 | 350 | 3,000 | 307 | 260 |
Clonazepam metabolite (7-amino) | BENZO | 0.755 | 2,334 | 800 | 7,500 | 1,000 | 288 | 5,700 |
Clobazam | BENZO | 0.796 | 218 | 500 | 700 | 250 | 237 | 260 |
Buprenorphine | OPIA | 0.783 | | | No effect | | | No effect |
Oxycodone | OPIA | 0.800 | No effect | 17,000 | 20,000 | 16,000 | > 75,000 | 2,550 |
Oxymorphone | OPIA | 0.847 | No effect | No effect | 40,000 | 40,000 | | > 20,000 |
Amoxapine | TCA | 0.508 | No effect | | No effect | No effect | | No effect |
Table 2
Drugs that can produce false positives on broad specificity DOA/Tox screening immunoassays
Phentermine | AMPH | 0.778 | No effect | No effect | 750,000 | No effect | | 25,000 |
Levofloxacin | OPIA | 0.560 | 1,700,000 | No effect | 125,000 | 60,000 | 200,000 | |
Dextromethorphan | PCP | 0.565 | 12,900 | No effect | 500,000 | No effect | No effect | 12,000 |
Meperidine | PCP | 0.538 | 34,650 | No effect | No effect | No effect | No effect | 25,000 |
Carbamazepine | TCA | 0.460 | 29,972 | | No effect | No effect | | No effect |
Cyclobenzaprine | TCA | 0.565 | | | 2,000 | 450 | | |
Prochlorperazine | TCA | 0.630 | 999 | 100,000 | | | | |
Quetiapine | TCA | 0.485 | 2,484 | | No effect | No effect | | 100,000 |
In clinical practice, drugs are commonly classified by their therapeutic class, but this does not explicitly define how similar drugs may be to one another in terms of chemical structure and their potential for cross-reactivity in DOA/Tox screening immunoassays. Therefore, we have utilized a computational method known as similarity analysis between molecules [
20,
21]. Variables that can be included in similarity calculations are extensive and include those related to molecular structure, electrostatic potential, shape, and electron density. Similarity analysis has been used widely in the pharmaceutical industry as a 'virtual' screen for identifying drug-like molecules and predicting drug toxicity, and can be valuable in narrowing the number of compounds subjected to
in vitro, animal, or human testing [
20,
22,
23]. In our analysis, we have used two-dimensional (2D) similarity with the Tanimoto coefficient, which compares two compounds and generates a similarity measure that ranges from 0 to 1, with 0 being maximally dissimilar and 1 being maximally similar [
21,
24]. We have found that this similarity measure correlates well with cross-reactivity of immunoassays used clinically for DOA/Tox screening and therapeutic drug monitoring [
25,
26].
In this study, we applied similarity analysis as a quantitative tool to rationalize false positives and false negatives of DOA/Tox screening assays. We have also compiled historical data on prescription drug usage in the United States to demonstrate how changing patterns of drug use may influence clinical utility of DOA/Tox screening assays. Lastly, we present the results of our own investigation into the causes for positive screening results for PCP and TCA screening assays in our medical center, which has adult and pediatric EDs that serve as a regional toxicology referral center.
Discussion
DOA/Tox screening immunoassays are widely used in emergency medicine [
5,
7,
10]. These assays are also used by substance abuse treatment centers, chronic pain clinics, and psychiatric units, in addition to employee and competitive athlete drug screening programs [
5,
10]. The multiple uses of DOA/Tox screening tests probably provides substantial inertia to attempts to alter assay design and performance, as changes in assay design and detection cutoffs could have wide-ranging impacts. The most common set of DOA/Tox screening assays (e.g., amphetamines, barbiturates, benzodiazepines, cocaine metabolite, opiates, and PCP), and their antigenic targets, have remained remarkably similar across the last four decades.
As we have shown, false negatives for DOA/Tox screening assays aimed at drug classes can occur with drugs that became widely used in clinical practice in the United States after the 1970s and which have relatively low structural similarity to the classic antigenic targets of their associated immunoassays. For benzodiazepines, this includes alprazolam, clobazam, clonazepam, and lorazepam, while for opiates this includes buprenorphine, oxycodone, and oxymorphone. A few marketed immunoassays (e.g., Biosite Triage) have attempted to broaden specificity by using antibodies raised against multiple antigenic targets. The potential disadvantage of this approach is reduced specificity and increased false positives. Also, any alteration of these immunoassays has implications for workplace and athlete testing, leading to pressure to keep assay performance stable across many years of testing.
Many marketed DOA/Tox screening immunoassays have documented cross-reactive drugs that can produce false positives. In the medical setting, false positives can lead to incorrect diagnoses and treatment. One way to limit false positives is to use higher concentration cutoffs for determining what constitutes a positive screening result, although this has the trade-off of reducing sensitivity. This strategy is common in workplace DOA testing where cutoff concentrations for a variety of DOA screening tests are often higher than cutoffs used in the medical setting, so as to limit false positives that require costly and time-consuming confirmatory testing [
7,
10]. For example, using higher cutoffs helps reduce the issue of poppy seed ingestion causing a positive opiate screen [
70] or passive marijuana inhalation resulting in a tetrahydrocannabinol positive screen [
71].
We demonstrated that PCP and TCA screening assays are prone to false positives by common drugs that may be taken in overdose, either in suicide attempts or for psychotropic effects (e.g., dextromethorphan, meperidine). In our own medical center study, there were more false positives than true positives for both PCP and TCA screening assays applied to a clinical sample that included many ED patients. This brings into question the utility of these particular tests in settings where use/abuse of the target drug(s) is uncommon.
One application of our Tanimoto similarity assessment using the MDL keys would be to identify compounds that have a high likelihood of cross-reacting with marketed immunoassays. The 2D similarity method can readily screen very large databases of many thousands of drugs (including herbal products) and their metabolites. Compounds with high Tanimoto similarity to the immunoassay antigenic target(s) can then be prioritized for testing for cross-reactivity. This approach would provide a more systematic approach to cross-reactivity testing and may identify previously unknown clinically important cross-reactive drug or drug metabolites more quickly, leading to an increased recognition of potential cross-reactivity by clinicians.
The steady increase in prescription and over-the-counter medications available clinically presents a difficult challenge for future DOA/Tox immunoassay design. Some newer therapeutic classes of drugs that are often taken in overdose (e.g., atypical antipsychotics, SSRI antidepressants) are actually not that closely related to one another (see Additional file
1, tabs W and X). This is a contrast to older drug classes (e.g., barbiturates, benzodiazepines, and TCAs), where there is significant structural similarity between all drugs within the class. The clinical implication of this is that it would be difficult to design an 'SSRI overdose screen' or 'atypical antipsychotic overdose screen' with standard immunoassay technology. It also suggests that development of DOA/Tox immunoassays has not kept pace with the development of new drugs relevant to the ED community or with changes in patterns of abuse of illicit and prescription drugs.
The analytical methods currently used mainly for DOA/Tox confirmatory testing, such as GC/MS and liquid chromatography/tandem mass spectrometry (LC/MS/MS), can specifically identify (and in some cases quantitate) drugs and their metabolites. This technology, however, is technically demanding, labor-intensive, expensive compared to immunoassays, and usually available only at reference laboratories or in clinical laboratories associated with larger medical centers [
7]. A future goal would be to develop and adapt GC/MS, LC/MS/MS, or a novel technology in a manner to be more widely accessible clinically, so as to provide detailed drug exposure data with a rapid turnaround time, allowing ED physicians to make more specific diagnoses and treatment plans. A scientifically similar challenge is in emerging technology to develop portable yet analytically robust sensors for chemical warfare agents or environmental pollutants [
72], and there may be opportunities to develop clinical applications using related technology.
An important limitation of the 2D similarity approach used in our study is that this cannot account for the complex three-dimensional (3D) molecular interactions that mediate antibody-antigen binding as occurs in immunoassays. To our knowledge, a 3D structure of an antibody used in a marketed DOA/Tox screening immunoassay bound to its antigenic target has not been reported, although there has been structural determination of several other antibodies being evaluated as novel antidotes to DOA overdose (e.g., PCP [
73] and cocaine [
74,
75]), in which the antibody interacts with all portions of the target molecule. For DOA/Tox screening immunoassays where similar antibody-drug interactions apply, whole molecule similarity measures (as used in our study) seem appropriate for prediction; however, this may not always be the case. A crystal structure of morphine bound to a monoclonal antibody showed that the antibody interacted with the more hydrophobic portion of morphine, while the hydrophilic half was mostly solvent exposed [
76]. For target compounds like morphine, similarity searching using substructures may therefore be worth evaluating, although depending on the size of the molecule, complexity, and novelty this may yield many more molecules predicted as positives.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
MDK conceived of the study and structured the data. MDK and SE drafted the manuscript. AFP participated in the planning of the study, interpretation of data, and the historical data analysis. MGS and SG performed and analyzed the laboratory studies involving immunoassays and GC/MS. SE and MI performed the computational analyses. All authors participated in editing and revising the manuscript and approved the final version.