U: Urine; OF: Oral Fluid; DHHS: Department of Health and Human Services; LC-MS/MS: liquid Chromatography-Tandem Mass Spectrometry; MPA: Mobile phase-A; MPB: Mobile phase-B; SEM: Standard Error of the Mean; SN: Sensitivity; SP: Specificity; TP: True Positive; FP: False Positive; FN: False Negative; TN: True Negative; 6MAM: 6-Monoacetylmorphine

Drug testing has become essential in today’s society with substance abuse increasing daily, worldwide, and the need for prescription drug monitoring in the fields of pain management, behavioural health, and internal medicine, among others [1]. The most common test matrix is urine [2]. As the need for drug testing has expanded, there have been advances in other drug monitoring techniques and matrices, such as oral fluid testing [3]. The growing popularity in oral fluid as a desirable matrix has opened the discussion of its concordance with urine drug testing.

Oral fluid drug testing has many advantages over urine testing, the major advantage being that it uses a less invasive collection process to benefit the patient [4]. Furthermore, it limits the potential for adulteration, substitution, and diversion due to the ability to directly observe collections, and minimizes the concerns surrounding specimen integrity [3].

Limitations in recent and past publications of this nature include single subset patient populations, for example, pain patients [4], or are limited to certain drug classes [5]. In this clinical study, we set you to determine the concordance of urine and oral fluid drug testing collected on patient’s same day while expanding on patient populations included in the data (i.e., pain, treatment, internal medicine, etc.) as well as only including one urine and oral fluid specimen per patient.

For the purposes of this study, concentrations greater than or equal to the validated cut off level were detected by liquid Chromatography- Tandem Mass Spectrometry (LC-MS/MS).

Participants and specimen collection

This study was approved by Aspire IRB Santee, CA and was conducted over a 4-month period from August 2016 to November 2016, with specimens collected from 25 clinics across 12 states and shipped to Precision Diagnostics (San Diego, CA). Among the clinics, clinical specialties included 11 addiction/rehabilitation clinics, 6 internal or family medicine clinics, 4 pain management clinics, and 4 behavioural health clinics where both urine and oral fluid samples were collected on the same day from each patient. The time between each collection could not be determines as collection was completed within the individual clinics. The randomly selected specimens were analysed from 142 patients undergoing prescription drug management programs within addiction, pain, behavioural, or internal medicine practices. Only one urine and oral fluid specimens pair collected same day from each patient was used in the analysis to prevent bias. Collection and storage of urine and oral fluid specimens were completed and regulated at the corresponding physician offices and sent to Precision Diagnostics. Oral fluid specimens were collected with Quantisal^® saliva collection devices per the Quantisal^® manufacturing instructions. The collector was placed under the patient’s tongue until the indicator turned blue. Then the collector was placed in the buffer solution of the transport tube and packaged for transport. The analyses of analytes in each specimen tested were limited to those deemed medically necessary by the requesting physician.

Laboratory analyses

The urine specimens were prepared and analysed using the previously published method by Krock et al. [6]. A high-sensitivity LC-MS/MS method was developed and validated capable of detecting low concentrations of up to 71 drugs and metabolites important for monitoring medication adherence and substance use disorder for the urine specimens. The test results were analysed on logarithmic distributions fitted with a trend line to estimate the required cut-off level necessary to capture the normal distribution of each drug and metabolite. An LC-MS/MS method was developed and validated, capable of quantitating up to 33 drugs and metabolites important for monitoring medication adherence and substance use disorder for the oral fluid specimens. A Shimadzu (Kyoto, Japan) 20 series binary pump system, well-plate auto sampler, and temperature-controlled column oven was paired with a Sciex (Framingham, MA) 6500 triple quadruple mass spectrometer. LC-MS grade water was obtained from a Sartorius (Bohemia, NY) ultrapure water system, LC-MS grade methanol, acetonitrile and ammonium hydroxide were obtained from EMD Millipore (Billerica, MA), and LC-MS grade formic acid was obtained from Covachem (Loves Park, IL). A Phenomenex (Torrance, CA) Kinetex^® phenyl-hexyl column with dimensions of 50 × 4.6 mm and 2.6 μm particle size was used for chromatographic separation. The binary pump system delivered a mixture of the mobile phases at the proportion and flow rate displayed in Tables 1 and 2. Mobile phase-A was 0.1% formic acid in LC-MS grade water and Mobile Phase-B (MPB) was LC-MS grade methanol containing 0.1% formic acid.

Table 1: The gradient of mobile phase delivered during oral fluid analysis allowed separation of the isobaric compounds and interferents and provided a high organic phase portion to clean the column before the next analysis.

Table 2: Cohen’s Kappa values drugs and metabolites in oral fluid and urine.

The mass spectrometers used the following settings common to all analytes; curtain gas of 35 L/min, collision gas of 10 L/min, positive mode Ion Spray voltage of 2500 V, source temperature of 450°C, ion source gas 1 of 60 L/min and ion source gas 2 of 50 L/min. Two transitions for each analytes were optimized for declustering potential, collision cell energy and exit potential. Analytes and internal standards were obtained from Cerilliant (Round Rock, TX). Fourpoint calibration curves were prepared from the cut-off level to 25 times the cut-off level for each analytes. Two quality control samples were analysed with each batch of specimens to ensure acceptability of results. Samples were prepared by an automated solid phase extraction procedure, outlined here. Automated sample handling tips filled with 10 mg of cation exchange resin (DPX Technologies, Coloumbia, SC) were wetted with a mixture of 50%each methanol and deionized water, by volume. A mixture of stable isotope internal standards was added to the specimens and drawn into the solid phase extraction tips. Unbound material was rinsed away with deionized water and the analytes were eluted into a 96-well plate with 5% ammonium hydroxide in acetonitrile. This plate was dried to completeness using a MiniVap (Porvair Sciences, Wrexham, Wales, UK) drying station set to 40 L/ min of nitrogen and heated to 45°C. Prior to analysis, the specimens were reconstituted with 100 μL of a mixture of 20% methanol and 80% water by volume, with 0.1% formic acid. Five μL of prepared specimen was injected onto the analytical column for LC-MS/MS analysis and a flow diversion valve sent the first 1.3 minutes of eluent to waste. Results were analyzed using Indigo Bio Automation’s (Indianapolis, IN) ASCENT software. A four-point calibration curve was used with a linear fit and 1/x weighting. Calibrator acceptability was within ± 20% of the expected concentration with an R2 value of greater than 0.98. The area ratio of the analyte to a deuterated internal standard was used to account for ion suppression. All analytes had a signal to noise calculation of greater than 10 at the lower limit of quantitation. The precision and accuracy of the assay was evaluated over five days for both intraday and interday variability and all analytes were within 20% CV. Recovery was determined to be within ± 20% for all analytes.Drugs and metabolites detected in urine and oral fluid are listed in Table 3 along with the associated cut off levels used as developed and validated per Krock et al. [6].

Table 3: Cut off levels for drugs and metabolites in oral fluid and urine.

Data analyses

For each analyte, the mean, Standard Error of the Mean (SEM), and the concentration range were determined in Table 4.The Sensitivity (SN), how likely to detect analytes in both urine and oral fluid, and Specificity (SP), how likely to have a negative in urine or oral fluid were determined and used to evaluate the validity of oral fluid drug testing as an alternative or replacement for urine drug testing [7,8]. The equations used to determine sensitivity and specificity using the previously published method in Fenn Buderer [9], where True Positive (TP) is the number of urine and oral fluid positives; False Positive (FP) is the number of urine negatives and oral fluid positives; False Negative (FN) is the number of urine positives and oral fluid negatives; True Negative (TN) is the number of urine and oral fluid negatives (Table 5). Furthermore, the predictive values, both positive, proportion and negative were determined using the determine TP, FP, FN, and TN values resulted in Table 5.

Table 4a: Prevalence, mean, median, and range of drug and metabolite concentrations in urine and oral fluid.

Table 4b: Prevalence, mean, median, and range of drug and metabolite concentrations in urine and oral fluid.

Table 5: Test agreement between urine and oral fluid.

Urine and oral fluid in this study are considered the ‘judges’ operating independently from the analytes or ‘categories,’ to determine the percent agreement and ultimately the reliability of the results, or Cohen’s Kappa [10]. Cohen’s Kappa values less than 0 are considered ‘poor’, 0-0.20 are ‘slight’, 0.20-0.40 are ‘fair’, 0.40-0.60 are ‘moderate’, 0.60-0.80 are ‘substantial,’ and 0.80-1.00 are in almost ‘perfect agreement [11]. The agreement between the two ‘judges’ was determined by the proportion in which the units agree, or po, and the proportion of expected agreement, or pe; where po and pe are found per Cohen [10]. (Table 2) The same analysis was completed for all analytes and as well as two subsets of drugs set out by the Department of Health and Human Services (DHHS) for federal workplace practice.

Prevalence and concentration

The prevalence and concentration of 26 drugs and metabolites were evaluated for 142 patients, where a urine and oral fluid specimen was collected in concordance. The analyses of 142 patients in urine and oral fluid specimens submitted from August 2016 to November 2016 were confirmed by LC-MS/MS and are listed in Table 4. Analytes that could not be tested for in the oral fluid matrix were not included for the purposes of this study. For example, results for many metabolites such as noroxycodone, norhydrocodone, alphahydroxyalprazolam, 7-aminoclonazepam, and nordiazepam were excluded from this study as they are not currently incorporated into the LC-MS/MS oral fluid testing.

Patients excluded from this table: Patients with out of range creatinine (<20 mg/dL), patients suspected of contamination of the oral cavity and patients suspected of pill shaving.

Buprenorphine was detected most frequently in both oral fluid and urine. Buprenorphine is followed in urine by its metabolite norbuprenorphine then naloxone, hydromorphone, and amphetamine more frequently. Buprenorphine is followed in oral fluid by methamphetamine, amphetamine, THC and oxycodone most frequently.

From the results listed in Table 4, in urine, norbuprenorphine, hydromorphone, naloxone, oxazepam and temazepam tested positive most prevalently while their oral fluid counterparts were negative for these analytes. In oral fluid, methamphetamine, heroin metabolite (6MAM), amphetamine, morphine, codeine and cocaine tested positive most prevalently in oral fluid while their urine counterparts were negative for these analytes.

In urine when methadone, methamphetamine, carisoprodol, and meprobamate were positive, the oral fluid counterparts were also positive. In oral fluid, if naloxone, norbuprenorphine, zolpidem, and lorazepam were positive, the urine counterparts were also positive. It was also noted when temazepam and oxazepam were positive in urine, the oral fluid counterparts were always negative.

In many cases, the concentrations found in urine were 10 to 100 times greater compared to oral fluid concentrations. Per the methods used in Cohen (1960), the Pearson product moment correlation coefficient were determined for 6 analyte data sets (morphine; r=0.305, oxycodone; r=0.052, buprenorphine; r=0.079, amphetamine; r=0.51, methamphetamine; r=0.22, and naloxone; r=0.09). This is the measure of strength of a linear relationship between two subsets where r=1 is a perfect correlation, and r=0 indicated no correlation [12]. From these results, there was a positive relationship for amphetamine, morphine, and methamphetamine. Figure 1 demonstrates an example of this relationship for amphetamine concentrations in urine and oral fluid.

Figure 1: Amphetamine concentrations in urine and oral fluid.

Agreement between oral fluid and urine

The overall agreement between the urine and oral fluid samples is shown in Table 6. This table included all n=142 patients with no exclusions. Based on statistical references the agreement of all analytes listed in Table 5, between urine and oral fluid was found as 52.1% [10,13,14]. These values fall into range with Heltsley et al. study of comparison, however the Cohen’s Kappa of 0.501, indicated a lesser strength, ‘moderate,’ between urine and oral fluid results compared to Heltsley et al. study [4]. All 26 drug analyte results were evaluated using urine and oral fluid as the two raters.

Table 6: Validity, agreement and predictive value.

Two subsets of the data per SAMHSA’s Workplace Program Guidelines were also examined using the same methodology. The first subset examined were the analytes set out in the Department of Health and Human Services (DHHS) which included amphetamine, methamphetamine, THC, cocaine, codeine, morphine, and hydromorphone. The agreement and Cohen’s Kappa was 73.1% and 0.682 respectively. The second subset was the extended DHHS, added in hydrocodone and oxycodone which resulted in an agreement of 68.5% and Cohen’s Kappa of 0.644. Both subsets were found to have a ‘substantial’ agreement between urine and oral fluid (Table 6) [10,15].

This study compared the use of urine and oral fluid drug testing for licit and illicit drugs in the population of addiction, internal medicine, behavioural medicine, and pain patients, with the potential for oral fluid matrix to be used as an alternative to the standard urine analysis. All urine and oral fluid specimens were analysed by LC-MS/MS.

There was ‘moderate’ agreement found between the test results in the oral fluid and urine specimens as a 52.1% agreement and 0.501 Cohen’s Kappa coefficient between analytes was calculated. This suggests that there is a ‘moderate’ relation between urine and oral fluid drug testing across all observed analytes in this study, indicating oral fluid to be a substitute, but not replacement to urine drug testing [10].

There are many factors that can be attributed to the observed differences between urine and oral fluid drug testing. Pharmacokinetic factors involved in the two matrices can have a large effect on the concordance of results. Many drugs are subject to ‘phase 1’ drug metabolism and undergo some degree of metabolism before entering ‘phase II’ in which the drugs are rendered more polar enabling the final products of drug metabolism to be excreted in urine [16,17]. Once bound or glucuronidated, the polarity and charge on the drug or metabolites make it hard to passively diffuse into oral fluid for detection [1,16,17]. This phenomenon is particularly observed in benzodiazepines which are highly protein bound and weakly acidic allowing for ease of detection with urine rather than oral fluid drug tests [1].

Alternatively, oral fluid samples tend to detect drugs in the unbound or free form. Therefore, initial confirmatory tests target parent drugs. For example, the carboxy metabolite of THC has almost no presence in oral fluid, hence oral fluid testing targets the parent drug THC, whereas urine detection involves the detection of the carboxy metabolite, THC-COOH [1,18]. There are some exceptions to this rule of thumb, like the conversion of cocaine to its metabolite, benzoylecgonine which is very well captured in oral fluid [1,18]. A main distinction of oral fluid testing is its variability in pH and the influence it plays in the detection of certain analytes [1,16]. Weak bases are detected in higher concentrations and for longer time in oral fluid. Depending on the pKa and lipophilicity, these analytes are subjected to ‘Ion trapping’ due to the difference of pH between blood and oral fluid. The low pH in oral fluid allows weak bases to ionize, increasing oral fluid drug concentrations [16,17].

The detection of an analytes can also be dependent on many factors involving the individual patient/subject such as dose, time of dose, and route of administration [1-3]. In this study, the dose and therefore time of dose of a patient relies on limited knowledge. Consequently, an analyte positive in urine and negative in oral fluid could be due end elimination of a particular drug. For example, results in this study show a large prevalence of hydromorphone detection in urine than in oral fluid. Although pharmaceutically available, in urine hydromorphone is commonly detected as a metabolite of morphine or hydrocodone use [19]. Depending on last use, hydromorphone may not present in oral fluid unless taken in free form i.e., Dilaudid [1]. Detection of a drug may also be affected by the route of administration, for instance, drugs detected in oral fluid may be due to residual contamination of the oral cavity due to recent use after smoking or snorting a certain drug [18].

The most prevalent positives in urine with negative results in oral fluid consisted of nor buprenorphine, hydromorphone, naloxone, oxazepam and temazepam (Table 5). With the higher prevalence of drug metabolites detected in urine, these results suggest, that urine drug testing can be better used when monitoring drug elimination over time due to differences in detection time between urine and oral fluid, along with its ability to detect metabolites of the parent licit and illicit drugs. Results of this study, indicates urine analysis as fit for monitoring medication compliance and licit drug use, while also being successful in determining illicit drug use such as methamphetamine, THC, and cocaine (Table 5).

For those analytes positive in oral fluid and negative in urine, it is possible the drug had not had the allotted time needed for metabolism to then excretes the analytes in urine or may be due to contamination of the oral cavity, indicating recent use [18]. Oral fluid also tends to have a high prevalence of illicit drugs, like methamphetamine, due to the lower pH in the oral cavity causing ion trapping of the positively charged drug [11,18]. Results for the illicit drugs in this study: heroin metabolite (6MAM), methamphetamine, cocaine, and THC, all tested positive at a higher incidence in oral fluid than in urine (Table 5). This suggests that the oral fluid matrix can be used for monitoring illicit drug use and is a good indicator of recent drug use. Based on these results, it can be concluded that oral fluid testing would be successful in the treatment and rehabilitation environments, monitoring patient’s illicit use of drugs. Additionally, since we see a very high concordance of buprenorphine and methadone, prescribed medications used to treat opioid addiction, oral fluid testing can be considered a successful alternative to urine monitoring [18].

In this study there was a stronger agreement between the drugs indicated in the DHHS and ‘extended’ DHHS subsets showing substantial agreements of 73.1%, with a Choen’s Kappa of 68.2% and 68.5%, with a Choen’s Kappa of 64.35%, respectively. Due to the drugs evaluated: amphetamine, methamphetamine, THC, cocaine, codeine, morphine, hydromorphone, and the addition of hydrocodone and oxycodone in the extended subset, there was a strong agreement between urine and oral fluid as seen in Table 5. These drugs capture well in both matrices as they involve illicit drugs, and parent drugs. Without the addition of metabolites for the opioids involved, the agreement between urine and oral fluid is strengthened. This supports the conclusion that oral fluid can act as a successful alternative, but not a true replacement to urine drug testing especially in the realm of pain medicine.

Due to the diverse patient population in this study, a high rate of positives within the pool of analytes was anticipated. This study differs from previous studies in the fraction of the test subjects still using illicit drugs. Additionally, the specimens used in this study consisted of the first urine and oral fluid specimens collected on the same day for each patient to help rule out any bias of including multiple specimens for the same patient. Lastly, when compared to other studies comparing urine and oral fluid drug matrices, the cut off levels are much lower than the industry standard cut off levels allowing for higher rates of positives [6] (Table 3).

The comparison of urine and oral fluid testing resulted in a lower Cohen’s Kappa coefficient meaning a ‘moderate’ agreement between urine and oral fluid matrices. This can be attributed to the many different factors in which equivalence between urine and oral fluid is not always expected; route of administration, free drug versus protein bound drug, lower cut off levels in oral fluid testing, inability to monitor adulteration, substitution, or diversion with the urine samples, and the difference in detection windows. In conclusion, analyses of urine and oral fluid specimens collected from addiction, family medicine, pain, and behavioural clinics concurrently provided a ‘moderate’ agreement between the two matrices, and that oral fluid testing can be used as a substitute to the gold standard of urine but may not act in full as a replacement.