Analysis of Drugs in Saliva of US Military Veterans Treated for Substance Use Disorders Using Supported Liquid Extraction and Surface-Enhanced Raman Spectral Analysis

According to the Center for Disease Control, there were more than 107,000 US drug overdose deaths in 2021, over 80,000 of which due to opioids. One of the more vulnerable populations is US military veterans. Nearly 250,000 military veterans suffer from substance-related disorders (SRD). For those seeking treatment, buprenorphine is prescribed to help treat opioid use disorder (OUD). Urinalysis is currently used to monitor buprenorphine adherence as well as to detect illicit drug use during treatment. Sometimes sample tampering occurs if patients seek to generate a false positive buprenorphine urine test or mask illicit drugs, both of which can compromise treatment. To address this problem, we have been developing a point-of-care (POC) analyzer that can rapidly measure both medications used for treatment and illicit drugs in patient saliva, ideally in the physi-cian’s office. The two-step analyzer employs (1) supported liquid extraction (SLE) to isolate the drugs from the saliva and (2) surface-enhanced Raman spectroscopy (SERS) to detect the drugs. A prototype SLE-SERS-POC analyzer was used to quantify buprenorphine at ng/mL concentrations and identify illicit drugs in less than 1 mL of saliva collected from 20 SRD veterans in less than 20 min. It correctly detected buprenorphine in 19 of 20 samples (18 true positives, 1 true negative and 1 false negative). It also identified 10 other drugs in patient samples: acetaminophen, amphetamine, cannabidiol, cocaethylene, codeine, ibuprofen, methamphetamine, methadone, nicotine, and norbuprenorphine. The prototype analyzer shows evidence of accuracy in measuring treatment medications and relapse to drug use. Further study and development of the system is warranted.

. SERS of Sample 1 first derivative (black) overlaid with (A) first derivative of buprenorphine (red), (B) overlaid with first derivative of nicotine (red). (C) Overlay of Sample 1 spectrum (red) on the predicted spectrum composed of 54% buprenorphine and 46% nicotine (black). (D) Overlay of Sample 13 spectrum (red) on predicted spectrum composed of 95% ibuprofen and 5% norbuprenorphine (black). Conditions: 100 mg/L sample in water, 40 mW at 785 nm, 1 sec acquisition.
Compared to urinalysis, SERS correctly identified buprenorphine in 19 of 20 samples; 18 true positives, 1 true negative (Sample 16), and 1 false negative (Sample 13), which contained norbuprenorphine. The following additional comparisons of SERS to urinalysis were noted (Table 1). Cannabinoids were detected by SERS in only four of the six positive urinalysis samples. Amphetamine was detected by SERS in Sample 3 and by urinalysis. Methamphetamine was detected in Sample 2 by SERS, but not by urinalysis. Methadone was detected in Samples 10 and 14 by SERS, but not by urinalysis. Yet methadone was detected by urinalysis in Sample 12, which was not detected by SERS. Cocaine was detected in Samples 17 and 20 by urinalysis, while it was detected as cocaethylene in Samples 7 and 20 by SERS. These discrepancies are likely due to the fact that urinalysis detects the benzoylecgonine metabolite, while SERS detects cocaethylene metabolite. In addition, their relative concentrations may be significantly different in saliva and urine.
Opioids were detected by SERS in three samples, but none matched the three samples that tested positive for opioids by urinalysis. This discrepancy and others indicated in Table 1 may be due to the differences in saliva versus urine metabolites, as well as relative concentrations. Compared to urinalysis, SERS correctly identified buprenorphine in 19 of 20 samples; 18 true positives, 1 true negative (Sample 16), and 1 false negative (Sample 13), which contained norbuprenorphine. The following additional comparisons of SERS to urinalysis were noted (Table 1). Cannabinoids were detected by SERS in only four of the six positive urinalysis samples. Amphetamine was detected by SERS in Sample 3 and by urinalysis. Methamphetamine was detected in Sample 2 by SERS, but not by urinalysis. Methadone was detected in Samples 10 and 14 by SERS, but not by urinalysis. Yet methadone was detected by urinalysis in Sample 12, which was not detected by SERS. Cocaine was detected in Samples 17 and 20 by urinalysis, while it was detected as cocaethylene in Samples 7 and 20 by SERS. These discrepancies are likely due to the fact that urinalysis detects the benzoylecgonine metabolite, while SERS detects cocaethylene metabolite. In addition, their relative concentrations may be significantly different in saliva and urine.
Opioids were detected by SERS in three samples, but none matched the three samples that tested positive for opioids by urinalysis. This discrepancy and others indicated in Table 1 may be due to the differences in saliva versus urine metabolites, as well as relative concentrations.

Buprenorphine Quantitation
The second objective required identifying the best buprenorphine spectral peak to use for quantitation. Two factors were examined: (1) the pH spectral dependence of buprenorphine and (2) the spectral interferences by the other drugs in the samples. The SERS

Buprenorphine Quantitation
The second objective required identifying the best buprenorphine spectral peak to use for quantitation. Two factors were examined: (1) the pH spectral dependence of buprenorphine and (2) the spectral interferences by the other drugs in the samples. The SERS intensity of drugs often has a pH dependence, due to the fact that protonated and deprotonated molecules interact with the surface plasmon of gold nanoparticles to varying degrees [38]. This dependence also influences the molecular-to-gold surface orientation and hence the relative intensities of the various functional group spectral peaks. Consequently, the pH dependence of buprenorphine samples at 100 µg/mL, adjusted from pH 5 to 8 by adding HCl or NaOH, was measured by SERS. The following SERS peaks were observed and have been assigned as follows: 505 cm −1 to a-/c-ring CH bending, 638 cm −1 to a-ring C=CH out-of-plane bending, 735 cm −1 to out-of-plane C=O bending, 835 cm −1 to c-ring CH bending, 1210 cm −1 to c-ring CCC out-of-plane bending, 1310 cm −1 to d-ring piperidine CH stretch, and 1590 cm −1 to c-ring CC stretching (Figures 4 and 5).
intensity of drugs often has a pH dependence, due to the fact that protonated and deprotonated molecules interact with the surface plasmon of gold nanoparticles to varying degrees [38]. This dependence also influences the molecular-to-gold surface orientation and hence the relative intensities of the various functional group spectral peaks. Consequently, the pH dependence of buprenorphine samples at 100 µg/mL, adjusted from pH 5 to 8 by adding HCl or NaOH, was measured by SERS. The following SERS peaks were observed and have been assigned as follows: 505 cm −1 to a-/c-ring CH bending, 638 cm −1 to a-ring C=CH out-of-plane bending, 735 cm −1 to out-of-plane C=O bending, 835 cm −1 to c-ring CH bending, 1210 cm −1 to c-ring CCC out-of-plane bending, 1310 cm −1 to d-ring piperidine CH stretch, and 1590 cm −1 to c-ring CC stretching (Figures 4 and 5).  It was found that the 835 cm −1 peak was the most intense at pH 7 and the 1210 cm −1 peak was the most intense at pH 6, while the 638 cm −1 peak, although modest in intensity, was relatively insensitive from pH 5 to 8 ( Figure 4A). An examination of the 36 drugs in Table 2 revealed only five drugs that contained peaks in the region of the 638 cm −1 buprenorphine peak; heroin, codeine, diazepam, norbuprenorphine, and hydrocodone ( Figure  4B). These drugs were included in the analysis of the 20 samples described above. Of these drugs, only norbuprenorphine and codeine were detected in the 20 samples. Norbuprenorphine was detected in Samples 5, 6, 8, and 13. There is no confusion that both bupren- intensity of drugs often has a pH dependence, due to the fact that protonated and deprotonated molecules interact with the surface plasmon of gold nanoparticles to varying degrees [38]. This dependence also influences the molecular-to-gold surface orientation and hence the relative intensities of the various functional group spectral peaks. Consequently, the pH dependence of buprenorphine samples at 100 µg/mL, adjusted from pH 5 to 8 by adding HCl or NaOH, was measured by SERS. The following SERS peaks were observed and have been assigned as follows: 505 cm −1 to a-/c-ring CH bending, 638 cm −1 to a-ring C=CH out-of-plane bending, 735 cm −1 to out-of-plane C=O bending, 835 cm −1 to c-ring CH bending, 1210 cm −1 to c-ring CCC out-of-plane bending, 1310 cm −1 to d-ring piperidine CH stretch, and 1590 cm −1 to c-ring CC stretching (Figures 4 and 5).  It was found that the 835 cm −1 peak was the most intense at pH 7 and the 1210 cm −1 peak was the most intense at pH 6, while the 638 cm −1 peak, although modest in intensity, was relatively insensitive from pH 5 to 8 ( Figure 4A). An examination of the 36 drugs in Table 2 revealed only five drugs that contained peaks in the region of the 638 cm −1 buprenorphine peak; heroin, codeine, diazepam, norbuprenorphine, and hydrocodone ( Figure  4B). These drugs were included in the analysis of the 20 samples described above. Of these drugs, only norbuprenorphine and codeine were detected in the 20 samples. Norbuprenorphine was detected in Samples 5, 6, 8, and 13. There is no confusion that both bupren- It was found that the 835 cm −1 peak was the most intense at pH 7 and the 1210 cm −1 peak was the most intense at pH 6, while the 638 cm −1 peak, although modest in intensity, was relatively insensitive from pH 5 to 8 ( Figure 4A). An examination of the 36 drugs in Table 2 revealed only five drugs that contained peaks in the region of the 638 cm −1 buprenorphine peak; heroin, codeine, diazepam, norbuprenorphine, and hydrocodone ( Figure 4B). These drugs were included in the analysis of the 20 samples described above. Of these drugs, only norbuprenorphine and codeine were detected in the 20 samples. Norbuprenorphine was detected in Samples 5, 6, 8, and 13. There is no confusion that both buprenorphine and norbuprenorphine are present, because the analysis employs the full spectrum (400 to 1800 cm −1 ) and each drug has unique peaks that were detected: 835 and 1445 cm −1 for buprenorphine and 1015, 1135, and 1635 cm −1 for norbuprenorphine. Codeine was detected in Sample 9 and also has spectral peaks that differentiate it from buprenorphine at 535 and 1255 cm −1 . Consequently, the 638 cm −1 peak was used to quantify buprenorphine in the samples. Next, buprenorphine samples were prepared in purchased, deidentified pooled saliva at 6.25, 12.5, 25, 50, 100, and 200 ng/mL, extracted and reconstituted as described in Method 3, and measured by SERS to produce a calibration curve. The baseline of the spectra was set to 0 at 665 cm −1 , and the 638 cm −1 peak height was plotted as a function of the prepared concentrations and fit with a straight line: [BUP] = 0.029 × Peak Height + 0.19, with an R 2 value of 0.99 ( Figures 5C and S2).
The equation was then used to calculate the concentration for each of the 20 samples using the SERS measured and baseline-corrected: 638 cm −1 peak height (Table 1, right-most column, and Figure S3). The concentrations for samples containing norbuprenorphine and codeine, as described above, were corrected using the spectral fit percent results for these samples. For example, the Sample 5 buprenorphine concentration was reduced from 39 to 35 ng/mL, since the 638 cm −1 peak was composed of 89% buprenorphine. Similarly, for Sample 9, containing codeine, the buprenorphine concentration was reduced from 80 to 45 ng/mL (56%). The uncorrected concentrations are shown in parentheses in Table 1.

Analytical Figures of Merit
The analysis reproducibility, limit of detection (LOD), and limit of quantitation (LOQ) were also determined for the measurement procedure. Analysis reproducibility, encompassing sample preparation, extraction, reconstitution, and measurement, was determined by measuring nine independently prepared 50 ng/mL buprenorphine samples, consisting of 40 µL drops,~5 mm in diameter, deposited on glass slides. It was found that the percent standard deviation for the 638 cm −1 peak height was 3.5% (Figures 6 and S4, Table 3). Note that SERS measurement repeatability, performed by measuring nine positions of a single 50 ng/mL, 40 µL drop, yielded a percent standard deviation of~1%, indicating that the colloid and buprenorphine were evenly distributed in the sample.  (Figure 7). The latter spectral region was chosen as it contains only background noise, while the width was selected to match the peak width. The 50 ng/mL sample was used for the calculation. The signal was taken as the 638 cm −1 peak height, baselinecorrected from 610 to 665 cm −1 (S = 1705), and the root mean squared (rms) noise (15.6) between 1810 and 1865 cm −1 was used (Figure 7). The latter spectral region was chosen as it contains only background noise, while the width was selected to match the peak width.

Discussion
The main goal of this study was to test the ability of an SLE-SERS-POC prototype analyzer to determine patient compliant use of buprenorphine. In this regard, the prototype performance was very good, correctly identifying 18 adherent patients and 1 not adherent patient to medication in agreement with urinalysis. The prototype only misidentified one patient sample as nonadherent who was adherent according to the urinalysis results. It is possible that the patient spiked their urine sample with buprenorphine. While the prototype provided quantitative buprenorphine concentration for all of the samples, there was no relation between the patient SERS-based saliva concentrations and their administered dosage. This could be caused by two factors: variability of the patient's metab-

Discussion
The main goal of this study was to test the ability of an SLE-SERS-POC prototype analyzer to determine patient compliant use of buprenorphine. In this regard, the prototype performance was very good, correctly identifying 18 adherent patients and 1 not adherent patient to medication in agreement with urinalysis. The prototype only misidentified one patient sample as nonadherent who was adherent according to the urinalysis results. It is possible that the patient spiked their urine sample with buprenorphine. While the prototype provided quantitative buprenorphine concentration for all of the samples, there was no relation between the patient SERS-based saliva concentrations and their administered dosage. This could be caused by two factors: variability of the patient's metabolism, or more likely, the time the saliva samples were collected with respect to when the patient took their dose. It has been shown that the saliva concentration is as high as 1000 ng/mL within the first hour after sublingual administration of a 1 mg tablet and does not reach a steady state until 10 h after administration due to "holding" in the oral cavity and buccal permeability [33,39] In fact, some patients supplied saliva samples within an hour after taking a dose. Furthermore, the saliva concentration is relatively stable from 10 to 24 h, suggesting that the best time to perform a saliva measurement would be right before the tablet is administered. In addition, measurement of a patient at the same time for several days would be advantageous to setting dosage based on metabolism and thereby improve patient performance. It is also worth noting that the prototype LOD, LOQ, and R 2 values are similar to LC-MS/MS measurements of buprenorphine in saliva at 5 ng/mL, 10 ng/mL, and 0.9986, respectively [23].

Materials 1: Purchased Materials
All chemicals and solvents used to prepare samples, colloids, and perform extractions were obtained from Sigma-Aldrich (St Louis, MO, USA). The drugs used to prepare the spectral library were purchased as 1 mg/mL methanol forensic samples from the same supplier (Table 2)

Materials 2: Prepared Materials
The gold colloid solution used for SERS measurements was synthesized following a modified Lee-Meisel method [40]. Briefly, 240 mg of gold chloride (HAuCl4•3H2O) was dissolved in 500 mL of water and heated to 100 °C, at which temperature 50 mL of 1% sodium citrate was added and then allowed to boil for 1 hr. The gold colloids have a shelf life of over a month and were prepared in advance. The forensic drug samples were diluted to 100 µg/mL using distilled water. Twenty microliter aliquots of these diluted drug

Materials 2: Prepared Materials
The gold colloid solution used for SERS measurements was synthesized following a modified Lee-Meisel method [40]. Briefly, 240 mg of gold chloride (HAuCl 4 •3H 2 O) was dissolved in 500 mL of water and heated to 100 • C, at which temperature 50 mL of 1% sodium citrate was added and then allowed to boil for 1 hr. The gold colloids have a shelf life of over a month and were prepared in advance. The forensic drug samples were diluted to 100 µg/mL using distilled water. Twenty microliter aliquots of these diluted drug samples were mixed with 20 µL of the gold colloids. Similarly, a concentration series of buprenorphine was prepared by diluting a 1 mg/mL methanol forensic sample to 5 ng/mL in distilled water. For each concentration, 200 µL of buprenorphine in water was added to 200 µL of deidentified pooled saliva.

Materials 3: Patient Samples
VA Connecticut Healthcare System (VACHS, West Haven, CT, USA) patients being treated for OUD using buprenorphine who were already providing urine samples for analysis as part of a larger study were recruited to provide saliva samples in accordance with IRB Protocol 00008942 (Chesapeake IRB, Inc., Columbia, MD, USA). Twenty volunteer patients went through an informed consent process, stated that they understood this study, and signed the consent and Health Insurance Portability and Accountability Act documents. Most of the patients were taking buprenorphine for at least 2 weeks prior to providing a saliva sample, which was collected within 2 h of the urine samples. The patients also provided information regarding drug use for the previous 2 weeks. Buprenorphine was administered sublingually once or twice a day as Suboxone containing 2, 4 or 8 mg of buprenorphine and 0.5, 1 or 2 mg naloxone, respectively. Total daily patient doses were 8, 12, 16, 20 or 24 mg buprenorphine. Ten minutes prior to saliva sample collection, the patients were instructed to rinse their mouth out with bottled water. Sample collection was performed by spitting into plastic tubes until 1 to 2 mL of saliva was obtained. Everyone on buprenorphine had their dose the day the sample was collected and took a dose within a few hours before the saliva sample was collected. The saliva samples were sealed and frozen until saliva analysis was performed at Real-Time Analyzers (RTA).
The general demographics for the 20 patients were as follows (Table 4): all male, 12 Caucasian, 5 African American, 2 Hispanic, and 1 declined; 2 under age 40, 6 between 40 and 49, 4 between 50 and 59, and 7 between 60 and 65 years of age. The 20 samples were all collected over the course of one week. Table 4. Demographic information for 20 enrolled patients (see Table S1 for additional details).

Method 1: Urine Toxicology
Urine samples provided by patients were delivered to an on-site VACHS clinical laboratory. The samples were analyzed by a standard multiplexed sample and reagent handling system coupled to an electrogenerated chemiluminescence analyzer (e.g., Roche Hitachi 6001). Immunoassays consisted of standard ruthenium functionalized drug-specific antibodies on magnetic beads such that separation could be accomplished using an electrode to capture the beads and generate chemiluminescence. Magnetic beads, functionalized with a DNA sequence, were used to bind buprenorphine instead of an antibody, as used for all other drugs. The intensity of the luminescence signal was compared to a positive cut-off calibrant sample selected for each of the 9 drugs (Table 5).

Method 2: Liquid Extraction
A 200 µL saliva sample mixed with 200 µL of distilled water was added to a SLE column attached to the vacuum pump. The sample was adsorbed onto the support by applying a negative pressure of 15 inch of Hg for 1 sec. After a 5 min wait, 2 sequential aliquots of 900 µL dichloromethane were drawn through the support, first using gravity for 5 min, then again using −15 inch of Hg for 1 min. The collected sample was dried under a gentle stream of nitrogen for 5 min, reconstituted using 40 µL of the gold colloid solution, of which a 10 µL drop was deposited a onto a glass slide, which in, turn was placed in the sample compartment of the Raman spectrometer for analysis.

Method 3: Raman Spectroscopy
The 5 µL aliquots of library drugs, buprenorphine concentration series, and patient saliva samples mixed with 5 µL of the gold colloids were each deposited on glass slides and placed into the sample compartment above the focal point of the 785 nm laser of the Raman spectrometer (Wasatch Photonics, model FPR-785-WS, Orlando, FL, USA). Spectra were recorded from~200 cm −1 to 2300 cm −1 with~20 cm −1 resolution (peak width at half height), and peak positions are reported to the nearest 5 cm −1 (except the 638 cm −1 buprenorphine peak). A laptop computer was used to control the laser power, acquisition time, and perform spectral analysis. Each spectrum consisted of a 1 sec acquisition using 40 mW of 785 nm laser excitation focused to~200 µm at the sample. RTA's Chem-ID and S-Quant software were used to identify drugs and quantify buprenorphine in the samples, respectively. A complete surface-enhanced Raman spectral analysis of the drugs described here has been published [32].

Conclusions
The SLE-SERS-POC prototype analyzer successfully quantified buprenorphine in samples in less than 20 min for 20 VA patients. A simple supported liquid extraction method was successfully developed to isolate the drugs from the saliva samples for analysis by surface-enhanced Raman spectroscopy. Semiautomated spectral analysis, employing a spectral library, identified 25 of 30 drugs in the samples. This included the identification of six drugs without prior knowledge of their presence in the 20 saliva samples. Furthermore, the presence of these additional drugs did not interfere with the measurements. Nevertheless, the method could be improved by using a library that includes only those drugs that could be reasonably expected in patient samples. We believe that an SLE-SERS-POC production analyzer could greatly improve patient adherence by eliminating drug spiking and aid physicians in setting dosage and monitoring buprenorphine and thereby improve treatment. Future work will focus on determining the best method to collect saliva, (e.g., passive drool or swab) and automating the sample extraction and analysis software. A next-stage prototype will be used to perform initial clinical trials that include comparing