EVIDENCE LIBRARY
Can Portable Raman Detect THC in Saliva but Miss It in Urine Without Separation?
A 2019 primary study showed why urine urea masked THC-associated Raman signals and how TLC separation helped. It did not validate impairment or forensic accuracy.
Short answer: A 2019 laboratory investigation demonstrated THC-associated surface-enhanced Raman signals in two saliva samples obtained shortly after reported cannabis use, but untreated urine produced strong interfering signals. The researchers could observe THC-associated features in prepared urine after thin-layer chromatographic separation. That is a promising analytical proof of concept, not a validated roadside THC test, a forensic confirmation procedure, or a method for determining whether a person was impaired.
The crucial finding concerns specimen-matrix interference. A technique that produces a recognizable chemical spectrum in one fluid may yield an uninterpretable spectrum in another, even when the target molecule is present. The experimental source is Sivashanmugan, Zhao, and Wang, Biosensors (2019), DOI: 10.3390/bios9040125.
What scientific question did the investigators actually test?
They investigated whether a portable Raman spectrometer, coupled with a gold-nanoparticle-coated diatomaceous substrate, could recognize THC-associated spectral features in saliva and urine. For urine, they additionally tested whether a simple thin-layer chromatography (TLC) step could separate the THC-associated signal from abundant background substances. The experiment addressed feasibility of a chemical sensing platform, not the clinical validity of a diagnostic decision rule.
That distinction matters when an instrument is described as “portable.” Portability means the optics can be taken to a sample. It says little, by itself, about a validated limit of detection in authentic specimens, analytical specificity, collection-to-result turnaround, chain-of-custody performance, or acceptable error rates in a legal proceeding. Those claims require separate evidence.
Was this a controlled human dosing trial?
No. The research combined reference THC solutions, urine deliberately fortified with THC, and biological samples collected from volunteers. The publication explicitly describes two cannabis-user saliva samples in its spectral figure. It also describes a cannabis-user urine example but does not provide an independently powered, demographically characterized human diagnostic cohort or a clear urine-donor denominator for estimating population-level sensitivity and specificity.
Saliva was collected within approximately 15 minutes of reported cannabis use. The researchers diluted and centrifuged it before analysis. Urine specimens were obtained from reported users and nonusers, with nonuser urine deliberately spiked for concentration experiments. The resulting comparisons should not be construed as dozens of independently collected positive and negative patient samples merely because numerous spectra were recorded.
The investigators did not administer a controlled known dose to participants, quantify systemic absorbed THC in each person, or measure driving, attention, balance, or other impairment endpoints. An exposure-associated spectrum therefore cannot be translated into a dose-response relationship.
How does Raman spectroscopy differ from conventional forensic chromatography?
Raman spectroscopy measures changes in light scattered by molecular bonds and produces a pattern of spectral shifts, generally reported in inverse centimeters (cm−1). Surface-enhanced Raman scattering (SERS) increases the detectability of weak signals by placing molecules near engineered metallic nanostructures. In this study the sensing surface incorporated gold nanoparticles and porous silica derived from diatoms.
By contrast, a routine confirmatory forensic procedure often identifies compounds after chromatographic separation using a mass-selective detector and validated identification criteria. Raman spectral similarities are chemically informative, but a visually recognizable peak is not automatically equivalent to a validated unique molecular identification in a complex specimen. Overlapping peaks, sample preparation, instrument settings, and reference materials remain important.
For a discussion of the differences between screening indications and defensible drug identification, see confirmatory drug testing. The new Raman work did not itself establish equivalence to an accredited confirmatory LC–MS/MS or GC–MS protocol.
What did the portable instrument and substrate actually do?
The authors employed a portable 785-nm Raman instrument with a stated 420-mW excitation laser, a 5-mm optimized working distance, and 10-second spectral integration; they reported averaging 20 consecutive acquisitions. The gold nanoparticles were approximately 60 nm in diameter, while the diatomaceous particles were roughly 10–25 micrometers across. These physical design choices were intended to concentrate analyte and boost weak Raman scattering.
For reference THC material, a prominent band appeared near 1601 cm−1 and other associated bands contributed to a more complete fingerprint. The investigators estimated an enhancement factor of about 1.4 × 106 for gold nanoparticles on diatomite, compared with about 1.2 × 102 for gold nanoparticles alone under their reference conditions.
Those enhancement factors describe signal amplification in an experimental optical system. They are not specimen-specific limits of detection, percentages of THC recovered, clinical diagnostic accuracy, or probabilities that a particular person’s urine contains THC.
Why were saliva spectra easier to interpret?
Figure 3 of the original paper displays spectra from two saliva specimens collected after cannabis use. A THC-associated peak near 1601 cm−1 was recognizable in both. The authors also discussed a feature near 1681 cm−1, which they associated with a carboxyl-containing metabolite structure. They did not provide orthogonal reference-standard identification of a specific metabolite in those exact biological specimens. Assignment of that band is thus not independently confirmed metabolite speciation.
Importantly, the researchers did not simply put unprocessed whole oral fluid directly into the instrument. Their methods included dilution and centrifugation to remove debris. The meaningful contrast is that saliva did not require the additional TLC separation used for urine in this experimental configuration. It is not proof that all saliva samples are interference-free.
Saliva can also contain recent oral contamination after smoking or consuming cannabis. A chemical signal in oral fluid does not establish an independently sampled venous whole-blood THC concentration, a specific amount absorbed into the central circulation, or a person’s psychomotor condition at the time of a traffic event.
Why did direct urine analysis fail?
Urine contains substantial quantities of urea and other endogenous compounds that interact with the SERS surface. In the initial untreated-urine analysis, the recorded spectra were dominated by background bands associated with the urine matrix. The researchers reported that direct SERS testing could not reliably reveal the sought THC bands in either THC-spiked urine or a user’s urine under those conditions.
This is a basic lesson in analytical toxicology: failure to observe a target signal in an interfering matrix is not evidence that the target is absent. A negative or obscured optical reading might reflect competition for the sensing surface, an unresolved spectral overlap, inadequate separation, low analyte abundance, or an otherwise unsuitable method. Those possibilities cannot be distinguished merely by looking at one broad peak.
The specimen distinction is also critical. Urine testing generally addresses excreted analytes or metabolites, whereas whole-blood testing addresses a different biological compartment and a different legal question. Additional context appears in biological matrices in forensic toxicology.
What changed after thin-layer chromatography?
The team used porous diatomaceous-earth TLC channels to move components of a small urine aliquot through a stationary phase before acquiring spectra from the separated region. They compared channel widths of 1, 2, and 3 mm. The 1-mm channels yielded the strongest THC-associated signal under the reported test conditions and were selected for subsequent experiments.
After TLC, THC-fortified urine produced a clearer feature near 1605 cm−1, along with additional assigned bands. Figure 5 also shows THC-associated signals near 1003 and 1603 cm−1 in a cannabis-user urine example after separation; before TLC, urea-associated background dominated. Thus the study’s strongest urine result was not just the instrument hardware, but the combination of physical separation and optical measurement.
The paper presents apparent separation positions and spectral traces, not a complete forensic validation of chromatographic resolution, fraction purity, analyte identity in authentic urine, or recovery over a large variety of donors and interferents. The demonstration should therefore be described as improved visibility of assigned THC-related features, not unequivocal chemical confirmation in every future urine sample.
Which results were actually numerical, and what do they mean?
| Reported experiment or measurement | Original result | What it supports—and does not support |
|---|---|---|
| THC reference solutions in methanol | Concentrations from 1 to 1,000 ppm were explored; the main associated band appeared near 1601 cm−1. | Reference spectral behavior on this substrate, not a validated authentic-specimen sensitivity estimate. |
| Substrate signal enhancement | Estimated enhancement factors: approximately 1.4 × 106 for gold-on-diatomite versus 1.2 × 102 for gold alone. | Relative optical enhancement under experimental assumptions, not population diagnostic accuracy. |
| Saliva from reported cannabis users | Two separately depicted saliva samples showed a THC-associated band near 1601 cm−1. | Feasibility in those two specimens, not a two-person controlled-dose trial or sensitivity validation. |
| Laboratory-fortified urine after TLC | THC was added at 10–1,000 ppm; spectra after separation showed the assigned feature near 1605 cm−1. | Demonstration at fortified concentrations, not forensic detection of low-ng/mL authentic urine. |
| Raman acquisition | 10-second integrations averaged from 20 measurements, with a 785-nm laser. | Instrument settings and repeated optical observations; not 20 different people or total field turnaround time. |
Concentration units are especially important. A ppm concentration in a water-like laboratory solution is on the order of milligrams per liter; 10 ppm is approximately 10,000 ng/mL under that convention. That conversion illustrates scale only: actual specimen density and preparation must be accounted for, and the experiment did not establish a formal validated urinary lower limit of quantitation at 10 ppm.
Did principal component analysis validate a THC screening cutoff?
No. The authors used principal component analysis (PCA) to describe spectral patterns from six specimen or preparation categories, including reference THC, user saliva, control saliva, urine, spiked urine, and user urine. They reported that PC1 explained approximately 98.25% of the variance in the analyzed spectra, with PC2 about 0.75% and PC3 about 0.50%.
That 98.25% is a variance-explained statistic. It is not 98.25% correct identification, sensitivity, specificity, predictive value, accuracy, or the proportion of impaired subjects identified. The authors used approximately 20 spectra per sample type for this visualization. Multiple spectra from related preparations are not independent biological donors; PCA clustering in a development dataset does not substitute for blinded external validation.
Figure 6 shows visually separated groupings but does not supply a prespecified, externally tested clinical threshold with a confusion matrix. A future field instrument would need independent studies of near-cutoff positives, drug-free specimens, closely related cannabinoids, mixed drugs, medication matrices, day-to-day substrate batches, and other environmental sources of variability.
What was the study’s biological denominator?
The relevant independent biological units are actual collected donor specimens—not the number of laser scans, spectrum points, TLC channels, microscopic images, or PCA markers. Two cannabis-user saliva specimens were explicitly plotted. For the urine feasibility portion, the available original text does not identify a sufficiently detailed independent participant denominator, sampling scheme, and paired reference-test results to calculate a defensible clinical false-negative or false-positive percentage.
That missing information has practical consequences. Even if all two depicted saliva samples contain a recognizable band, it would be invalid to report 100% real-world detection sensitivity. The experiment did not establish how often an untreated specimen masks THC in a large unselected driver population or how an optical result correlates with a confirmatory assay.
How should a forensic expert separate the different claims?
| Claim in a forensic dispute | Evidence provided here | Evidence still needed |
|---|---|---|
| “This optical platform can produce THC-associated signals.” | Yes, for reference preparations and selected saliva and TLC-treated urine examples. | Reproducibility across batches and authentic diverse matrices. |
| “A negative direct-urine spectrum rules out THC.” | No. Strong urea/matrix interference obscured target-associated bands. | Appropriately validated testing and, where indicated, confirmation. |
| “A 1681-cm−1 signal uniquely proves a specific THC metabolite.” | No. The authors proposed a functional-group assignment without authentic-specimen orthogonal structural confirmation. | Reference standards, additional identification criteria, and validated specificity. |
| “PCA explained 98.25%, therefore the test is 98.25% accurate.” | No. Explained spectral variance is not a diagnostic classification metric. | Independent blinded positives/negatives and a prespecified operating rule. |
| “A Raman result proves impairment or time of use.” | No human driving, dose-time reconstruction, or psychomotor outcome was tested. | Incident-specific evidence and independently validated causal/functional interpretation. |
Can this Raman method replace mass-spectrometric confirmation?
The publication does not support that conclusion. The study’s most compelling result is identifying a specific problem with complex specimens and showing that a simple separation step changes the observed spectral pattern. It does not provide the comprehensive method validation ordinarily necessary for evidentiary confirmation: cross-reactivity challenges, specimen-specific calibrators and quality controls, interference panels, carryover, within- and between-day imprecision, false-positive and false-negative rates, robust analyte identification, and demonstrable blind performance.
Nor did it establish quantitative comparability to GC–MS or LC–MS/MS on matched human specimens. A SERS-assigned spectral feature could be a useful screening or research observation while still requiring a separate validated analytical method for a legally consequential identification or concentration.
Good practice begins with the requested forensic question, the sample actually collected, collection and storage conditions, and the analytical decision rule. The complementary discussion of specimen collection and its limitations explains why an instrument’s laboratory performance cannot rescue an unsuitable or unverified specimen history.
What does the work imply for driving, exposure timing, and individual causation?
Very little directly. No subject performed a standardized field sobriety test, simulator drive, divided-attention examination, real-road driving task, or clinical impairment assessment in the reported Raman experiment. The investigators did not obtain contemporaneous venous whole-blood THC values that could be compared with the oral-fluid or urine signals. A recent saliva-associated spectrum could reflect oral-cavity deposition as well as other exposure pathways; urine signals raise different questions about excretion and analytical specificity.
Consequently, even a successfully confirmed THC finding would not, by itself, prove that an individual was impaired when driving, establish the precise moment of use, reconstruct an earlier blood concentration, or identify the cause of a collision. Each proposed inference needs its own validated scientific and evidentiary foundation. This proof-of-concept paper is a reason to evaluate analytical methods carefully, not a substitute for case-specific pharmacology and observed evidence.
Frequently asked questions
1. Did the researchers directly test whole-blood THC?
No. Their new portable sensing experiments involved reference solutions, saliva, urine, and THC-fortified urine. Blood concentrations cited in introductory background were not new whole-blood measurements made for this sensing study.
2. Does a clear Raman peak near 1601 cm−1 uniquely confirm THC?
No. It is consistent with an assigned THC-associated feature in the study’s reference framework, but authenticated biological specimens may contain overlapping or interfering bands. The paper did not establish fully validated confirmatory identification criteria.
3. Why was an extra chromatography step required for urine?
Direct urine SERS spectra were dominated by endogenous compounds, notably urea. Separation improved visibility of assigned THC-associated bands in fortified and selected authentic samples.
4. Did the study prove detection at ordinary low-ng/mL forensic cutoffs?
No. The reported urine spike experiments covered 10–1,000 ppm, a much higher concentration scale than many forensic cannabinoid thresholds. They did not establish a validated near-cutoff clinical limit of detection or limit of quantitation.
5. Were 20 Raman spectra equal to 20 independent cannabis users?
No. The paper describes repeated spectral measurements, and only two user saliva examples are explicitly shown. Technical replicates are not independent human participants.
6. Is 98.25% PCA variance the same as 98.25% THC-test accuracy?
No. Variance explained by a statistical component does not tell how often a test correctly classifies independent positives and negatives.
7. Could this device establish that a defendant was impaired at a specific time?
No. The study did not evaluate driving, human impairment, controlled individual dose-time curves, or validated legal thresholds. A separate case-specific assessment would be required.
Bottom line for interpreting the original experiment
Portable diatomaceous SERS produced THC-associated spectral signals in two reported cannabis-user saliva examples and, after TLC separation, in THC-fortified and selected user urine examples. The negative direct-urine result was a matrix-interference problem, not a demonstrated absence of THC. The study offers a plausible sensing architecture but not clinically established accuracy, confirmatory chemical identification, or impairment evidence.
Primary research: Sivashanmugan K, Zhao Y, Wang AX. Tetrahydrocannabinol Sensing in Complex Biofluid with Portable Raman Spectrometer Using Diatomaceous SERS Substrates. Biosensors. 2019;9(4):125. https://doi.org/10.3390/bios9040125.