Can a Carbon-Paste Electrode Detect THC in Saliva? Practical Versus Theoretical Limits

Short answer: A 2015 laboratory experiment detected deliberately added delta-9-tetrahydrocannabinol (THC) in commercial synthetic saliva using an optimized graphite–mineral-oil carbon-paste electrode and five minutes of preconcentration. The lowest THC concentration actually demonstrated in that saliva preparation was 0.50 micromolar (µM). The same research reported much smaller, mathematically extrapolated nanomolar detection limits for two aqueous buffer conditions—not for authentic human saliva. Those different claims must not be combined into a single clinical, roadside, or forensic performance figure. No person provided saliva, took THC, drove, or underwent an impairment assessment in this experiment.

The relevant original primary paper is Nissim and Compton, Absorptive stripping voltammetry for cannabis detection, Chemistry Central Journal 9:41 (2015), doi:10.1186/s13065-015-0117-0. Its figures establish a proof of analytical signal formation in prepared liquids. They do not establish that a positive field result could identify who consumed cannabis, when consumption occurred, whether a person was impaired, or whether cannabis contributed to a particular incident.

What did the investigators actually test?

Investigators Rita Nissim and Richard G. Compton built a working electrode from graphite powder and a liquid binder, then immersed the electrode in solutions with known added parent THC. The paste accumulated material during an open-circuit waiting interval. The electrode was subsequently moved into a separate blank supporting-electrolyte solution for an electrochemical scan. This is why the authors describe the technique as absorptive stripping voltammetry: the target is collected in the bulk paste and later produces an oxidation signal during measurement, rather than simply being detected directly in an unprepared oral-fluid sample.

The experimental materials were controlled laboratory reagents. One series involved aqueous borate buffer, another stirred that buffer, and a third employed commercial synthetic saliva mixed with borate buffer. The researchers studied analytical behavior under these three distinct configurations. They did not recruit cannabis users, collect saliva in a clinical setting, examine paired blood specimens, or conduct a diagnostic comparison against confirmed human samples. Consequently, the study’s sample size cannot be described as a cohort of patients or drivers. Error bars on its calibration figures refer to separate electrode preparations, whose total independent count is not specified as a participant sample.

The experiment is nevertheless useful: it asks whether a relatively uncomplicated carbon-paste electrode can generate a concentration-related response to a chemically specified cannabinoid in a controlled liquid. A method-development answer is valuable, provided it is not silently promoted into a validated human toxicology test.

Why graphite and mineral oil, rather than a conventional rigid electrode?

THC can undergo electrochemical oxidation, but oxidation products may foul some electrode surfaces and cause a declining or irreproducible response. A carbon paste provides a renewable analytical interface and a volume of binder into which hydrophobic analyte can partition. Nissim and Compton compared two graphite pastes, one using dioctyl phthalate and the other mineral oil, during an initial three-minute uptake experiment. Both produced THC-related oxidation peaks, but the graphite–mineral-oil preparation yielded the more favorable signal in the tested conditions and became the basis of later measurements.

The comparison used THC-containing stationary borate-buffer solutions in the approximately 7–80 µM range for the initial experiment; this was not a low-level saliva validation. Figure 1 documented oxidation peaks around +0.37 V for the mineral-oil paste and +0.39 V for the alternative binder, relative to a saturated calomel reference electrode. The choice of mineral oil is thus an observed laboratory configuration result, not proof that every mineral-oil formulation or commercially manufactured device would behave the same way.

The electrode surface was refreshed by cleaning its holder and packing new paste between scans. That important procedural detail explains why variability between preparations matters and why laboratory repeatability should not be mistaken for durability after hundreds of unreplaced roadside uses. A proposed field instrument would require separate evidence regarding storage, contamination, cross-reactive chemicals, temperature, operator variation, and recalibration.

What does five-minute preconcentration mean?

In the later experiments, the authors extended the electrode accumulation period to five minutes and increased the mineral-oil proportion of the paste while holding graphite mass constant. The electrode sat in the THC-containing source solution under open-circuit conditions, permitting THC to enter the paste. Only afterward was an electrical measurement carried out in a separate blank aqueous buffer containing potassium chloride. The five minutes were an accumulation interval, not a comprehensive sample-to-report turnaround time, and not a biologically measured interval after cannabis consumption.

The reported test setup also used controlled pH, a thermostated bath around 25 °C, a nitrogen atmosphere, and a three-electrode arrangement comprising the working carbon paste, a platinum counter electrode, and saturated calomel reference electrode. The square-wave scan extended approximately +0.20 to +0.60 V versus the reference, with a 100 Hz frequency, 1 mV potential step, and 40 mV amplitude. These conditions are relevant to replicability; removing the apparatus from a controlled laboratory would introduce questions that the source did not test.

Accumulation also has a limit. Once uptake into the paste no longer increases proportionally with THC concentration, the calibration response flattens. Longer accumulation is therefore not evidence of an unlimited, perfectly quantitative dynamic range. The experiment directly shows why any real reporting method needs defined lower and upper analytical boundaries.

What were the measured results in each liquid?

Three optimized configuration results are central: stationary aqueous buffer, stirred aqueous buffer, and stationary synthetic saliva. Their lowest practically demonstrated THC concentrations, concentration-response slopes, and calibration behavior differed. The following table preserves the actual liquid and measurement distinctions rather than presenting one pooled detection limit.

These values refer to parent THC fortified into prepared liquids, not to THC-COOH in urine or THC in actual blood after human use. The sensitivity values are slopes of electrochemical peak current against the nominal THC concentration. They are not diagnostic sensitivity percentages, and the R² values are fits in selected calibration regions rather than probabilities that a positive human test is correct. The original figures also reveal loss of proportional response at higher concentrations; those plateaus must be part of an accurate interpretation.

Why did stirring make the buffer result more sensitive?

Stirring replenished THC near the paste as material partitioned from the source liquid into the electrode. Under stationary aqueous conditions, nearby THC could become depleted during the five-minute period. Mixing reduced that local transport limitation and increased the slope of the peak-current response from 0.12 to 0.84 µA per µM. In the tested aqueous conditions, the lower practical detection point changed from 0.50 to 0.10 µM.

This does not mean stirred and stationary results are interchangeable. They are different procedures, and the lower 0.10 µM practical detection observation belongs to stirred buffer, not to the stationary synthetic saliva experiment. A description that says “the saliva sensor detected 0.10 µM THC” would move a result from one source matrix and mixing state into another without evidentiary support. The same problem would occur if a report treated the aqueous stirred calibration slope as a sensitivity measured in oral fluid.

It is also incorrect to assume more stirring necessarily improves every clinical specimen. Real oral fluid has variable viscosity, proteins, mucins, debris, pH, and sample collection differences that were not represented by this calibrated artificial liquid experiment. The direction of an effect in buffer supports a mass-transport explanation; it is not a substitute for a matrix-matched human validation program.

What is the difference between a practical limit of detection and an extrapolated theoretical limit?

This is the pivotal issue for forensic use. A practical detection level identifies the lowest concentration the investigators actually prepared and visibly detected under a specified experimental procedure. Nissim and Compton reported practical aqueous values of 0.50 µM without stirring and 0.10 µM with stirring. In contrast, they calculated much smaller theoretical LODs by projecting the response of a fitted calibration line toward a blank-signal noise threshold. Their formula was LOD = 3σ/s, with σ the blank signal’s standard deviation and s the calibrated current-versus-concentration slope.

Under stationary buffer conditions, the mathematical LOD was 0.48 nanomolar and the mathematical limit of quantification (LOQ), using 10σ/s, was 1.61 nM. Under stirred buffer conditions, calculated LOD and LOQ were 0.41 and 1.38 nM. The units matter: one micromolar equals one thousand nanomolar. Thus the lowest actually demonstrated stationary-buffer concentration, 0.50 µM, is approximately 1,042 times the corresponding 0.48 nM extrapolated LOD. The stirred-buffer practical 0.10 µM point is approximately 244 times its 0.41 nM theoretical LOD.

Those comparisons do not invalidate the calculation; they define its evidentiary status. A regression-based prediction beyond the observed concentration range is a modeled analytical estimate. It is not a demonstrated positive finding at the predicted concentration. In particular, a theoretical nanomolar buffer threshold must not be advertised as an experimentally verified nanomolar detection or quantification limit in genuine human oral fluid.

What did the saliva experiment actually establish?

The authors used commercially prepared synthetic saliva, not collected oral fluid. They fortified it with known concentrations of parent THC, adjusted the solution using their borate-buffer system at pH 10, held the paste electrode in the stationary preparation for five minutes, and subsequently recorded oxidation peaks in a separate electrolyte. Their Results report an experimental source range of 0.50 to 16 µM THC, a practical saliva detection point of 0.50 µM, and a calibration-response slope of 0.067 µA per µM. The peak current was proportional over approximately 0.50 to 7.0 µM, above which the response reached a plateau.

The source did not supply a separately calculated nanomolar theoretical saliva LOD or LOQ. The authors specifically did not calculate those values for the synthetic-saliva condition because of the limited linear range. The publication’s saliva experiment therefore cannot inherit the numerical theoretical LOD or LOQ values calculated for aqueous buffer. Nor can its 0.50 µM practical detection point be portrayed as a validated decision threshold for a person suspected of recent cannabis use.

The plotted Figure 4 caption describes a broader starting range than the Results prose. That reporting inconsistency does not supply a demonstrated lower saliva detection threshold; the authors’ stated practical saliva result remains 0.50 µM. For responsible use, the measured range and authors’ reported validated observation take precedence over an unsupported inference from a caption. A formal implementation would additionally need multiple authentic donor pools, interference challenges, independently verified concentration assignments, and prespecified acceptance criteria.

Does a high R² mean 99% diagnostic accuracy?

No. A calibration R² near 0.99 means that the fitted straight line explains most of the variation in the plotted calibration points within the fitted region, under the statistical model used. It does not mean a human oral-fluid test has 99% sensitivity, 99% specificity, 99% probability of a correct positive, or a 1% false-positive rate. Those are diagnostic-performance quantities that require tested positive and negative human specimens with an appropriate reference method and defined classification rules.

In this original investigation, the source concentrations were prepared by the experimenter, and the electrode output was compared against those known prepared values. The researchers did not report true-positive and false-positive counts in an independent population of actual clinical oral-fluid specimens. There was no receiver-operating-characteristic analysis, no patient-based threshold, and no assessment of how other cannabinoids or environmental chemicals might behave at a similar oxidation potential.

High linear fit within part of a calibration also says little about concentrations above the point of saturation. In synthetic saliva the signal began plateauing above approximately 7 µM. A concentration beyond the proportional interval might produce a similar peak height to a substantially lower concentration. Any attempt at quantification needs calibration-range restrictions and an explicit way to handle off-range specimens.

Where did the assay stop being proportional?

The investigators observed different upper ranges depending on how THC was accumulated. With stationary aqueous buffer, the calibration response increased through lower values and then leveled off around 5 µM. The stirred buffer experiment reached a plateau around 4 µM, despite showing a steeper current slope at the lower end. In stationary synthetic saliva, the current-response relationship was approximately linear to 7 µM and then flattened at higher additions. These are observed limitations, not insignificant details to omit from an optimistic sensor summary.

In mechanistic terms, a finite capacity for THC uptake into the paste is plausible: at some point additional THC in the surrounding solution no longer produces a proportional increase in the accumulated quantity available for oxidation. The paper interprets the plateau in that manner. It does not present a validated kinetic capacity model by which a high off-range specimen can be back-calculated precisely. The lesson is that detection of a peak is not identical to reliable quantification over every concentration.

For an evidentiary laboratory method, the reportable range should match actual standards, quality-control material, and the accepted calibration function in the same specimen matrix. A linear equation copied from the wrong part of a curve can generate unjustified numerical precision and misleading comparisons among individuals.

Which numerical findings were calculated rather than directly demonstrated?

To avoid mixing measured and modeled quantities, the following table separates the actual experimental lower concentration from the mathematical lower thresholds and the behavior that constrained quantitative use. The theoretical figures pertain only to the two aqueous-buffer series. Where the authors reported no derived saliva threshold, the correct entry is “not calculated,” not an estimate borrowed from another matrix.

Neither table contains a number of authentic tested drivers or a rate of field-test error because this original paper did not produce those data. The original publication also does not report inferential p-values assessing a real-patient positive-versus-negative diagnostic contrast. A lack of such outcomes is not a statistically nonsignificant clinical test; the clinical question was simply not investigated.

Is the analyte parent THC or the urinary metabolite THC-COOH?

The target in this experiment was parent delta-9-THC, the molecule intentionally added to aqueous and synthetic-saliva laboratory solutions. No separate calibration for 11-hydroxy-THC or 11-nor-9-carboxy-THC (THC-COOH) was established here. The reported peak around +0.37 V describes the electrochemical behavior of the fortified parent under specified conditions, not molecular confirmation of every cannabinoid that might be present in real oral fluid.

This distinction matters because specimen choice controls the question a toxicologist can address. Oral fluid is not interchangeable with venous whole blood, separated plasma or serum, urine, hair, vitreous humor, or an administered cannabis product. The collection environment and analytical chemistry differ, and a result in one matrix cannot be converted to a concentration in another using a generic factor. A complementary explanation of those distinctions appears in the independently published biological matrices guide.

A purported quantitative urine result for THC-COOH would require a method validated for THC-COOH in urine, rather than an electrode calibration for parent THC in artificial saliva. Even a measured parent THC value in genuine oral fluid would need its own interpretation; detection and a specific individual’s physiological impairment are separate propositions.

Was the electrochemical peak a confirmatory identification?

The study provides mechanistic and instrumental evidence that known added THC produced a reproducible oxidation-associated current in the tested systems. That is appropriate for a proof-of-concept electrochemical sensor. But an oxidation peak at an expected potential is not, by itself, a comprehensive test of analytical specificity against all possible substances in authentic human oral fluid. Other electroactive constituents, different cannabinoids, medication compounds, contaminants, or sample conditions could affect a candidate signal. The paper does not report a full independent selectivity panel sufficient to resolve that possibility.

A forensic confirmatory claim ordinarily requires a validated identification scheme for the molecule and physical specimen being reported, with appropriate calibrators, blanks, reference standards, interference assessments, and documented quality control. Electrochemical response in fortified synthetic saliva does not supply a patient-level identification result from a chemical reference method. Readers seeking the general distinction between a preliminary signal and a confirmed drug result can consult the already published confirmatory drug testing guide.

The original source’s comparison with detection limits in other published technologies appears in its Table 1, but those figures come from other investigators’ cited studies. They are not additional independently tested Nissim-and-Compton clinical samples, and they do not establish cross-method comparability when matrices, calibration assumptions, or identification rules differ.

Why might real human saliva behave differently from synthetic saliva?

Synthetic saliva permits the experimenter to control a starting matrix; that simplifies a feasibility experiment and reduces uncontrolled variation. Real oral fluid, however, is collected from people under changing physiological and procedural circumstances. Its flow rate, pH, protein and mucin content, debris, volume collected, and possible contamination can differ markedly from a laboratory surrogate. A prepared liquid’s performance cannot simply be carried forward as a claim about all donors.

The Nissim method also involved intentionally establishing alkaline pH and transferring the electrode into a fresh solution for scanning. Those handling steps should be described accurately when assessing feasibility. A marketed roadside instrument might have to combine collection, preconcentration, instrumental readout, specificity controls, cleaning or replacement, calibration, and legally adequate documentation. None of those end-to-end processes was prospectively evaluated with actual roadside use in this experiment.

Before collection results can be compared among individuals, a program would need to establish how and when specimens were collected, transported, stored, and analyzed. The published specimen collection guide provides contextual information, but it is not a substitute for validation of this particular electrode. Neither the source nor these contextual links should be read as a report that this carbon-paste assay has already met forensic clinical acceptance requirements.

Can this measurement establish recent use, impairment, or responsibility for a crash?

No. The study established an analytical signal from known THC additions in buffered and synthetic liquids. It did not study drug administration, human absorption, redistribution, metabolism, elimination, human oral-fluid concentration-time profiles, psychomotor performance, driving, or collisions. A five-minute electrode preconcentration is an equipment setting, not a five-minute interval since last consumption. A 0.50 µM detectable signal in synthetic saliva is not a universal impairment threshold.

Several scientifically distinct inferential steps would be required even after authentic human oral-fluid validation. A specimen would first need a reliable, documented, legally admissible identity and concentration result. That result would then have to be connected to a validated biological interpretation for its exact collection time and matrix. Any inference about individual functional impairment would require evidence appropriate to that person, circumstances, and task. A claim that the drug caused a collision, medical event, or alleged injury would demand a still separate causal analysis.

This paper supplies no individual-level probability of error for these downstream questions. Its absence of human subjects is a study-design boundary, not an attack on the potential usefulness of the electrochemical approach. Research into fast, inexpensive screening may be promising while remaining categorically different from validated forensic evidence in a litigated incident.

What would a defensible validation study need to add?

A next-stage study would need genuine human oral-fluid specimens from an appropriately described study population, including drug-negative samples and a range of independently confirmed THC concentrations. The comparator should identify the molecule with a suitable validated chemical method in the same or a demonstrably comparable physical specimen. Investigators would need to report how many individuals and unique specimens contributed to each analysis, rather than equating multiple scans or electrode preparations with independent biological observations.

At each relevant reporting threshold, a diagnostic study should quantify false-positive and false-negative results, sensitivity, specificity, and uncertainty, with explicit treatment of specimens near the decision boundary. It would also examine CBD, CBN, THC metabolites, other plausible electroactive interferents, food and oral contamination, and realistically varying pH and sample composition. Low-end calibration should actually include measurements near any claimed nanomolar working threshold instead of depending only on line extrapolation from much higher concentrations.

Practical controls would include lot-to-lot electrode variability, storage stability, transfer and deoxygenation requirements, number of scans achievable before repacking, field operators, temperature variation, carryover, upper-range saturation, and independent reproducibility. A method might ultimately become useful for a particular screening purpose, but the supported use and any required confirmatory pathway would have to be defined prospectively. The 2015 experiment neither supplies nor contradicts that later evidence.

Seven questions about the carbon-paste THC detection experiment

1. Was THC detected in actual saliva from cannabis users?

No. Investigators deliberately fortified commercial synthetic saliva with known parent THC concentrations. They did not collect oral fluid from living participants or establish how frequently a human saliva test would return an accurate positive or negative result.

2. What is the lowest THC concentration actually detected in synthetic saliva?

The source reports a practical lowest detectable value of 0.50 µM in stationary synthetic saliva under the specified pH-10, five-minute paste-accumulation protocol. The lower 0.10 µM practical value was obtained in a different condition: stirred aqueous buffer.

3. Did the authors prove that saliva THC could be detected at 0.41 or 0.48 nM?

No. Those two figures were calculated from buffer calibrations using blank-signal statistics and slopes. They were not experimentally demonstrated at those concentrations, and the paper specifically did not calculate equivalent nanomolar LODs or LOQs for synthetic saliva.

4. Does R² = 0.99 mean the saliva assay was 99% accurate?

No. It describes how well a linear equation fitted selected prepared calibration measurements. Human diagnostic accuracy requires a separate comparison of authentic positive and negative specimens against an appropriate reference method, with actual counts of errors.

5. Did the electrode measure THC-COOH in urine or THC in a driver’s blood?

Neither. The target was parent THC in fortified aqueous solutions and synthetic saliva. There was no experiment involving urine THC-COOH, directly collected human whole blood, measured THC in separated human plasma, or any individual driving case.

6. Why does the signal stop increasing proportionally at higher THC concentrations?

The study describes plateaus when the paste’s THC accumulation becomes limited. The approximate upper proportional ranges differed by condition: around 5 µM stationary buffer, 4 µM stirred buffer, and 7 µM synthetic saliva. Beyond a calibrated linear region, simple concentration extrapolation is not justified.

7. Can a positive reading from this prototype establish when THC was used or whether a person was impaired?

No. The experimental detection signal was developed in THC-fortified laboratory liquids, not in humans. The source contains no administered THC dose, verified last-use interval, impairment measurements, actual driving performance, or case-specific event-causation analysis.

Bottom line for forensic interpretation

Nissim and Compton showed that a graphite/mineral-oil carbon-paste electrode can accumulate and electrochemically signal parent THC under controlled conditions. Stationary synthetic saliva supported a demonstrated practical 0.50 µM detection point; separately studied stirred buffer reached 0.10 µM. Calculated buffer LODs of 0.48 and 0.41 nM are useful modeling outputs, but substantially below actually tested lowest concentrations and inapplicable as proven saliva performance.

The scientifically supportable conclusion is a laboratory sensor feasibility result with method-specific lower and upper calibration boundaries. It is not a clinical validation of oral-fluid cannabinoid testing and does not transform one oxidation peak into a finding about ingestion time, human intoxication, crash responsibility, or the causes of an individual event. The original evidence and its limits can be reviewed directly in Nissim and Compton’s primary research article.