Can a Sativex Registry Prove THC Does Not Impair Driving? What 941 Prescriber Reports Actually Show

Short answer: No. A postmarketing registry of 941 patients prescribed the THC–CBD oromucosal medicine Sativex recorded 63 patients whose driving ability was described by prescribers as improved and 19 as deteriorated. But 514 records placed the driving question in a “not appropriate” category, another 40 had no recorded answer, and the study did not perform standardized on-road driving, simulator, psychomotor, or chemical blood-testing assessments. The numbers describe reports entered by clinicians; they do not determine whether THC improved driving, whether a particular patient was unimpaired, or whether a particular collision was caused by treatment.

The underlying primary research is Etges and colleagues’ 2016 observational safety registry of nabiximols in the United Kingdom, Germany, and Switzerland: An observational postmarketing safety registry of patients prescribed Sativex (THC:CBD) oromucosal spray. The same dataset also recorded clinically significant falls, psychiatric symptoms, adverse events, treatment continuation, and perceptions of benefit. It is useful for identifying what clinicians noticed during routine prescribing. It was not designed to calculate a causal effect of THC on road safety.

What question did this registry actually investigate?

The registry followed patients who had already been prescribed a commercial medicine containing both delta-9-tetrahydrocannabinol (THC) and cannabidiol (CBD). Its purpose was long-term pharmacovigilance: finding potential safety signals in ordinary clinical use after marketing authorization. That objective differs from testing whether a controlled THC dose alters lane position, reaction time, divided attention, or the ability to drive safely.

Prescribers were periodically invited to submit anonymized patient information. The case report forms requested clinically significant adverse events and answers concerning issues of special interest, including falls requiring medical attention, serious psychiatric symptoms, suicidality, suspected misuse, and changes in driving ability. Participating physicians volunteered their reports retrospectively, and the manufacturer maintained the registry. There were no patient-completed questionnaires, no assignment to a noncannabinoid control group, and no blinding of the prescribing physician.

This distinction determines the permissible inference. A prospective controlled driving experiment tests performance under defined conditions. A passive or semistructured postmarketing registry shows what was documented within a clinical reporting system. It cannot, without additional evidence, sort improvement from changes caused by the underlying disease, symptom relief, clinician expectations, selective entry, or natural variation over time.

Who was included—and who was not?

The registry contained 941 distinct patients: 761 from the United Kingdom, 178 from Germany, and two from Switzerland. Women accounted for 540 and men for 401. Age at treatment initiation was available for only 852 patients, with a reported mean of 51.2 years. These are patient counts, not independent observations for each twice-yearly collection period; the same person could appear on more than one reporting form.

Multiple sclerosis was the principal prescribing context. The investigators described 729 patients as treated for an MS-related indication, 125 for other indications and 87 with an indication not recorded. The cohort is therefore not a random sample of motorists, healthy recreational cannabis users, arrestees, crash-involved drivers, or people taking inhaled cannabis products. The natural history of MS, including mobility limitations, fatigue, spasticity, medication changes, and varying ability to drive at all, complicates any comparison of performance over time.

In the UK, only 761 of 3,493 eligible treated patients appeared in the registry—approximately 22%. The denominator indicates substantial potential for selection. Clinicians could select which patients to report; no validated procedure establishes that unreported patients resembled reported ones. Even an apparently large registry can have major ascertainment bias when participation is discretionary.

How much THC and CBD were prescribed, and was either measured in blood?

Each 100-microliter Sativex spray actuation was described as delivering 2.7 mg of THC plus 2.5 mg of CBD. Daily-dose information was recorded for 798 patients. Among these records the median dose was 4.4 sprays per day (interquartile range 3.0–7.0), and the mean was 5.4 sprays per day (standard deviation 4.9). Multiplying an actuation count by labeled milligrams estimates drug in administered product; it does not establish that all of the labeled amount entered systemic circulation.

The study did not report analytical measurements of THC, CBD, 11-hydroxy-THC, or THC-COOH in forensic venous whole blood, separated serum, plasma, urine, exhaled breath, or oral fluid. There was no specimen collection tied to the driving responses. No validated blood-THC concentration, within-person concentration trajectory, impairment cutoff, or pharmacokinetic parameter can be derived from the registry’s prescribed spray counts. Any claimed conversion from this cohort’s average sprays per day into a case-specific blood concentration would be an unsupported assumption.

For an actual specimen, the relevant questions include collection time, blood fraction, anticoagulant, analytical specificity, and reporting units. A primer on different biological matrices in toxicology provides the framework: a product label in milligrams and an assayed parent-drug concentration in ng/mL are not equivalent measurements.

The main evidence: what the registry recorded

The following table preserves the study’s changing denominators. A count of 941 does not mean that every covariate, exposure duration, or benefit judgment was available for all 941 people. The observation window extended through March 31, 2015, whereas the article was published in 2016; these results should not be presented as a present-day safety census.

Did prescribers report better or worse driving?

The primary article’s special-interest table divides driving responses into six mutually accounted-for patient categories. Improved: 63 (6.7%). Deteriorated: 19 (2.0%). Both improved and deteriorated at different collection periods: two (0.2%). No change: 303 (32.2%). Not appropriate: 514 (54.6%). Not recorded: 40 (4.3%). The categories sum to the full cohort of 941. Crucially, the two patients assigned “both” are not silently counted again in the separate improved and deteriorated rows.

Descriptions of improvement or decline were entered by prescribers, not produced by an instrumented on-road assessment, a validated hazard-perception test, or direct reaction-time measurement. The paper did not document when the patients drove, baseline driving skill, hours behind the wheel, traffic conditions, crashes, enforcement outcomes, medication interactions, or contemporaneous THC concentrations. A report of “no change” is likewise not a scientifically measured demonstration of no impairment.

It would be misleading to say 7% of Sativex users became better drivers. That number is 63 divided by the entire registry, including 514 for whom the driving question was not appropriate and 40 with missing answers. Conversely, 2% reported deterioration does not estimate the chance that a randomly selected person driving after an actuation would be impaired. The exposed, eligible, actually driving population and standardized outcome definitions were not established.

Why might a clinician perceive improvement without THC improving psychomotor ability?

The medicine was used largely for MS-associated spasticity. Symptom relief can alter comfort, mobility, sleep, fatigue, or practical ability to participate in daily activity. It is conceivable that a prescriber describing a patient as better able to drive was observing improved control of an underlying symptom rather than superior reaction time. That is a clinical hypothesis, not something this registry causally demonstrated. Conversely, patients who felt more functional might drive more frequently, increasing exposure to risk without any change in underlying psychomotor competence.

A proper driving study would distinguish disease severity and benefit, dose timing, concurrent medications, objective task metrics, actual driving exposure, and prespecified impairment criteria. The registry collected none of that information systematically enough to resolve competing explanations.

What was reported about falls, psychiatric symptoms, and adverse events?

Driving responses should not be considered in isolation from the other safety signals. In the same cohort, clinicians recorded 61 patients with falls requiring medical attention, 55 with other significant psychiatric or psychotic events, and 15 with suicidal thoughts or attempts. These are reported events in treated people; neither the occurrence nor the absence of a report establishes that the drug caused, prevented, or had no relation to the event.

The investigators also recorded 24 patients with a serious adverse event regarded by the prescriber as treatment related (2.6%). There were 32 patients with serious adverse events leading to death; the authors noted that one lung-cancer report included a remote possibility of a relationship, but significant tobacco exposure complicated that assessment. These details prohibit simplistic assurances that the product could not be associated with severe harm; they also prohibit claiming the treatment caused every such outcome.

Among treatment-attributed events, dizziness was documented for 22 patients (2.3%) and fatigue for 16 (1.7%). All-causality reports of the same preferred terms were higher: dizziness 29 (3.1%) and fatigue 20 (2.1%). Somnolence appeared among ten (1.1%) all-causality adverse event reports. Dizziness, fatigue, and somnolence can be relevant hypotheses in a driving inquiry, but the registry did not cross-tabulate them with validated moment-of-driving performance or measured drug concentration.

Does the exposure-adjusted event rate tell us the risk caused by the medicine?

No. The authors divided numbers of patients with an event by 2,213.98 cumulative patient-years. For medically attended falls, 61 divided by that exposure corresponds to about 0.028 reported patients with a fall per patient-year. This is a descriptive quotient, not a randomized risk ratio, a prospective recurrent-event hazard, or evidence that the treatment reduced falls compared with no treatment. The cohort lacked a matched untreated group and a common independently verified fall-surveillance method.

Time-at-risk also varied substantially. Duration was known for 848 patients, ranging from one to 4,289 days, with a mean of 954 days and standard deviation of 924. A cumulative exposure denominator can convey the amount of follow-up while still obscuring whether every event would have been ascertained. Events not brought to a clinician, deemed not clinically significant, or omitted from a retrospective form need not appear as incident reports.

In the source’s onset table, 77 patients were reported with a first adverse event within the first 28 days. The later duration categories have changing denominators because fewer patients were still exposed long enough to be at risk. There were also 95 adverse-event reports without a recorded time to first onset. Accordingly, apparent early clustering does not establish an individual patient’s future risk after a certain treatment interval.

How should treatment continuation and perceived benefit be interpreted?

Continuing therapy is not the same as being unimpaired. Status was known in 868 patient records; 563 were still receiving the medicine and 305 had stopped. Of those with perceived-benefit information, 680 of 822 were considered by clinicians to receive worthwhile benefit at one or more reporting periods. Different denominators are indispensable: dividing 680 by all 941 or calling it patient-rated benefit changes what the source measured.

Among people who stopped, reasons included lack of effectiveness, availability or cost, adverse events, death, and difficulty using the spray; more than one reason could be reported for a person. A patient who tolerates and continues a prescription may still experience transient performance effects at particular times after administration. A patient who discontinues may do so for reasons wholly unrelated to psychomotor impairment. These are not mutually exclusive classes of “safe” and “unsafe” drivers.

The apparent balance of efficacy and tolerability in a clinician-reported registry also depends on who appears in the system. Those receiving long-term prescriptions represent people who have already survived an initial course of treatment and have access to continuing care. This survivorship and treatment-selection process cannot supply the causal contrast that randomization would provide.

Did the registry establish that misuse, dependence, or abuse never occurred?

No. A special questionnaire was added later rather than administered uniformly to the original full cohort. Prescribers supplied the targeted responses for 427 of 941 patients (45%). Thirteen respondents had at least one higher concern score requiring follow-up. Some concerns were investigated and not substantiated as dependence or diversion, but incomplete screening and follow-up cannot rule out every occurrence in the other patients or outside the observation windows.

This is an example of the difference between no detected pattern in the available surveillance and proof that a risk is absent. For forensic questions, a clinician’s low concern rating is not a laboratory measurement of drug adherence, a toxicological test for abuse, or independent proof about another person’s behavior.

What can a forensic expert properly use this study to establish?

The study can support a historically accurate proposition: under routine clinical prescribing, voluntary physician reports from a largely MS population described certain experiences, events, continuation rates, and apparent benefits over a substantial span of recorded exposure. It establishes the existence and frequency of those reported categories within this registry, subject to the stated denominators and missingness. The report is relevant to pharmacovigilance and the limits of real-world evidence.

It does not validate a proposition that every therapeutic THC recipient is fit to drive, that THC improves performance, or that a given treatment contributed to or caused a collision. It does not supply the blood-THC concentration of a specific person, establish a dose-response impairment curve, or substitute for contemporaneous clinical evidence. A correctly framed opinion must separate population registry reporting from measurements or observations about the individual in dispute.

A competent case review therefore examines the actual timeline of medication use, symptoms, co-administered drugs, medical condition, objective conduct, examination findings, and specimen evidence. The distinction between screening and a chemically confirmed result is discussed in confirmatory drug testing; the registry included neither kind of direct analytical drug measurement for its driving category.

What records would matter in litigation over a particular driving event?

Begin with the prescription details, actual number and times of sprays, adherence history, contemporaneous diagnoses and functional baseline, symptom trajectory, sedating or otherwise relevant co-medications, the precise timing and factual circumstances of driving, witnesses’ descriptions, and any available telematics or video. Where a biological specimen exists, document the collection, timing, preservation, custody, test method, parent analyte and metabolites, concentration units, and interpretive uncertainty. The specimen collection and handling framework is important because laboratory data cannot be interpreted without knowing what was collected and how.

Ask whether anyone actually observed an error in tracking, judgment, divided attention, speed regulation, or vehicle control. Distinguish a symptom such as fatigue from proof that fatigue caused a maneuver. Do not silently equate a physical fall with a motor-vehicle crash or a prescriber’s “improved driving” category with a standardized driving-performance result. Alternative causes should be assessed even when THC exposure is established.

Case-specific toxicology can sometimes inform exposure timing or plausibility, but this particular registry offers no calibration for it. A contemporaneous breath-alcohol measurement, if present in an unrelated case, cannot be turned into THC whole-blood concentration; a plasma or serum THC result also cannot be treated as whole blood without relevant methodological evidence. The scientific task is to respect matrix, analyte, and endpoint boundaries rather than filling those gaps with a favorable registry percentage.

Why do the registry’s sponsorship and collection rules matter?

The paper states that the study was maintained by GW Pharmaceuticals and that its authors were employees of the company. The registry was part of a postmarketing risk-management program, and an independent advisory board reviewed the data submitted to regulators. Those design facts do not invalidate the numerical entries, but they are material when assessing possible reporting, collection, interpretation, and sponsorship biases.

The researchers acknowledged voluntary, retrospective submissions and broad prescriber judgment about what constituted a clinically significant adverse event. The collection system could capture the same patient across periods but did not require direct patient-completed questionnaires. Nor did it compel prescribers to report every eligible patient. These constraints should travel with the study whenever the dataset is used in an expert report or deposition.

Importantly, the cohort’s results are not a comparative risk estimate. Without a suitable comparator observed in a comparable fashion, more safety events in the registry do not prove harm from THC, and fewer observed events than in an unrelated population do not prove a protective effect. The original study’s discussion speculated about possible improvement in some clinical functions; that hypothesis must not be converted into an established ability-to-drive finding.

Frequently asked forensic questions

1. Did 7% of Sativex-treated patients objectively drive better?

No. Prescribers categorized 63 of 941 patients as improved in driving ability. No standardized driving performance was tested. More than half of the cohort was assigned “not appropriate” for the driving question, and a further 40 records lacked an answer.

2. Does a 2% deterioration report prove that only 2% could be impaired?

No. Nineteen records described deterioration, but the survey did not establish who drove, when they drove, or whether subtle psychomotor changes went unrecognized. The reported percentage is not the incidence of THC-related impairment per journey, per treatment, or per exposed motorist.

3. Does the registry report blood THC concentrations or a threshold?

No. It reports the amount of THC and CBD present in the prescribed spray and summarized use, not assayed drug concentrations in venous whole blood, serum, plasma, or another biological matrix. Neither a legal threshold nor an individual pharmacokinetic estimate follows.

4. Can the 61 medically attended falls be attributed to nabiximols?

Not from these data alone. Falls were documented in treated patients, but disease severity, mobility, other medications, and how reporting occurred all matter. No comparable untreated control group or case-level causal adjudication establishes what caused each fall.

5. Does no newly detected safety signal mean no serious event occurred?

No. The investigators documented 113 patients with at least one serious adverse event of any attributed cause and 32 patients with a serious adverse event leading to death. “No new signal” is a surveillance interpretation, not a statement that no severe adverse outcome occurred or that risk is zero.

6. Were the 941 patients surveyed directly about their driving?

No. Clinicians provided anonymized retrospective case-report forms. There were no patient questionnaires and no blinded driving examinations. A physician’s narrative observation is not an independently measured driving score.

7. What is the scientifically defensible courtroom conclusion?

The registry provides descriptive real-world safety observations for patients prescribed combined THC–CBD medicine, with important gaps and selection biases. It cannot, by itself, establish that a particular person was unimpaired, impaired, truthful, safe to drive, or responsible for a collision. Any such opinion requires independent case-specific evidence and a method suited to the actual question.

Primary study and interpretive boundary

Etges T, Karolia K, Grint T, Taylor A, Lauder H, Daka B, Wright S. An observational postmarketing safety registry of patients in the UK, Germany, and Switzerland who have been prescribed Sativex (THC:CBD, nabiximols) oromucosal spray. Therapeutics and Clinical Risk Management. 2016;12:1667–1675. Original primary research DOI: 10.2147/TCRM.S115014. This analysis uses the full primary methods, numerical registry tables, results, and limitations, not an abstract or secondary review. The clinically recorded changes in driving ability are observations with missing and inapplicable categories—not a validated chemical or behavioral test of personal impairment.