Can Lisdexamfetamine Reduce Cocaine Craving Without Lowering Positive Urine Tests? A 43-Person Trial

Answer first: In a randomized, double-blind, 43-person pilot study, daily oral lisdexamfetamine (LDX) reduced a self-reported measure of needing cocaine, but it did not produce a statistically established decrease in positive urine cocaine-metabolite tests in the full randomized sample. With missing urine results counted as positive, adjusted benzoylecgonine-positive rates were 92.4% for placebo and 86.5% for LDX (p = .4728). By contrast, average ratings of “needing cocaine” were 28.7 and 17.5 on a 100-mm visual-analogue scale, respectively (p = .0176). A positive craving result does not convert the nonsignificant biochemical comparison into a positive cocaine-use trial.

The evidence comes from Mooney, Herin, Specker, Babb, Levin, and Grabowski, Pilot Study of the Effects of Lisdexamfetamine on Cocaine Use: A Randomized, Double-Blind, Placebo-Controlled Trial (2015), Drug and Alcohol Dependence 153:94–103, original primary study DOI. The results are relevant to interpreting clinical drug-testing evidence, but the experiment did not determine any participant’s cocaine concentration in forensic venous whole blood, individual time of use, fitness to drive, or cause of an incident.

What did the researchers establish—and what did they not?

They established that two kinds of measurement could point in different directions. Participants assigned LDX reported feeling less need for cocaine on a specified questionnaire item, whereas the preplanned comparison of urinary cocaine-metabolite positivity in all randomized patients did not show a significant medication effect. Both findings may be accurate descriptions of the study because they assess different constructs. Craving is an experienced psychological state. Urinary benzoylecgonine positivity records whether an assay crossed a decision threshold at a collection visit. Neither is a direct measurement of the other.

Nor is a failure to demonstrate a difference proof that the medications were equivalent. This pilot randomized 43 people rather than the substantially larger number the investigators estimated they would need. Frequent missed visits, uneven adherence, and substantial attrition affect what can be inferred from any apparent numerical difference. The study supports a limited conclusion: a subjective craving signal worth further research, alongside a negative full-sample test for its principal observable cocaine-use outcome. It does not validate LDX as a treatment that reliably produces abstinence for an individual.

Who participated, and how many people were independently studied?

The trial enrolled 43 treatment-seeking adults meeting the then-current DSM-IV criteria for cocaine dependence. Randomization allocated 22 people to LDX and 21 to placebo; these are independent human participants, not specimens or observations. A smaller selected set of 27 completed the 14-week treatment phase: 12 originally allocated LDX and 15 allocated placebo. Participants attended repeated clinical visits and supplied repeated urine samples, but those multiple measurements do not increase the independent biological sample size above 43.

Eligibility excluded several conditions that could materially change treatment safety or generalizability, including cardiovascular disease, a relevant history of seizures, and various other psychiatric diagnoses. Patients were seeking treatment, so the cohort was not a random survey of community cocaine consumers and cannot be generalized automatically to emergency patients, arrestees, drivers, people with competing diagnoses, or people exposed to different stimulant doses. The groups were described as similar on recorded baseline variables, which helps the randomized comparison but does not eliminate problems from missing follow-up data.

Was the study drug cocaine, amphetamine, or methamphetamine?

The experimentally assigned medication was lisdexamfetamine, an orally administered prodrug that is converted in the body to the active stimulant dextroamphetamine. It was not methamphetamine and was not a controlled administration of cocaine. The people in the trial already had cocaine dependence; investigators followed cocaine-use outcomes during treatment. Any cocaine that participants consumed outside the research clinic was not a uniform investigator-administered dose with a known administration time.

The original paper described a maximum daily LDX treatment dose of 70 mg, roughly corresponding to a 30-mg dextroamphetamine component for the authors’ clinical comparison. That approximate equivalence is a statement about the prescribed preparation, not a measured dextroamphetamine concentration in a participant’s whole blood. A dose in milligrams cannot be translated into an individual forensic drug result in nanograms per milliliter without the appropriate specimen, chemical method, collection time, and pharmacokinetic information.

How were treatment and follow-up organized?

Researchers conducted a one-week intake phase, a week of medication induction, fourteen weeks of active blinded treatment, and a final medication-reduction phase. Participants in both arms received weekly manualized cognitive-behavioral therapy, so the drug comparison was LDX added to a shared clinical treatment program versus placebo added to the same program. It was not stimulant therapy compared against no clinical support.

Follow-up involved approximately twice-weekly urine collection opportunities and weekly subjective measures. Attrition and nonattendance can affect these repeated observations in two separate ways: they reduce what is directly observed and create analytical choices about how to represent missing results. A participant who missed several visits does not thereby contribute repeated evidence of biochemical cocaine use unless the investigators choose to impute those missing tests as positive. The paper therefore reported multiple analysis strategies, and their denominators and assumptions should never be interchanged.

What exactly was measured in the urine?

At clinical visits, collected urine was screened semi-quantitatively for benzoylecgonine (BE), a metabolite associated with cocaine exposure. The investigators identified the PROFILE-V MEDTOXScan drugs-of-abuse system and treated BE at or above 150 ng/mL in that urine testing framework as positive. This is a study-defined urine-screen threshold. It is not a forensic impairment limit, a quantitative venous whole-blood cocaine concentration, or a clock that independently reveals the exact time of consumption.

The parent compound cocaine and its metabolite BE are chemically distinguishable. The laboratory test centered on the latter, not a mass-spectrometric determination of parent cocaine in a simultaneous venous blood sample. The study did not provide a matched series of whole-blood concentrations for parent cocaine, cocaethylene, dextroamphetamine, or lisdexamfetamine. Neither a reported 150-ng/mL urine threshold nor an observed percentage of urine-positive visits can be copied into an unrelated specimen matrix or treated as a validated legal threshold.

For a broader explanation of why sample type matters, see the published guide on biological matrices in toxicology. When identification specificity matters in an actual case, the separate subject of confirmatory drug testing becomes essential; this trial’s semi-quantitative clinical screen should not be described as independent confirmatory GC–MS or LC–MS/MS testing.

Why do missed urine samples change the answer?

The authors analyzed the randomized sample under two missing-data rules. Their intention-to-treat-style strategy designated an uncollected urine test as cocaine-positive. This conservative assumption for abstinence research increases the counted positive proportion when people miss visits. In the companion strategy, missing test results remained missing and did not automatically become observations of cocaine exposure. A third distinction separated everyone originally randomized from only those who completed the full treatment course.

These analytical choices matter because an uncollected specimen is not a measured positive specimen. An imputed positive can be reasonable for one prespecified treatment-outcome analysis without becoming a documented laboratory result about the person. Conversely, simply discarding missing specimens can distort comparisons when missingness is related to treatment, drug use, symptoms, adherence, or dropout. The crucial randomized and completer results are presented separately below rather than pooled into a single efficacy claim.

Evidence table 1: randomized versus selected-completer urine findings

The percentages in this table are adjusted rates of positive urine test observations under different modeling and missingness rules, not proportions of unique participants who ever used cocaine. One person could supply many tests. The subset who finished treatment was determined after randomization: analysis restricted to those people can change the balance of prognostic factors even when baseline summaries do not identify a significant difference. The authors themselves cautioned against treating the completer-only finding as causal evidence. Its significance within a selected follow-up subset does not supersede the full randomized comparison.

Did the full randomized sample demonstrate lower cocaine use?

No statistically established reduction was detected. Under the report’s missing-as-positive rule, the adjusted 5.9-percentage-point separation was accompanied by p = .4728. Under the alternative missing-left-missing rule, the numerical separation was larger, but the authors still reported a nonsignificant comparison (p = .1106). Both randomized-sample results fail the study’s conventional .05 test criterion. The second analysis should not be singled out solely because its percentage contrast looks more favorable.

It would be inaccurate to say that the trial proved LDX had no possible effect on cocaine use. A small experiment may fail to reject a null hypothesis even when a clinically relevant effect exists, and the missing-data structure amplifies uncertainty. It would be equally inaccurate to say the study proved a beneficial medication effect on cocaine use based solely on the selected completers. The credible public account reports both possibilities as uncertainty: no full-sample demonstration, coupled with an exploratory signal requiring replication.

Why was the completer analysis not a second randomized trial?

The 27 people who finished the treatment phase were not newly randomized into their final subgroup at week fourteen. They were the survivors of the original treatment and follow-up process. Under the missing-left-missing approach, their modeled positive-rate difference was large, and the published report described the contrast as statistically significant. Under the alternate missing-as-positive rule, however, even the completer-only comparison did not reach the usual significance criterion (p = .0622). This sensitivity to analytic decisions is highly relevant when the headline claim concerns objective drug-use reduction.

Completers can differ from noncompleters on unrecorded factors. They may be more adherent, more motivated, differently symptomatic, or more likely to provide valid urine samples. Neither treatment completion nor urine collection is an independent randomized exposure. Accordingly, the exploratory analysis is useful for choosing future hypotheses, not for claiming that any particular patient’s cocaine use would necessarily fall after taking a prescribed prodrug.

Did patients actually report less cocaine craving?

Yes, for a specified subjective item. The original report’s between-treatment comparison of “needing cocaine” was 28.7 ± 3.21 standard-error units on the 100-mm rating scale for placebo and 17.5 ± 3.30 for LDX. The group comparison was F(1, 62.6) = 5.94, p = .0176. These values reflect a statistically supported average difference in a patient-reported craving measure over the study period.

A visual-analogue score asks a patient to express intensity on a rating scale; it does not quantify cocaine or BE in urine or plasma. A lower feeling of need could occur alongside continued use, sporadic use, or a different pattern of uncollected specimens. Reporting the craving result does not require suppressing it, but interpretation must keep it separate from objective positive-urine testing. Blinding quality also deserves attention: many patients correctly guessed their medication assignment, potentially affecting subjective reporting and expectations.

What happened to retention and medication adherence?

Retention showed a numerical difference that was not statistically established. Of 21 placebo participants, 15 completed treatment (71.4%). Of 22 LDX participants, 12 completed. The survival comparison yielded log-rank χ²(1) = .9318, p = .3344. Sixteen people did not finish, mostly lost to follow-up, so statements about measured outcomes in the full original group need a clear account of those missing observations.

To assess taking the dispensed capsules, study medication included additional riboflavin, and urine fluorescence was measured. Values at or below 20 fluorescence units were treated as nonadherence, but the authors noted that dietary or supplemental riboflavin could generate misleading readings. Modeled adherence was 93.4% with placebo and 80.3% with LDX; the treatment comparison was p = .0550, which does not cross .05. A biomarker included in the medication capsule helps check adherence to that capsule; it does not confirm whether cocaine was consumed, prove an individual LDX concentration in legal blood, or measure impairment.

Evidence table 2: clinical and safety outcomes are different endpoints

The first row reports mean ratings with standard errors, not standard deviations. The adverse-event percentages describe at least one self-reported occurrence, not severity, number of episodes, or a causal diagnosis. The reported safety pattern warrants scrutiny; it cannot establish an individual’s susceptibility, nor can the absence of a significant between-group difference for a vital sign be converted into proof that drug-related clinical risk does not exist.

What do the serious adverse events establish?

Two serious events were described in the clinical report. One person assigned placebo was hospitalized with psychosis following reported cocaine binge use and a previously undisclosed schizoaffective diagnosis. One LDX-assigned participant was hospitalized for chest pain and diagnosed with angina that the researchers considered possibly related to study medication. Both discontinued treatment. These accounts are clinical safety observations with important factual context, not randomized event-rate estimates precise enough to determine cause in a specific unrelated individual.

Additional findings did not show significant study-medication differences for the reported group comparisons of body weight, systolic blood pressure, or heart rate. That statement is an account of the authors’ tests in this selected, relatively healthy group; it does not exclude rare cardiovascular events or validate the combination of prescribed stimulant treatment and unsupervised cocaine exposure as safe in other populations. The study was not designed to identify fatal drug concentrations, establish a legal toxicology threshold, or determine the mechanism of a hypothetical later chest-pain episode.

Why does the pilot’s small sample change the strength of its conclusion?

The authors aimed for a larger efficacy study than they ultimately completed. They randomized 43 people and estimated that, given the observed effects, approximately 102 people per group could be required for adequate statistical power. The study therefore had limited ability to separate modest treatment benefits from chance variation, especially after some participants dropped out and repeated urine specimens were missing. A nonsignificant result in such a pilot need not demonstrate equivalence; likewise, an unexpected positive subgroup analysis can be unstable.

Several additional uncertainties are important: all participants received behavioral therapy, the groups could not be completely insulated from expectations about the active drug, investigators did not obtain an independent series of confirmatory blood cocaine tests, and the semi-quantitative urine assay captured visits rather than verified ingestion events. Blinded treatment assignment helps with confounding at baseline, but postrandomization selection and imperfect blinding still affect the weight of later comparisons.

How is a cocaine-treatment urine screen different from a forensic drug test?

In this trial, urinary BE positivity was used to count study follow-up outcomes. A forensic question may be very different: what analyte was detected in which specimen, when was it collected, was the result confirmed by a chemically specific method, and does it support a particular opinion about recent ingestion, an individual’s clinical condition, or an event? The trial did not produce the information needed to answer these latter questions. The participants’ observed urine tests were not direct measurements of legal whole-blood cocaine or dextroamphetamine concentration.

The measured BE in urine is also different from cocaine parent compound in plasma and from cocaethylene, which can form in the presence of ethanol. There was no paired quantitative blood series to derive an individual urine-to-whole-blood conversion, and no on-road driving assessment. A urine BE positive at the study’s decision threshold is a positive in that assay and matrix under those conditions; it is not proof of intoxication at collection, recent driving impairment, or an incident’s medical or legal cause.

Can 150 ng/mL be treated as a cocaine impairment threshold?

No. The 150-ng/mL value described how the researchers classified BE in urine for an outcome analysis. It did not define impaired driving, a fatal concentration, or a statutory concentration in human venous whole blood. Units alone do not make chemically and physiologically different specimens interchangeable. In particular, measuring a metabolite in urine, which reflects excretion and collection conditions, is not equivalent to determining parent cocaine in circulating blood at a known time.

If an actual legal or clinical interpretation turns on drug identification and instrument specificity, the reported screening method must be evaluated on its own terms. The trial does not establish confirmatory test performance, sensitivity to actual cocaine consumption under a verified dose/time protocol, or a validated cutoff for individual culpability. Thresholds chosen for treatment trials are decision rules for those trials, not shortcuts to reconstructing an event.

Could a lower craving score predict safer driving?

The original experiment did not test driving. It did not measure lane position, braking latency, attention while driving, on-road safety, collisions, or independent field impairment assessments. A difference in one craving scale therefore cannot establish restored psychomotor capacity or exclude stimulant-related behavioral effects. Nor does a person with a lower craving score necessarily have a negative urine drug result or a sufficiently low blood concentration for any particular activity.

Even when different biological and behavioral measurements are available, forensic opinions require careful alignment of specimen, assay, collection time, timing of conduct, clinical signs, alternative causes, and uncertainty. This study supplies useful evidence about a treatment hypothesis in a small cohort; it supplies no individually validated bridge from self-rated craving to case-specific intoxication, reaction time, last cocaine use, or accident causation.

What can forensic professionals responsibly take from the study?

The central interpretive lesson is to identify the endpoint before making the claim. The randomization tested a treatment assignment, the urine screen classified a cocaine metabolite under a particular rule, the craving scale recorded a psychological state, and the riboflavin reading offered an imperfect marker of taking the prescribed capsules. These are different measurements with different sources of error. When a question uses the broad phrase “drug use,” it is critical to identify which of those quantities was actually observed.

A scientifically defensible reading recognizes the full randomized urine comparisons as nonsignificant, acknowledges that craving ratings differed, and labels the completer-only finding as exploratory and vulnerable to selection. The paper supports further investigation of LDX for cocaine-use disorder; it does not prove an individual consumed cocaine, specify blood concentrations, validate a legal threshold, or determine liability. For the separate role of dose history and lab matrix in therapeutic settings, see the published guide to therapeutic drug monitoring in toxicology. The trial itself did not provide quantitative LDX therapeutic drug monitoring.

Seven frequently asked forensic-toxicology questions

1. Did lisdexamfetamine significantly reduce cocaine-positive urine tests in all 43 randomized patients?

No. The authors reported p = .4728 when missing tests were treated as positive and p = .1106 when missing tests remained missing. Neither meets the .05 criterion. The numerical direction of the adjusted percentages should not be described as a confirmed medication effect on cocaine use in the full randomized sample.

2. Why was craving statistically lower if urine positivity did not clearly improve?

Craving is self-reported experience; urine benzoylecgonine testing is a biological measurement of a metabolite under a cutoff. The average craving response can change without proving a parallel biochemical effect. Treatment expectations, missing visits, measurement variability, and the distinct endpoints also matter.

3. Was the positive result among 27 treatment completers evidence of an effective cocaine therapy?

It was an exploratory, restricted analysis, not the full randomized comparison. Completion occurred after randomization, and the signal depended in part on the missing-data rule. It is appropriate for designing a larger replication study, not proof of causal efficacy for an individual patient.

4. Did the researchers directly measure cocaine, lisdexamfetamine, or dextroamphetamine in venous whole blood?

No quantitative forensic venous whole-blood drug series was reported for those analytes. The principal biochemical cocaine-use outcome was semi-quantitative urine benzoylecgonine screening. Medication adherence used urine riboflavin fluorescence, a separate marker unrelated to a blood concentration.

5. Is the study’s 150-ng/mL urine benzoylecgonine rule a statutory driving-impairment limit?

No. It classified urinary BE screens for the trial’s repeated follow-up analyses. It was neither a legal venous whole-blood cutoff nor a validated individual driving-performance threshold. The study did not include driving tests to create such a relationship.

6. Does a failure to find significant differences in retention or vital signs mean LDX was harmless?

No. The study was small, participants were selected for medical eligibility, and one LDX-assigned patient experienced chest pain diagnosed as angina that investigators considered possibly drug-related. A statistical failure to detect a group effect is not an individual clinical safety guarantee.

7. Can this paper identify precisely when a particular person last used cocaine or why an incident occurred?

No. Repeated urine BE screens, self-rated craving, and nominal study-drug doses lack the matched chemical and temporal evidence required to reconstruct an individual’s last dose, legal intoxication, impaired driving, or a particular collision or injury. Such opinions require independent case-specific data, not this pilot trial alone.

Primary scientific source and interpretation boundary

Mooney ME, Herin DV, Specker S, Babb D, Levin FR, Grabowski J. Pilot Study of the Effects of Lisdexamfetamine on Cocaine Use: A Randomized, Double-Blind, Placebo-Controlled Trial. Drug and Alcohol Dependence. 2015;153:94–103. Read the complete original study by DOI. This commentary preserves the distinction between the trial’s observed positive subjective outcome, its nonsignificant randomized urinary outcome, exploratory subgroup analysis, safety observations, and outcomes it never measured. No one study establishes individual forensic conclusions outside its design.