Does Switching From Amphetamine to Lisdexamfetamine Improve Adult ADHD? Why Two Placebo Patients Limit the Evidence

Answer first: A published post hoc analysis found substantial improvements in ADHD symptom ratings among 39 adults given lisdexamfetamine (LDX) who had been taking another amphetamine formulation at screening. That observation is clinically interesting, but it does not establish that switching to LDX is superior to continuing the previous treatment or to placebo in this subgroup. Only two previously amphetamine-treated participants were assigned placebo, the investigators used descriptive rather than comparative efficacy statistics for the subgroup, and the original trial did not randomize patients between different amphetamine formulations. The study also measured clinical symptoms and routine safety variables—not amphetamine concentrations in forensic whole blood or individual driving impairment.

The original empirical analysis is Babcock, Dirks, Adeyi, and Scheckner (2012), Efficacy of lisdexamfetamine dimesylate in adults with attention-deficit/hyperactivity disorder previously treated with amphetamines: analyses from a randomized, double-blind, multicenter, placebo-controlled titration study, BMC Pharmacology and Toxicology 13:18, doi:10.1186/2050-6511-13-18. It presents a secondary analysis of patient-level observations collected in a four-week randomized clinical trial, rather than a newly designed switching experiment.

Why this question matters in clinical and forensic interpretation

A physician considering ADHD treatment may ask whether a patient who remains symptomatic on mixed amphetamine salts or dextroamphetamine could benefit from a different formulation. A forensic evaluator may encounter the same history in medication records, deposition testimony, a drug-screen narrative, or the chronology of a collision. Both questions require identifying exactly what the study measured and resisting an inference that its design cannot support.

LDX is an orally administered prodrug that is converted to pharmacologically active dextroamphetamine. It should not be described as methamphetamine, and its nominal capsule dose should not be treated as a measured blood concentration. A change in an ADHD rating after therapy is not a chemical exposure assay. Equally, a post hoc subgroup containing two placebo patients cannot support a reliable, precise group treatment-effect estimate merely because the parent trial was randomized.

What did the researchers actually do?

The investigators revisited data from a multicenter, double-blind, four-week trial that assigned adults with ADHD to placebo or oral LDX at 30, 50, or 70 mg per day under a forced-dose titration protocol. Participants were 18 to 55 years old. The parent trial’s efficacy population comprised 414 independently participating adults: 62 assigned placebo and 352 assigned LDX. The separate safety population comprised 420 randomized, treated adults: 62 placebo and 358 LDX. Those different denominators reflect their distinct analysis eligibility requirements, not different studies.

For the present exploratory analysis, the authors identified 41 of the 414 efficacy participants who were taking another amphetamine-containing medicine at screening. The historical regimens included mixed amphetamine salts, an extended-release mixed-salts product, and dextroamphetamine preparations. Two of these 41 were later randomized to placebo and 39 to LDX. Within the LDX subgroup, 11 received a 30-mg target dose, 16 a 50-mg target dose, and 12 a 70-mg target dose.

People had been taking a previous amphetamine for periods ranging approximately from two weeks to thirteen years. Because the earlier medicine was washed out before the randomized trial’s baseline rating, the experiment did not continuously compare a patient’s old formulation with the new one at matched treatment times. There was no concurrent randomized arm continuing the previous amphetamine. The history of previous medication is a grouping variable, not an independently randomized comparative intervention.

How many independent people were analyzed?

The decisive number for the previously amphetamine-treated subgroup is 41 adults, not 414, not the sum of repeated ADHD questionnaires, and not four weeks multiplied by 41. Among those 41, 39 received LDX and only two received placebo. The investigators further identified 36 people with screening ADHD Rating Scale IV (ADHD-RS-IV) total scores above 18 despite their previous amphetamine use: 34 subsequently received LDX and two placebo.

The symptomatic subset of 36 and the full prior-treatment subgroup of 41 overlap. They must never be added together as though they represented 77 independent patients. Likewise, parent-trial safety denominators of 420 and efficacy denominators of 414 must be respected whenever their respective rates or symptom outcomes are discussed.

Which clinical endpoints were measured?

The principal symptom instrument was the 18-item ADHD-RS-IV using adult prompts, measured at screening, after the previous therapy washout at baseline, and during randomized treatment. Higher scores reflect more severe reported ADHD symptoms. The trial also used the clinician-rated Clinical Global Impressions severity (CGI-S) and improvement (CGI-I) scales. For its exploratory analyses, the investigators defined clinical response as both an ADHD-RS-IV reduction of at least 30% from baseline and a CGI-I rating of 1 or 2. They defined symptomatic remission operationally as a post-baseline ADHD-RS-IV total score of 18 or below.

These thresholds were study-specific methods for summarizing behavioral or clinical scores. They are not toxicology concentration thresholds, intoxication cutoffs, validated forensic driving-performance limits, or proof that a patient was symptom-free under all real-world conditions. The investigators also collected treatment-emergent adverse events, blood pressure, pulse, routine laboratory variables, and electrocardiographic findings.

Evidence table 1: where the 41-person subgroup fits into the parent trial

The symptom changes show why the investigators considered LDX a plausible treatment option in symptomatic people with previous amphetamine exposure. They do not isolate the effect of switching formulations, because prior treatment was neither maintained as a randomized comparator nor standardized for dose, adherence, tolerability, or duration. A table of mean changes alone cannot supply the missing causal contrast.

What symptom improvement was observed with LDX?

For the 34 symptomatic patients previously receiving an amphetamine and then treated with LDX, the mean decrease from the randomized post-washout baseline ADHD-RS-IV score was 17.8 points by endpoint. Among the two similarly symptomatic prior-treatment participants assigned placebo, the mean decrease was 13.5 points. The investigators explicitly acknowledged that meaningful placebo–LDX comparison within this tiny subgroup was not warranted. The numerically greater LDX reduction must therefore remain a description, not a demonstrated statistically significant switching benefit.

In the entire efficacy population, irrespective of prior amphetamine use, the mean baseline-to-endpoint ADHD-RS-IV score decreased by 17.5 points with LDX and 7.8 points with placebo. That parent-trial contrast is better populated, but it answers a broader efficacy question about LDX in trial participants rather than the narrower causal question of what happens when someone switches from one amphetamine formulation to another. Borrowing statistical confidence from 352 active-treated people does not manufacture a reliable effect estimate for the 39-person exploratory subgroup.

Why do the two placebo patients matter so much?

A treatment-effect estimate depends on the variability and comparability of both treatment groups. Two people cannot adequately characterize the natural variation of a prior-amphetamine-treated placebo population, particularly when one participant’s outcome can change a group percentage by fifty percentage points. The placebo participants also happened to be both female, while the LDX subgroup included 20 men and 19 women. That imbalance illustrates how small the comparator is and why adjustment for potentially relevant participant differences would be extremely unstable.

Chance baseline differences, expectation effects, cessation and washout of the previous therapy, regression toward the mean, and clinical fluctuation can all influence repeated symptom scores. The investigators could not reliably distinguish these sources of variation from a new-formulation treatment effect within the two-versus-39 subgroup. Calling the parent study “randomized” is accurate; describing this post hoc subset as a well-powered randomized comparison is not.

Was this a prospective head-to-head switch trial?

No. Every participant in the parent trial received a randomized LDX dose or placebo after the requisite baseline procedures. The researchers did not randomly assign symptomatic amphetamine users either to continue their established amphetamine formulation or to switch to LDX while otherwise preserving the same care, monitoring, and follow-up. Nor did they prospectively stratify a sufficient number of previous amphetamine users to provide balanced randomized groups for a planned switch-effect analysis.

Consequently, the design cannot answer whether LDX outperformed a patient’s previous mixed-salts or dextroamphetamine product at an equipotent dose, whether a longer trial of the previous medicine would have worked, whether a different dosing schedule would have mattered, or which individual would respond best. An observed improvement following a sequence of treatments is not the same proposition as proof that the second treatment caused a greater improvement than the first.

Did the investigators test significance for the 41-person switch subgroup?

For this subgroup, the paper says the efficacy summaries were descriptive. It expressly states that no statistical comparison was conducted between the prior-amphetamine subgroup and the overall study population because that comparison was not part of the prospective study design. The tiny subgroup placebo count further made placebo-versus-LDX comparisons inappropriate. A calculated P-value or confidence interval presented as an original subgroup switching result would therefore be invented.

This is an important distinction between a result that was nonsignificant and a result that was never adequately tested. The investigators supplied numerical improvement but not a substantiated positive or negative inferential test of LDX superiority specifically among previous amphetamine users. An absence of such a test is not evidence that the therapies are equivalent, and it is not evidence of superiority.

What do the response and remission definitions mean?

Response combined a reduction of at least 30% in the ADHD-RS-IV score with a CGI-I rating of “very much improved” or “much improved.” Symptomatic remission was a total ADHD-RS-IV score of 18 or lower after baseline. Both could be summarized over weekly follow-up assessments, but the repeated visits remained observations of the same individuals rather than new biological participants. They are clinical constructs based on ratings, not measured drug-exposure biomarkers.

The authors reported broadly comparable patterns of clinical responders and remitters in LDX-treated prior-amphetamine users and the much larger LDX-treated overall efficacy population. Without a properly populated prior-treatment placebo control and a planned contrast, similarity of plotted responder percentages does not establish noninferiority, therapeutic interchangeability, an expected response rate for a new patient, or the causal superiority of any switch.

What did the separate safety analysis show?

The safety set of 420 people included 358 LDX-treated and 62 placebo-treated participants, while the previous-amphetamine subgroup had 39 and two respectively. Treatment-emergent adverse events are events reported after randomized treatment, not automatically events caused by the study drug. In the prior subgroup, 22 of 39 LDX-treated participants reported at least one such event; both of the two placebo participants also reported an event. A two-person placebo rate of 100% must not be interpreted as a stable population safety estimate.

Within the prior-amphetamine LDX group, dry mouth and headache each affected five people; fatigue and insomnia each affected three. The investigators reported no serious treatment-emergent events among those prior users. For the larger overall safety population, event reporting was more frequent in the LDX arm, and two serious events occurred in LDX-treated patients; investigators considered those two not related to treatment. A four-week observation window and a small exploratory prior-user subgroup cannot rule out uncommon, delayed, or patient-specific harms.

Evidence table 2: selected safety findings and genuinely tested results

Safety comparisons should preserve their source context. The reported pulse difference, P = .0018, relates to the overall safety population, not proof of an effect within the exploratory 41-person prior-amphetamine subgroup. Conversely, the paper describes endpoint systolic and diastolic pressure differences versus placebo as not statistically significant and reports no significant meaningful between-group QTcF change. Those are genuine source-reported nonsignificant observations; they are not evidence that no individual can experience cardiovascular effects.

Does the formulation’s prodrug chemistry prove a clinical advantage?

LDX consists of dextroamphetamine covalently linked to l-lysine and is converted to the active drug after administration. That mechanistic distinction is relevant when comparing products, but the Babcock post hoc study did not directly quantify circulating LDX, dextroamphetamine, metabolites, distribution kinetics, or a patient’s concentration–response relationship. Its discussion refers to other pharmacokinetic and preclinical studies, which are not independent blood measurements performed on these 41 participants.

It would therefore be incorrect to claim that the observed reduction in ADHD-RS-IV scores proves a specific red-blood-cell conversion rate, a distinctive forensic whole-blood concentration profile, resistance to misuse, or a measured distribution of drug levels in responders versus nonresponders. Hypotheses about different formulations may motivate research, but this secondary clinical analysis cannot supply missing chemical observations.

Which biological matrices and laboratory methods were actually used?

The researchers collected routine clinical safety information and laboratory findings, but the paper did not provide a forensic quantitative assay of parent LDX or dextroamphetamine in venous whole blood, separated plasma, separated serum, urine, oral fluid, hair, sweat, vitreous humor, or exhaled breath. It reported no specimen-specific chromatographic calibration, deuterated internal standard, limit of detection, limit of quantification, ng/mL individual drug concentrations, or matched pharmacokinetic time series for the post hoc patients.

That absence must not be obscured by the fact that the compound is known to generate active dextroamphetamine in blood. A mechanistic statement about where conversion occurs is not the same as collection and chemical testing of a specific specimen. For a case requiring interpretation of an actual result, biological matrix, sampling time, laboratory identification, and uncertainty are central; see the site’s published primer on distinguishing biological matrices in toxicology.

Can a prescription or ADHD rating identify a legal whole-blood level?

No. The administered 30-, 50-, or 70-mg LDX product dose is an oral treatment allocation, not a venous whole-blood dextroamphetamine result. Milligrams of an administered prodrug cannot be equated to ng/mL of its metabolite in a legally collected specimen without appropriate sampling, validated chemistry, pharmacokinetic context, and individual information. The same medication history may be consistent with very different specimen concentrations depending on dose timing, adherence, metabolism, and other factors.

Likewise, ADHD-RS-IV and CGI scores are clinical response instruments. They are neither indicators of drug concentration nor validated substitutes for reaction time, lane tracking, divided attention under actual driving conditions, or an individual assessment of impairment at a historical event. A laboratory report, if one exists in another case, would require its own confirmation and uncertainty assessment; see principles of confirmatory drug testing.

What are the strongest sources of bias and uncertainty?

The most serious limitation is the imbalance of two placebo patients and 39 LDX patients in the prior-amphetamine subgroup. The analysis was post hoc, not powered around switching outcomes, and not designed to test whether the patient should remain on the previous medication. Participants were selected for having previous treatment, but prior dose, duration, exact formulation, clinical management, and reasons for suboptimal symptom control varied. Washout before randomized baseline creates another interval in which symptom scores can change independently of the next prescribed product.

Additional limitations include a short four-week randomized treatment period, dependence on rating scales, trial selection of adults meeting diagnostic and rating criteria, an uneven sex distribution between exploratory arms, and a small number of adverse-event observations in the prior-treatment comparator. The authors were affiliated with the manufacturer sponsoring the original research, making funding and potential conflicts relevant to how conclusions are appraised. Those considerations do not negate the recorded data, but they do narrow what the data can establish.

What additional experiment would settle the switching question?

A stronger design would prospectively enroll a sufficiently large and clearly defined group of adults who continue to meet symptom criteria despite an adequately characterized amphetamine regimen. Random assignment would compare a maintained or optimized prior formulation with a protocolized switch to LDX, ideally with masked outcome assessment, balanced groups, specified baseline exposure/adherence procedures, and adequate follow-up. The clinically relevant difference should be pre-specified with confidence intervals, not inferred after examining an unbalanced subgroup.

Researchers would also need to distinguish symptom efficacy, actual everyday functioning, tolerability, and harms. If the study intended to make any pharmacokinetic assertion, it would need separately validated specimen collection and quantitative drug measurements in an explicitly stated matrix. Even a carefully designed clinical switch trial would not automatically yield legal driving-impairment thresholds without a dedicated and validated performance component.

How should this paper be cited or characterized in expert work?

A defensible characterization is that the investigators found descriptive symptom improvement during LDX treatment among adults who reported prior amphetamine therapy, but the two-person subgroup placebo arm and secondary design prevented a reliable comparative conclusion about switching. The overall randomized trial contains a wider clinical efficacy context, and its safety results include a significant overall pulse comparison alongside reported nonsignificant overall blood-pressure and QTcF findings.

It would overstate the evidence to assert that the experiment demonstrated that LDX is a superior replacement for all previous amphetamine preparations, that the post hoc treatment difference is statistically proven, or that a prescription signifies a measured or legally relevant blood concentration. The first-party empirical publication can be located at the original DOI record. Each application to a particular patient or legal matter would require separate facts and methods beyond those provided here.

Frequently asked questions

1. Did this trial show that switching to lisdexamfetamine beats staying on Adderall or another amphetamine?

No. The former treatment was not continued as a randomized control regimen. The exploratory prior-amphetamine subgroup contained only two placebo patients and 39 LDX patients, so it could not provide a reliable placebo comparison, much less a prospective head-to-head contrast against the prior mixed-salts or dextroamphetamine therapy. Symptom improvement after LDX is an observed sequence, not proof of switch superiority.

2. Were the 414 trial participants all previous amphetamine users?

No. The overall efficacy population of 414 consisted of 352 active LDX recipients and 62 placebo recipients. Just 41 had amphetamine therapy at screening, and 36 of those met the report’s screening symptomatic cutoff. These groups are nested, not independently additive; the safety population of 420 uses a different eligibility denominator.

3. Was the ADHD symptom improvement statistically significant within the 41-person subgroup?

The original paper did not report a valid formal placebo-versus-LDX efficacy significance test in this subgroup. It treated those results descriptively, explicitly noting the inappropriate two-person placebo comparator. Numerical changes of −17.8 versus −13.5 points in the symptomatic subset should not be relabeled as significant, nonsignificant, equivalent, or noninferior without an appropriate test.

4. Which outcomes were genuinely significant or nonsignificant?

The overall safety population’s endpoint pulse difference versus placebo was reported significant at P = .0018. In that same larger population, endpoint systolic and diastolic pressure differences were described as not statistically significant, and no significant meaningful QTcF between-group change was reported. No subgroup-switch P-value was supplied; the clinical symptom changes are descriptive.

5. Did the authors chemically measure amphetamine in venous whole blood?

No. They measured clinical scales, adverse-event reports, routine safety measures, and related observations. The study did not publish validated quantitative dextroamphetamine or lisdexamfetamine results for forensic human venous whole blood or separated plasma/serum, and it supplied no corresponding drug assay detection limits or individual ng/mL results.

6. Can the 30-, 50-, or 70-mg dose determine whether an individual was impaired while driving?

No. A nominal oral prodrug dose and a therapeutic symptom rating are not direct measurements of a person’s drug concentration or driving performance. The study did not test driving, crashes, last-ingestion timing, specimen-specific concentrations, or a validated legal impairment threshold. None can be reconstructed reliably from these subgroup averages alone.

7. How should the 56.4% adverse-event rate be interpreted?

It means 22 of 39 previously amphetamine-treated LDX patients reported at least one treatment-emergent event during this short clinical trial. It does not show that LDX caused every event, establish an individual’s risk, or show LDX was safer than placebo in that subgroup: the comparison involved only two placebo patients. A larger and prospectively specified safety comparison is needed for stable inference.