Author: toxlaw

Restricting an analysis to people who took more study medication can increase an observed treatment difference. It can also change which people are being compared. Those two possibilities must be considered together.

A pilot trial within opioid treatment

Schottenfeld and colleagues’ 2018 Malaysian pilot randomized men with opioid and amphetamine-type stimulant dependence to atomoxetine or placebo alongside buprenorphine/naloxone and counseling. Sixty-six were randomized; one participant without follow-up data was excluded from the reported intention-to-treat analysis. After inpatient stabilization, outpatient follow-up lasted 16 weeks. Oral atomoxetine began at 40 mg/day and increased to 80 mg/day after seven days.

During weeks 9–16, estimated stimulant-negative urine proportions were 0.77 versus 0.67, with Cohen’s d=0.26. The overall medication comparison was nonsignificant (p=0.41). Among participants reporting at least 60 days of study-medication use, the proportions were 0.90 versus 0.64 and d=0.56; that subgroup comprised 12 atomoxetine and 16 placebo participants. Self-reported stimulant-abstinent days did not differ overall (p=0.42).

What changed when adherence defined the sample?

Adherence was self-reported, urine testing was weekly, and the study population was all male and receiving opioid treatment. The outcome concerned amphetamine-type stimulants, not a validated methamphetamine-only consumed dose.

The adherence subgroup was defined using behavior occurring after assignment. Randomization of treatment does not automatically randomize adherence. Participants who continued medication may differ from those who did not in motivation, attendance, symptoms or other determinants of the outcome.

A larger subgroup estimate therefore cannot, by itself, establish how much benefit every patient would receive if adherence improved. That causal question requires a design or analysis capable of addressing the reasons adherence differed. Simply removing less adherent people does not solve it.

Effect size, significance and the specimen outcome

Cohen’s d describes a standardized difference. It is not the percentage of participants cured, the probability that an individual will benefit or a direct measure of milligrams avoided. A pilot estimate can inform the planning of a larger trial while remaining too uncertain for a definitive efficacy claim.

The urine outcome also needs its original name. A proportion of negative collected tests describes observations under the study’s schedule. It should not be restated as the proportion of all days biologically verified drug-free, especially when testing is intermittent.

For a research-based opinion, the defensible account places the small overall estimate, the nonsignificant overall comparison and the selected adherence result together. This preserves the reason for further investigation without treating the subgroup as an independent confirmation. The historical study does not establish a current medication recommendation or generalize automatically to women, people without opioid dependence or settings without the accompanying treatment.

Source

Schottenfeld RS, et al. Atomoxetine for amphetamine-type stimulant dependence during buprenorphine treatment: A randomized controlled trial. Drug Alcohol Depend. 2018;186:130–137. doi:10.1016/j.drugalcdep.2018.01.017. Full text.

Relief of withdrawal symptoms is a meaningful outcome to measure, but it is not interchangeable with evidence of reduced methamphetamine consumption. A historical replacement-therapy trial illustrates why both outcomes belong in a research summary.

The tested regimen and the null use result

Galloway and colleagues’ 2011 randomized, double-blind trial enrolled 60 treatment-seeking adults with methamphetamine dependence. Participants received oral sustained-release dextroamphetamine 60 mg/day or placebo for eight weeks, with psychosocial treatment. After the first day, the daily dose was divided into two equal doses.

Methamphetamine urine testing occurred twice weekly, using an initial 1,000 ng/mL cutoff and GC–MS quantification of positive specimens. The number of methamphetamine-negative samples did not differ significantly (p=0.894), nor did self-reported abstinent days (p=0.842). Withdrawal and craving scores favored dextroamphetamine in the reported models. No serious adverse events were reported.

Why an amphetamine prescription does not invalidate a methamphetamine outcome

The study assessed methamphetamine, rather than simply counting every amphetamine-class result as continued methamphetamine use. This analytical distinction is essential to interpreting a trial in which the assigned medication itself is dextroamphetamine.

For any comparable laboratory record, the first question is what the assay reports: a drug class, a specified parent compound or a metabolite. A result for one of those categories should not silently become a result for another. The specimen method and confirmation procedure determine which comparison is defensible.

The study cutoff also defines its negative category. A specimen below that boundary does not establish that it contains no methamphetamine at any concentration, and it cannot establish uninterrupted abstinence between collection times. That is a limitation on the classification’s meaning, rather than a reason to discard the classification.

Can craving be treated as a surrogate for abstinence?

If reduced craving is proposed as a substitute for reduced use, the relationship must be demonstrated rather than assumed. Here the reported outcomes point in different directions: symptom measures favored medication, while the measured use comparisons did not establish superiority.

That pattern supports asking further research questions about symptom relief and the conditions under which it might influence behavior. It does not show that symptom improvement inevitably produces abstinence. Nor does the use null demonstrate that withdrawal relief was imaginary or unimportant.

The trial was small and brief, and its adverse-event experience cannot establish long-term or uncommon-risk estimates. It supplies historical evidence about the tested regimen in selected participants, without establishing a current prescribing instruction or a general prediction for every person with methamphetamine use disorder.

Source

Galloway GP, et al. A randomized, placebo-controlled trial of sustained-release dextroamphetamine for treatment of methamphetamine addiction. Clin Pharmacol Ther. 2011;89:276–282. doi:10.1038/clpt.2010.307. Full text.

A person reporting less stimulation has not necessarily experienced less cardiovascular stress. Subjective ratings, pulse and blood pressure answer different questions, even when they are collected during the same drug challenge.

What did the historical trial find?

Verrico and colleagues’ 2016 double-blind laboratory study randomized 60 non-treatment-seeking adults with methamphetamine use disorder; complete data were available for 59. Oral perindopril conditions were placebo, 4, 8 or 16 mg/day. After 5–7 days, participants received 15- and 30-mg intravenous methamphetamine challenges on alternate days.

The 8-mg perindopril group had lower peak anxious and stimulated ratings than placebo after correction for multiple comparisons (p=0.0009 and p=0.0070). A greater perindopril dose did not produce a progressively greater effect. Posttreatment cardiovascular treatment comparisons were nonsignificant; methamphetamine still increased the cardiovascular measures. Treatment-by-methamphetamine interactions were nonsignificant across the subjective ratings and cardiovascular measures.

A treatment main effect is not a drug-specific interaction

The study excluded important medical and psychiatric comorbidities and used a brief inpatient protocol. It did not establish outpatient abstinence or prevention of cardiovascular events.

A main effect of treatment describes a difference across treatment groups under the fitted model. An interaction asks whether the response to methamphetamine changes with treatment in the modeled comparison. Those propositions should not be treated as interchangeable merely because the experiment included both drugs.

The distinction is particularly useful when evaluating a claim that perindopril specifically blocked a methamphetamine effect. Such a claim should identify the relevant comparison and acknowledge the interaction result. The lower ratings remain findings; explaining their statistical form prevents them from acquiring a stronger mechanism claim than the analysis supplies.

Why the cardiovascular result must remain visible

It would be inappropriate to infer cardiovascular protection from reduced anxiety or stimulation scores. Demonstrating protection would require a suitable cardiovascular outcome and enough evidence to evaluate that outcome. A symptom rating cannot supply an unmeasured event comparison.

A nonsignificant cardiovascular difference also does not prove that the regimens are equivalent. That conclusion would require a suitable equivalence question, margins and precision. An ordinary null comparison and a successful equivalence test have different meanings.

The appropriate use of this paper is therefore narrow but valuable: it identifies a selected subjective signal within a controlled human experiment while retaining the cardiovascular null findings and the irregular dose pattern. It does not justify taking perindopril to counteract methamphetamine, nor does it predict an individual’s medical safety from how stimulated they feel.

Source

Verrico CD, et al. Subjective and Cardiovascular Effects of Intravenous Methamphetamine during Perindopril Maintenance: A Randomized, Double-Blind, Placebo-Controlled Human Laboratory Study. Int J Neuropsychopharmacol. 2016;19:pyw029. doi:10.1093/ijnp/pyw029. Full text.

A question about intended drug use and an opportunity to take the drug are different measures. A medication can change an answer on a rating scale without changing behavior in the laboratory task used to assess drug taking.

The participants and the actual choices

De La Garza and colleagues’ 2012 double-blind crossover experiment enrolled 20 adults with methamphetamine dependence who were not seeking treatment. Seventeen completed all conditions; excluding two methamphetamine nonresponders left 15 in the analysis. Oral rivastigmine conditions were 0, 3 or 6 mg/day, with treatment on days 2–8. Acute methamphetamine challenges were 15 and 30 mg intravenously.

Rivastigmine attenuated the peak “Likely to Use METH” rating (p=0.01), particularly with 6 mg rivastigmine and the 15-mg challenge. The peak desire result was nonsignificant (p=0.27). Rivastigmine did not reduce total methamphetamine choices in a separate task offering 5-mg intravenous infusions versus no infusion. No monetary alternative was offered.

A rating is not a validated substitute for drug taking

Calling the rating result reduced methamphetamine use would replace the measured outcome with a different one. The appropriate explanation keeps the words on the questionnaire visible and then reports the behavioral result separately.

That separation is more informative than declaring the whole experiment positive or negative. A subjective effect may help develop a hypothesis about how a medicine changes experience. The behavioral task asks whether that change carries into the specific opportunity to obtain drug. Neither outcome should erase the other.

The denominator also affects the inference. The final group excluded people who did not show the required acute methamphetamine response. A finding in that selected group cannot automatically describe everyone with methamphetamine dependence, including people who would not have met that response condition.

Why the alternative in a choice task matters

A decision between an infusion and no infusion is not the same decision as drug versus money, work, family commitments or another competing reward. The task defines what a choice costs and what can be obtained instead. Interpretation therefore requires the actual alternatives, rather than the phrase “self-administration” alone.

The null behavioral result does not establish that every possible rivastigmine regimen must fail. Conversely, the favorable rating does not demonstrate outpatient effectiveness. A treatment claim would need a suitable clinical comparison with outcomes that directly address the proposed benefit.

For an expert report, this study is useful as a concrete example of why intention, subjective drug effects and observed consumption require separate descriptions. It is historical human laboratory evidence, not a current recommendation to use rivastigmine for methamphetamine-related conditions.

Source

De La Garza R II, et al. Rivastigmine reduces “Likely to use methamphetamine” in methamphetamine-dependent volunteers. Prog Neuropsychopharmacol Biol Psychiatry. 2012;37:141–146. doi:10.1016/j.pnpbp.2011.12.014. Full text.

A pattern found after examining trial data can identify a research question without becoming a clinical prediction rule. The distinction is particularly important when a symptom measured during treatment is presented as if it predicted the outcome before treatment began.

Who was included in the reanalysis?

Zorick and colleagues’ 2011 paper reanalyzed a 12-week sertraline trial. Of 229 randomized participants, 129 completed at least eight weeks and entered this analysis: 61 assigned sertraline and 68 placebo. Sertraline began orally at 50 mg/day and increased to 100 mg/day on day eight. Participants also received psychosocial treatment.

The investigators compared baseline methamphetamine-positive urine proportions with weeks 8–12. A greater-than-15% increase classified 13 sertraline participants and five placebo participants as increased users. The 15% breakpoints came from the observed distribution, rather than an independently validated rule. Classification-tree analyses identified sustained in-treatment craving as characteristic of the sertraline subgroup with increased use. Only 18 people comprised the increased-use subgroup.

Three limits on interpreting the finding

First, a retained-participant analysis is not the same population as the original randomized sample. Requiring eight weeks of participation introduces selection after assignment. A report should preserve that condition before quoting the subgroup percentages.

Second, a data-derived threshold can be useful for exploration while remaining sensitive to the particular sample. Demonstrating prediction would require checking the rule in appropriate independent data, with attention to false positives, false negatives and calibration. A statistically significant group comparison does not perform that validation.

Third, craving measured during treatment and drug use during treatment share a timeline. Their association does not by itself establish that craving caused the increased use, or that sertraline produced the pattern through a particular biological mechanism. Temporal detail and a suitable causal design would be needed to distinguish those explanations.

What the urine outcome means

The classification was based on a change in the proportion of positive specimens. It was not a direct measurement of an increase in milligrams consumed. It also was not a test of driving impairment. Keeping the outcome in its original form prevents a percentage classification from acquiring a different biological meaning in an expert report.

The paper supplies a cautionary historical observation about a specific treatment experiment. It does not establish that the same response occurs with every serotonin-related medicine, nor does it replace assessment of an individual receiving an antidepressant for another indication. Treatment changes cannot be justified by relabeling this exploratory analysis as a validated personal forecast.

Source

Zorick T, Sugar CA, Hellemann G, Shoptaw S, London ED. Poor response to sertraline in methamphetamine dependence is associated with sustained craving for methamphetamine. Drug Alcohol Depend. 2011;118:500–503. doi:10.1016/j.drugalcdep.2011.04.015. Full text.

Failure to detect a relationship within a selected sample does not establish that the relationship is absent across all patients. This matters when an older-adult pharmacokinetic study is used to dismiss renal function as a consideration in interpreting amphetamine exposure.

What population received the dose?

Ermer and colleagues’ 2013 double-blind crossover study enrolled 47 healthy adults aged 55 years or older. Participants received single oral lisdexamfetamine 50 mg and placebo doses, separated by 7–14 days. The age groups were 55–64, 65–74 and at least 75 years. Mean baseline creatinine clearance was within the normal range in every group.

Active d-amphetamine exposure was modestly higher and clearance slower in the oldest group. Among women, mean peak plasma concentration was 51.0 ± 6.7 ng/mL at ages 55–64 versus 64.3 ± 12.1 ng/mL at ages at least 75; median peak times were 3.5 and 5.5 hours. The reported spreads are standard deviations.

The study found no relationship between d-amphetamine clearance and baseline creatinine clearance, and no consistent age pattern in pulse or blood-pressure changes. It excluded important comorbidities and did not represent medically complex older patients.

The age comparison was not randomized

Randomization determined treatment sequence, rather than age. The crossover strengthened the comparison between the administered treatments, but it did not randomly assign people to different ages. An observed age pattern therefore should not be described as an isolated experimental effect of aging.

Likewise, the renal finding concerns the range represented in the enrolled group. A conclusion that kidney function never affects amphetamine disposition would extend the inference beyond that range. It would require evidence from participants with the relevant degrees of impairment, rather than a null association in a healthy sample.

Can these concentrations predict an individual’s functioning?

The reported values describe concentration–time profiles after a known dose under study conditions. They do not provide a validated impairment threshold. A person’s measured concentration should not be matched to an age-group mean and then assigned the group’s presumed behavior.

Peak timing also deserves attention. A specimen collected at a particular hour is an observation at that hour. Comparing it with a study’s maximum concentration requires establishing how sampling time relates to administration and the observed curve, rather than assuming the sample coincided with the peak.

The useful application is to frame questions about age, specimen timing and the applicability of the comparison population. This historical single-dose study cannot establish dose proportionality in older adults, chronic safety or a dosing rule for patients with significant renal or hepatic dysfunction.

Source

Ermer J, et al. Double-blind, placebo-controlled, two-period, crossover trial to examine the pharmacokinetics of lisdexamfetamine dimesylate in healthy older adults. Neuropsychiatr Dis Treat. 2013;9:219–229. doi:10.2147/NDT.S38377. Full text.

A plausible explanation for a negative trial is still a hypothesis. If counseling intensity was identical across medication groups, the trial cannot establish what would have happened with less counseling.

Three doses, a common counseling program

Johnson and colleagues’ 2008 report described 150 adults in an eight-week, double-blind, placebo-controlled methamphetamine-dependence trial. Oral ondansetron doses were 0.25, 1 or 4 mg twice daily. Every group was offered 90-minute cognitive behavioral therapy sessions three times weekly.

No dose outperformed placebo on the primary methamphetamine-use measures, including urine outcomes, reported nonuse days and sustained abstinence. Craving changes also did not differ significantly (p=0.63). The urine classification used a 300 ng/mL boundary. Across groups, 44.19% of urine samples were missing; a methamphetamine-free week required the available samples to be negative.

What does the counseling explanation actually show?

The authors proposed that less intensive counseling might have made a medication effect easier to detect. Counseling intensity was not randomized. The report also identified limited sample size, short duration and loss to follow-up as interpretive limitations.

The randomized comparison estimates the added effect of ondansetron in the tested treatment setting. Because counseling was common to the arms, its presence is part of that setting. It does not invalidate the comparison. A different setting could produce a different result, but establishing that possibility requires another comparison rather than a revised description of this one.

There is also no demonstrated interaction between medication efficacy and counseling intensity here. To establish such an interaction, investigators would need evidence that the medication’s relative effect changed across counseling conditions. Offering the same counseling program to everyone leaves that question unanswered.

How missing urine samples change the reading

A week classified using available specimens should not be described as continuous biologically verified abstinence. The strength of the claim depends on the observation schedule and the handling of absent observations. A negative collected specimen and an uncollected specimen supply different information.

That distinction works in both directions. Missing samples should not automatically be called evidence of use unless that was an explicitly defined analytical rule, and they should not be silently counted as evidence of nonuse. The appropriate report identifies the rule and separates the recorded classification from the larger claim it might be used to support.

This historical trial does not support an ondansetron benefit under its tested conditions. It leaves room for further research into other conditions, without establishing efficacy in those untested settings or providing a current treatment recommendation.

Source

Johnson BA, et al. A preliminary randomized, double-blind, placebo-controlled study of the safety and efficacy of ondansetron in the treatment of methamphetamine dependence. Int J Neuropsychopharmacol. 2008;11:1–14. doi:10.1017/S1461145707007778. Publisher full text.

A trial can produce an unsuccessful primary result alongside observations worth investigating. The task is to preserve both, without promoting a secondary finding into the answer to the original question.

What was the trial designed to test?

Elkashef and colleagues’ historical eight-center, double-blind trial randomized 140 adults with methamphetamine dependence to oral topiramate or placebo for 13 weeks: 69 and 71 participants, respectively. Treatment started at 25 mg/day and targeted 200 mg/day during weeks 6–12, followed by tapering. Both groups received brief behavioral support.

The primary abstinence comparison during weeks 6–12 was not significant (p=0.13). A secondary analysis found at least a 25% reduction in median urine methamphetamine concentration in 64.2% of topiramate recipients versus 42.3% of placebo recipients (p=0.03). A subgroup of 26 participants with negative baseline urine also favored topiramate on abstinence. Retention and achievement of the target dose were limited; pill counts measured adherence.

Why the outcome hierarchy matters

The corrected PubMed abstract specifies the 25-mg starting dose; the archived abstract contains 50 mg, while the full methods also specify 25 mg. Preserving that correction avoids repeating an error in the treatment description.

The primary result asks whether assignment to the tested regimen improved the specified abstinence outcome in the trial population. A subgroup selected by baseline urine status asks a narrower question. A positive answer to the narrower question does not retroactively change the overall result, and a subgroup of 26 is not an independent replication.

Baseline urine status also should not be translated into a verified long history of abstinence. The classification describes evidence available at entry. A single negative result and a sustained pattern of abstinence are different observations.

Does lower urine concentration mean a known reduction in dose?

A concentration-based outcome and a count of days without use are not interchangeable. The units of measurement should remain visible: a urine result expresses analyte in a specimen, whereas a consumed dose expresses an amount administered. Calling the former a measured dose reduction would require an additional validated relationship.

In evaluating the paper, an expert should therefore identify which outcome supports each sentence. The primary abstinence analysis, the quantitative urine analysis and the baseline-negative subgroup belong in separate parts of the explanation. Doing so preserves a possible research signal while keeping the main null result visible.

These data can motivate a more focused trial. They do not establish a current prescribing recommendation, nor do they show that every person who begins treatment with a negative test will benefit.

Source

Elkashef A, et al. Topiramate for the treatment of methamphetamine addiction: a multi-center placebo-controlled trial. Addiction. 2012;107:1297–1306. doi:10.1111/j.1360-0443.2011.03771.x. Full text; corrected bibliographic abstract.

An interaction study can answer a focused exposure question without showing that a drug combination improves symptoms. That distinction matters when a paper’s conclusion about tolerability is used to support a broader treatment claim.

The actual combination and population

Faison and colleagues’ 2021 randomized, open-label crossover experiment enrolled 36 healthy adults without ADHD. Single oral treatments were viloxazine extended release 700 mg, lisdexamfetamine 50 mg, and the combination. Plasma was sampled through 96 hours. The authors selected the adult viloxazine dose to approximate exposures relevant to pediatric development; it was not a pediatric dosing experiment.

For active d-amphetamine, combination-versus-alone geometric mean ratios were 112.78% for peak concentration and 109.52% for exposure extrapolated to infinity. Their 90% confidence intervals were 109.93%–115.71% and 105.19%–114.03%, respectively. Viloxazine exposure measures also met the predefined 80%–125% bounds. No serious adverse events were reported.

“No interaction” needs its statistical definition

These results support the paper’s prespecified exposure classification. They do not mean that every concentration was identical. The amphetamine point estimates were above 100%, while their intervals remained within the chosen bounds. The authors also acknowledged that single doses may not represent chronic therapy.

This is why an expert explanation should state the criterion before using the phrase “no interaction.” A reader otherwise cannot tell whether the phrase means identical measurements, failure to reject a difference, or success against an equivalence margin. Those are different statistical propositions.

The measured stimulant analyte also matters. Lisdexamfetamine was the administered prodrug; the reported stimulant exposure comparison concerned d-amphetamine. A summary that treats the two names as interchangeable obscures which concentration was measured and which dose was swallowed.

What would establish added clinical benefit?

To evaluate whether the combination improves ADHD outcomes, a study would need participants with the relevant condition, a clinical outcome and an appropriate comparator. Establishing relative bioavailability does not supply a symptom-response comparison. It also does not rank the combination against other treatment strategies.

Likewise, a small short exposure study cannot provide the denominator needed to estimate uncommon harms during extended treatment. The absence of a serious event is a finding about the observed study experience. Calling it a guarantee would change both the population and the timescale without supplying additional evidence.

For a client reviewing a medication record, the defensible use is specific: this experiment informs a particular pharmacokinetic interaction question. Assessment of the person’s clinical response still requires their timing, symptoms and observations. The historical paper is not an instruction to begin, combine or alter medication.

Source

Faison SL, et al. Pharmacokinetics of Coadministered Viloxazine Extended-Release (SPN-812) and Lisdexamfetamine in Healthy Adults. J Clin Psychopharmacol. 2021;41:155–162. doi:10.1097/JCP.0000000000001361. Full text.

Combining a medicine associated with lower pulse and blood pressure with a stimulant does not establish a physiological balancing act. The useful question is whether the combination was actually tested for that outcome, in an appropriate population, for long enough to support the proposed conclusion.

What did the controlled experiment measure?

Roesch and colleagues’ 2013 open-label crossover study randomized 42 healthy adults; 40 completed it. Participants received single oral doses of guanfacine extended release 4 mg, lisdexamfetamine 50 mg, and their combination, separated by at least seven days. Plasma sampling continued through 72 hours.

The 90% confidence intervals for combination-versus-alone d-amphetamine peak concentration and total exposure were 0.967–1.019 and 0.983–1.060. Both met the study’s 0.80–1.25 criterion. Guanfacine peak concentration increased approximately 19%; its confidence interval, 1.066–1.321, exceeded that criterion, although total exposure met it.

Exposure equivalence does not establish cardiovascular cancellation

Observed supine pulse and blood-pressure changes with the combination resembled lisdexamfetamine alone, with modest increases. The authors explicitly cautioned that the experiment was not designed to assess cardiovascular effects robustly. Participants with relevant cardiac comorbidities were excluded, and the study used single doses rather than steady-state treatment.

There are two separate comparisons here. A pharmacokinetic comparison asks how coadministration changes the concentration–time profile. A cardiovascular comparison asks how the resulting drug effects interact in the person. Similar amphetamine exposure addresses the first comparison; it cannot, by itself, answer the second.

Nor does a guanfacine confidence interval outside the chosen boundary establish a dangerous interaction. The boundary supplies a prespecified statistical decision rule. Translating its failure into a clinical consequence requires evidence about the size, timing and relevance of the change. Conversely, satisfying that rule for amphetamine is not a guarantee against adverse effects.

How should this paper be used in a case review?

A report relying on this experiment should identify the proposed inference precisely: altered amphetamine exposure, an effect on blood pressure, or protection against a particular adverse event. Those claims require different evidence. A claim of cardiovascular protection would need an outcome comparison designed to test protection, rather than a concentration comparison relabeled after the fact.

For an individual record, the practical comparison remains the documented medication schedule, observations and measurements. An apparently balanced list of prescribed drugs cannot substitute for those observations. This historical healthy-volunteer experiment supplies evidence about the tested combination; it does not establish a universal rule for children, chronic treatment or patients with cardiovascular disease.

Source

Roesch B, et al. Pharmacokinetics of Coadministered Guanfacine Extended Release and Lisdexamfetamine Dimesylate. Drugs R&D. 2013;13:119–128. doi:10.1007/s40268-013-0014-8. Full text. This article interprets historical human experimental evidence.