CLINICAL TRIALS
`
`Acute tolerance to methylphenidate in the
`treatment of attention deficit hyperactivity
`disorder in children
`
`Objectives: To evaluate the efficacy of several drug delivery patterns of methylphenidate and to determine
`whether acute tolerance develops to this widely used stimulant medication in the treatment of children
`with attention deficit hyperactivity disorder.
`Methods: Double-blind trials were conducted in a laboratory school setting in which multiple measures of
`efficacy were obtained frequently in the morning and afternoon across the school day. In study I, relative
`efficacy was determined for three dosing patterns of methylphenidate: a standard twice-daily profile, a
`flat profile, and an ascending profile. In study II, tolerance was assessed by comparison of three-times-a-
`day regimens in which the time of the middle dose varied.
`Results: In study I, the efficacy of the ascending treatment increased across the day, and in the afternoon
`it was equal to the efficacy of the twice-daily treatment, indicating that an initial bolus was not required
`for efficacy. The efficacy of the flat treatment declined across the day, and in the afternoon it was signifi-
`cantly less than in the twice-daily treatment, suggesting that tolerance may be developing. In study II,
`acute improvements in efficacy were reduced to the second of two closely spaced but not to two widely
`spaced bolus doses, suggesting that shortly after exposure to high concentrations, efficacy is reduced to
`given concentrations of methylphenidate. In a concentration–effect model, a tolerance term was needed to
`account for counterclockwise hysteresis.
`Conclusions: Acute tolerance to methylphenidate appears to exist. This should be considered in the design
`of an optimal dosing regimen for the treatment of children with attention deficit hyperactivity disorder.
`(Clin Pharmacol Ther 1999;66:295-305.)
`
`James Swanson, PhD, Suneel Gupta, PhD, Diane Guinta, PhD, Daniel Flynn, MSW,
`Dave Agler, MA, Marc Lerner, MD, Lillie Williams, MD, Ira Shoulson, MD, and
`Sharon Wigal, PhD Irvine and Palo Alto, Calif, and Rochester, NY
`
`Stimulant medications have been used for about 50
`years1 to treat children with attention deficit hyperactivity
`disorder (ADHD).2 Standard clinical practice has
`remained essentially unchanged since amphetamine (INN,
`amfetamine)3 was replaced by methylphenidate4-6 about
`
`From the University of California Irvine, Irvine; ALZA Corporation,
`Palo Alto; and University of Rochester, Rochester.
`Supported by ALZA Corporation, Palo Alto, Calif.
`Received for publication Oct 26, 1998; accepted June 8, 1999.
`Reprint requests: James Swanson, PhD, University of California
`Irvine, Child Development Center, 19722 MacArthur Blvd, Irvine,
`CA 92612.
`Copyright © 1999 by Mosby, Inc.
`0009-9236/99/$8.00 + 0 13/1/100597
`
`25 years ago as the primary stimulant prescribed to treat
`ADHD. Methylphenidate use has increased dramatically,7
`and now more than 10 million prescriptions are written
`for methylphenidate each year in the United States.8
`Methylphenidate releases and inhibits uptake of cate-
`cholamines (primarily dopamine),9-10 and the resulting
`increase in these neurotransmitters is considered to be
`the basis for its clinical efficacy. Within 1 to 2 hours
`after oral administration of a clinical dose of
`methylphenidate, peak serum concentration is achieved
`and maximum clinical effects are manifested (ie,
`decreases in the symptoms of ADHD: hyperactivity,
`inattention, and impulsivity).11-15 Methylphenidate has
`a short pharmacokinetic half-life and an equally short
`
`295
`
`Page 1
`
`SHIRE EX. 2045
`KVK v. SHIRE
`IPR2018-00290
`
`

`

`296 Swanson et al
`
`CLINICAL PHARMACOLOGY & THERAPEUTICS
`SEPTEMBER 1999
`
`Fig 1. Study I: Simulated plasma methylphenidate concentrations for a 20-mg total daily dose
`delivered by twice-daily (bid), flat, and ascending dosing regimens.
`
`duration of efficacy of about 2 to 3 hours,11-15 so twice-
`daily (bid) or three-times-a-day (tid) dosing is typical.
`Because the clinically effective dose varies among chil-
`dren (from 5 to 20 mg per administration), individual
`titration is required. Years of clinical practice16-19 con-
`firm that bid or tid dosing regimens provide effective
`and safe treatment. Even though some observations of
`tolerance have been noted (eg, to side effects and in cer-
`tain high dosing regimens),20 in most cases clinical
`effectiveness is maintained over years of treatment with-
`out increasing dose.21 It therefore appears that children
`with ADHD do not develop long-term tolerance to treat-
`ment with typical clinical doses of methylphenidate.
`The short duration of efficacy of methylphenidate
`creates serious practical problems for effectiveness
`(because of waxing and waning of effects), for compli-
`ance (because of frequent missed doses with multiple
`daily administrations), for privacy (because of the need
`to administer medication at school or in other public
`settings), and for protection against diversion of this
`controlled drug (because of the difficulty in controlling
`access by others when it is stored outside the home).
`To address these problems, sustained-release prepa-
`rations intended for once-a-day administration of
`methylphenidate22-26 (and amphetamine)26-28 have been
`developed, but they are not considered to be as effective
`as multiple doses of immediate-release preparations and
`
`are not widely accepted for clinical use.8,24 The reasons
`for reduced efficacy of sustained-release preparations
`are unknown. Two characteristics of sustained-release
`patterns of drug delivery may contribute to reduced effi-
`cacy. First, a reduced or delayed bolus of drug (com-
`pared with the immediate-release pattern) may not result
`in sufficient increase in brain catecholamines to produce
`standard clinical effects.9,10,22,27,28 Second, the continu-
`ous rate of drug delivery may produce acute tolerance
`(tachyphylaxis).29
`Two studies were conducted to test whether these
`characteristics of drug delivery contribute to reduced
`efficacy of methylphenidate on measures of behavior
`and attention. Study I investigated the time course of
`efficacy produced by two experimental patterns of drug
`delivery established by frequent dosing (a large bolus
`followed by small constant doses that generates a flat
`profile and a small bolus followed by small increasing
`doses that generates an ascending profile), compared
`with the standard pattern of drug delivery in clinical
`treatment (two large bolus doses that generate the bid
`profile of peaks and troughs) and a placebo control con-
`dition. Study II established two experimental tid dos-
`ing regimens in which the timing of the second bolus
`dose was varied to establish patterns of peaks and
`troughs appropriate for the evaluation of acute toler-
`ance and for pharmacodynamic modeling of the rela-
`
`Page 2
`
`

`

`CLINICAL PHARMACOLOGY & THERAPEUTICS
`VOLUME 66, NUMBER 3
`
`Swanson et al 297
`
`tionship between simulated methylphenidate concen-
`tration and effect.
`
`METHODS
`Study I. Three methylphenidate delivery profiles
`(bid, flat, and ascending) and placebo were compared
`in a four-period, double-blind, randomized crossover
`study. Behavior, attention, and cognitive performance
`measures were taken frequently in a laboratory school
`setting in which children diagnosed with ADHD expe-
`rience repeated classroom sessions across each day.31-32
`The bid regimen (twice-daily dosing with immediate-
`release methylphenidate) was expected to produce peaks
`and troughs (Fig 1) in drug concentration during the typ-
`ical school day.9,12-15 The flat regimen was designed to
`provide an initial peak and then a constant methyl-
`phenidate concentration throughout the day. The ascend-
`ing regimen was designed to produce an increasing
`methylphenidate level from a low drug concentration
`(ie, the bid trough level) established early in the morn-
`ing to a high drug concentration (ie, the bid peak level)
`by the end of the day. All treatments were administered
`in identical capsules given precisely at 30-minute inter-
`vals throughout the day (an initial capsule at 7:30 AM
`followed by capsules at 8:30, 9, 9:30, 10, 10:30, 11,
`11:30 AM and at 12, 12:30, 1, 1:30, 2, 2:30, and 3 PM).
`The timing of actual drug administration differed across
`regimens to create the desired drug delivery pattern and
`expected concentration profiles. For example, only two
`of the capsules administered in the bid regimen con-
`tained methylphenidate, whereas all capsules adminis-
`tered in the flat and ascending regimens contained
`methylphenidate, but in differing amounts.
`Thirty-eight children (33 boys and five girls; age
`range, 7 to 12 years; mean age, 9.2 years), with clini-
`cal diagnoses of ADHD and receiving current treatment
`with methylphenidate doses of 5 to 15 mg administered
`two or three times per day, were recruited for this trial.
`Parents signed consent forms and children signed assent
`forms to enter a protocol approved by the University of
`California Irvine Institutional Review Board. A struc-
`tured interview (Diagnostic Interview Schedule for
`Children)30 was used to confirm the diagnosis of
`ADHD based on DSM-IV criteria, including onset by
`7 years of age, presence of at least six of the nine symp-
`toms in the Inattention or the Hyperactive-Impulsive
`domains, and significant impairment in at least two set-
`tings (eg, home and school).
`Each of two cohorts of children was evaluated in the
`laboratory school setting from 7 AM until 6 PM on five
`consecutive Saturdays. On the first Saturday, a cohort
`was first divided by age into two groups. These groups
`
`were then introduced to the staff and became familiar
`with the setting of the classroom (staffed with one
`teacher and one classroom aide) and the playground
`(staffed with four recess aides). On subsequent Satur-
`days, each child received (in random order) one of the
`following treatments: (1) bid: two doses of immediate-
`release methylphenidate (Ritalin hydrochloride) admin-
`istered 41⁄2 hours apart as the total daily dose; (2) flat:
`an initial loading dose of immediate-release methyl-
`phenidate equal to 80% of the morning dose of the
`bid condition, with the remaining amount of the total
`daily dose administered (starting 11⁄2 hours later) in small
`equal doses at 30-minute intervals over 6 hours; (3)
`ascending: an initial loading dose of immediate-release
`methylphenidate equal to 40% of the morning dose of the
`bid condition, with the remaining amount of the daily
`dose administered (starting 11⁄2 hours later) in small
`increasing doses administered at 30-minute intervals over
`5 hours; (4) placebo: lactose administered in all capsules.
`On study days, the teacher evaluated each child after
`four 30-minute group classroom sessions, and each
`child was tested on a computer in a 30-minute individ-
`ual laboratory session immediately before or after each
`of these classroom sessions. These laboratory school31
`evaluations were scheduled at 1 and 31⁄2 hours after the
`bid dosing times to coincide with expected peaks and
`troughs in methylphenidate plasma concentration in the
`bid regimen. Each classroom session had similar writ-
`ten seat work (eg, solving math problems) and group
`activities (eg, listening to and discussing a presentation
`to the class). Classroom rules that defined appropriate
`and inappropriate behavior were established and, after
`each classroom session, teachers completed the CLAM
`and the SKAMP rating scales32-34 to provide subjec-
`tive but systematic evaluations of several dimensions
`of behavior.35 The CLAM scale has 16 symptom-
`related items that are rated on a four-point scale (not
`at all, just a little, pretty much, and very much). It
`provides three established index scores based on aver-
`aging ratings from subsets of items (10 items for the
`Conners hyperactivity index, five items for the inatten-
`tion/overactivity index, and five items for the aggres-
`sion/defiance index). The SKAMP scale has 10 items
`describing problem behaviors in the classroom setting
`that are rated on a seven-point impairment scale (none,
`slight, mild, moderate, severe, very severe, or maxi-
`mal). It provides two established index scores based on
`averaging ratings from subsets of items (six items for
`the attention index and four items for the deportment
`index). In each laboratory session, children were tested
`on a display-memory scanning task to provide objec-
`tive measures (reaction time and accuracy) of cognitive
`
`Page 3
`
`

`

`298 Swanson et al
`
`CLINICAL PHARMACOLOGY & THERAPEUTICS
`SEPTEMBER 1999
`
`Fig 2. Study II: Simulated plasma methylphenidate concentrations for a 30-mg total daily dose
`delivered by the following dosing regimens: three times a day (tid), tid with the middle bolus given
`2 hours after the first (tid-AM), and tid with the middle bolus given 6 hours after the first (tid-PM).
`
`performance.36-37 In this task, a set of either one or four
`digits (the memory set) is presented on a computer
`screen, followed by a set of four digits (the display set).
`The subject is asked to press one of two buttons to indi-
`cate whether or not the display set contained any one
`of the memory set items.
`The desired methylphenidate profiles for each treat-
`ment were determined by simulation with use of pub-
`lished literature values9-15 and a nominal daily dose of
`20 mg. The doses required to produce the flat and
`ascending profiles were determined by deconvolution
`by use of a pharmacokinetic model with first-order
`absorption and one-compartment disposition. Fig 1
`shows the simulated plasma methylphenidate concen-
`tration–time graphs for bid, flat, and ascending drug
`delivery profiles.
`A mixed-effects ANOVA model was used to analyze
`the efficacy of the bid, flat, ascending, and placebo regi-
`mens. The ANOVA model included fixed-effect factors
`(regimen, session, sequence, and period) and the ran-
`dom-effect factors (intersubject and intrasubject effects).
`An overall among-regimen comparison at each time
`point (session) was conducted with an a = .05 signifi-
`cance level. In addition, three pairwise comparisons of
`methylphenidate regimens within each session were esti-
`mated if the overall among-regimen difference was sig-
`
`nificant. For these comparisons, the least-squares esti-
`mate of the mean difference between the two regimens
`and its 95% confidence interval were calculated. No fur-
`ther adjustments to the significance level were made.
`Study II. After a three-period, double-blind, random-
`ized crossover trial of three treatments (tid, ascending,
`and placebo), a parallel design was used to evaluate two
`experimental tid conditions. In all drug delivery regi-
`mens, children took one capsule every 1⁄2 hour for
`8 hours, starting at 7:30 AM. In the initial crossover phase,
`a tid regimen was used as the clinical standard (rather
`than the bid regimen as in study I) to extend the length
`of expected drug effects to 10 hours. Each subject
`received doses of methylphenidate at 7:30 AM,
`11:30 AM, and 3:30 PM, and each dose was equal to the
`child’s clinically titrated morning dose. In the ascending
`regimen, 80% of each subject’s usual morning dose was
`administered at 7:30 AM, followed by small increasing
`doses administered at 30-minute intervals across the day.
`In the parallel phase, subjects were randomly assigned
`to one of two tid conditions in which the middle bolus
`was varied to shift the second peak to an earlier or later
`time. In both experimental tid regimens, the first and last
`bolus doses were administered at 7:30 AM and at
`3:30 PM, but the time of the middle bolus was either
`9:30 AM (tid-AM) or 1:30 PM (tid-PM). In both of these
`
`Page 4
`
`

`

`CLINICAL PHARMACOLOGY & THERAPEUTICS
`VOLUME 66, NUMBER 3
`
`Swanson et al 299
`
`conditions, the first dose was always equal to the child’s
`usual morning dose and was ~33% of the total daily study
`dose. The deconvolution procedure and pharmacokinetic
`model that were applied in study I were used again in
`study II to select the second and third doses so that the
`magnitude of the peak after the second dose would match
`the magnitude of the midday peak in the standard tid con-
`dition (Fig 2). For the tid-AM condition, the second and
`third doses were set at ~21% and ~45%, respectively, of
`the total daily study dose, and for the tid-PM condition the
`second and third doses were set at ~39% and ~27%. The
`tid and ascending treatments were included for an effi-
`cacy analysis, which will be reported elsewhere; the
`experimental tid-AM and tid-PM regimens were used to
`provide data for the pharmacodynamic analysis of
`methylphenidate, which is the topic of this report.
`Thirty-two children (28 boys and four girls; age
`range, 7 to 12 years; mean age, 9.9 years) who had a
`diagnosis of ADHD and were being clinically treated
`with methylphenidate were recruited for this study.
`Their parents signed consent forms and the children
`assent forms approved by the University of California
`Irvine Institutional Review Board. The methods of study
`I for confirming a diagnosis of ADHD were used again
`in study II, and similar evaluation procedures were used
`for the evaluation of efficacy. The participating children
`were tested in the same laboratory school setting31 for
`4 study days, with at least 24 hours between each study
`day. The 32 children were evaluated in one cohort. On
`the first day, the cohort was divided into two groups of
`16 children based on age, and the groups became famil-
`iar with the staff and setting of the classroom (staffed
`with two teachers and two aides for 16 students) and the
`playground (staffed by eight recess aides). On each day,
`hourly cycles of activities31 were scheduled to allow for
`frequent classroom probes of attention and behavior
`across the day. The hourly cycle consisted of capsule
`administration (1 minute), computer math tests (9 min-
`utes), individual classroom seat work (20 minutes), cap-
`sule administration (1 minute),
`library quiet time
`(9 minutes), and group classroom activity (20 minutes).
`This cycle was repeated for 10 hours (8 AM to 5 PM),
`with substitutions in five of the cycles (ie, those start-
`ing at 9 and 11 AM and at 1, 3, and 5 PM) to allow for
`recess sessions and meals during the day.
`Plasma methylphenidate profiles were simulated with
`a nominal daily dose of 30 mg and deconvoluted with
`the same techniques as those used in study I. The sec-
`ond doses in the tid-AM and tid-PM treatments were
`designed to achieve earlier or later maximal plasma con-
`centrations equivalent to the midday peak after the mid-
`day second dose in the standard tid regimen (Fig 2).
`
`One subjective and one objective measure of efficacy
`were chosen for evaluations of acute tolerance. The sub-
`jective measure was the attention subscale of the
`SKAMP rating scale,32-34 expressed as the average rat-
`ing per item. The objective measure was a measure of
`activity obtained from a motion detector (Actigraph,
`Mini Motionlogger Actigraphs, Ambulatory Monitoring
`Inc, Ardsley, NY) worn on the nondominant wrist. A 5-
`second acquisition period was specified, and the counts
`were integrated over each of the precisely timed 20-
`minute periods of seat work activity in the classroom.
`These two primary efficacy measures (SKAMP
`attention and Actigraph activity) were obtained during
`the classroom seat work activities of the laboratory
`school cycles.31 For each of the 10 classroom sessions,
`mean values for attention and activity in the placebo
`condition were subtracted from the mean values for the
`two experimental conditions (tid-AM and tid-PM) to
`remove nondrug related within-day variability. To
`improve clarity in the visual display of the pharmaco-
`kinetic-pharmacodynamic relationship, this difference
`score was multiplied by –1 so that clinical improvement
`reflected in the pharmacodynamic measure would be
`positive and match the direction of change in the phar-
`macokinetic measure (the simulated plasma concentra-
`tion of methylphenidate). These pharmacodynamic
`measures (placebo-adjusted efficacy scores) were plot-
`ted versus the pharmacokinetic measures (expected
`drug concentrations), with time coded by arrows, to
`evaluate whether evidence of tolerance (a counterclock-
`wise hysteresis loop) was present in these data.
`A mathematical pharmacokinetic-pharmacodynamic
`model of the relationship between methylphenidate
`concentration and efficacy measures was used to assess
`tolerance.38-40 This model used methylphenidate clear-
`ance values reported in the literature as the basis for
`simulating individual concentration–time profiles for
`the various regimens. The validity of this approach was
`confirmed in a separate pharmacokinetic study of tid
`and ascending regimens administered to 21 adult vol-
`unteers (14 men and seven women): the shape of the
`pharmacokinetic curve in the ascending treatment
`(defined by the multiple small doses) matched the pre-
`dicted shape based on simulation, and the within-
`subject pharmacokinetic variability was low (<10%).
`For this study, modeling was conducted with use of
`the nonlinear mixed-effects approach (NONMEM)38
`and included all of the children’s data to estimate the
`mean and individual difference parameters. An Emax
`model was fitted to the simulated methylphenidate con-
`centrations and efficacy measures. The strengths and
`weaknesses of the Emax model have been discussed in
`
`Page 5
`
`

`

`300 Swanson et al
`
`CLINICAL PHARMACOLOGY & THERAPEUTICS
`SEPTEMBER 1999
`
`Fig 3. Study I: Peak and trough responses for an example efficacy measure (attention subscale of
`the SKAMP rating scale) for the bid, flat, ascending, and placebo treatments.
`
`detail elsewhere.38-40 The limitations imposed by the
`known weaknesses of the Emax model should be con-
`sidered when interpreting the analyses of this study.
`The following equation was used in this study:
`
`(1)
`
`50 + Ceh
`t)/(ECh
`Et = E0 + (Emax Ceh
`t)
`In this equation, Et is the efficacy score (SKAMP atten-
`tion or Actigraph activity) at time t, Emax is the maxi-
`mum efficacy score, and ECh
`50 is the concentration
`needed to reach the 50% efficacy score. Ceh
`t is the effect
`site concentration at time t, and h is the Hill coefficient.
`A tolerance model was fitted to the SKAMP assess-
`ment scores (equation 2). We used the tolerance model
`reported by Park et al40 in which the development of
`tolerance is modeled as a metabolite with antagonistic
`properties, as shown in the following formula:
`
`(2)
`
`Et = E0 + (S · Cet)/(1 + Cant/Cant50)
`In this equation, E0 is the baseline effect, S is the slope
`of the linear concentration–effect relationship, Cant is
`the antagonist concentration at time t, and Cant50 is the
`antagonist concentration that reduces the agonist effect
`by 50%. S is calculated as Emax/EC50 because a true
`Emax cannot be estimated. The following equations
`define Cant and Ce(t):
`
`Cant = f(ktol, Ce)
`
`(3)
`
`1
`Ce(t) = ^n
`
`i=1
`
`Dikakeo e–ka·ti
`———
`V ka(ke – ka) (keo – ka)
`
`(4)
`
`+
`
`Dikakeo e–ke·ti
`———
`V ke (ka – ke) (keo – ke)
`
`+
`
`Dikakeo e–keo·ti
`———
`V keo(ka – keo) (ke – keo)
`
`2
`
`In these equations, Di is the ith dose of the drug, ka is
`the apparent first-order absorption rate constant, keo is
`the distributional rate constant, V is the volume of dis-
`tribution of the drug, ke is the apparent first-order elim-
`ination rate constant, ti is the time since the ith dose,
`and t is the time since the first dose.
`These equations describe a one-compartment phar-
`macokinetic model for first-order absorption and dis-
`position of methylphenidate. As discussed in detail
`elsewhere,39,40 the pharmacokinetic and pharmacody-
`namic models are linked to relate the simulated plasma
`methylphenidate concentration to the observed efficacy
`measures. The link model relates the simulated plasma
`concentration (C) to the effect site concentration (Ce)
`through the rate constant (keo) that accounts for distri-
`butional disequilibrium. It is hypothesized that a non-
`competitive antagonist is produced with a first-order
`rate constant (ktol) that governs the rate of appearance
`and disappearance of the antagonist formed from Ce
`
`Page 6
`
`

`

`CLINICAL PHARMACOLOGY & THERAPEUTICS
`VOLUME 66, NUMBER 3
`
`Swanson et al 301
`
`B
`A
`Fig 4. Study II: Time–response (A) and concentration–effect plots (B) for the tid-AM treatment (n = 14).
`
`rather than C. The actual rate of tolerance development
`is the sum of keo and ktol.
`
`RESULTS
`Study I. Of the 38 children recruited for this trial, 34
`entered and 31 completed all 4 treatment days. All cap-
`sules were given within 90 seconds of the scheduled
`administration times, so the targeted dosing times were
`met that would generate the simulated plasma methyl-
`phenidate concentrations shown in Fig 1.
`The repeated-measures ANOVA revealed that the
`interaction of regimen and session was significant for
`multiple measures of behavior, attention, and cognitive
`performance. This interaction was significant for two of
`the three subscale scores from the CLAM rating scale
`(hyperactivity index [P < .0029] and inattention/overac-
`tivity index [P < .002]), for both subscales of the
`SKAMP rating scale (attention index [P < .0052] and
`deportment index [P < .0001]), and for the accuracy mea-
`sure from one of the conditions of the memory scanning
`task (memory load = 1 [P < .0095]). These significant
`regimen ·
`session interactions indicate that the specific
`time course of effects differed by dosing regimen. We
`will describe the time course differences for two of the
`efficacy measures to provide an example to the common
`pattern that held across the multiple measures.
`In Fig 3, the time course differences across condi-
`tions are shown for the deportment and attention mea-
`sures from the SKAMP rating scale.32-34 As expected
`
`from the extensive literature on effects of immediate-
`release methylphenidate,18,19 at all four assessment
`points these untransformed measures in the bid condi-
`tion were lower (reflecting greater efficacy) than in the
`placebo condition. Across all measures, the general pat-
`tern of effects of the two experimental conditions
`depended on the time of day: in the morning sessions
`the efficacy of the flat regimen was greater than the
`efficacy of the ascending regimen, but in the afternoon
`sessions the efficacy of the ascending regimen was
`greater than the efficacy of the flat regimen. At the
`afternoon peak (the most likely time for the occurrence
`of acute tolerance), paired comparisons of the mean
`scores in the bid and ascending regimens were not sig-
`nificant for any of the efficacy measures. However,
`paired comparisons of the mean scores in the bid or
`ascending regimens with the mean scores in the flat reg-
`imen revealed that the flat regimen was significantly
`less efficacious than the bid or ascending regimens for
`multiple measures of efficacy: the hyperactivity index
`(P < .001 and P = .037), the inattention/overactivity
`index (P < .001 and P = .044), the deportment index
`(P = .004 and P = .032), and attention index (P = .006
`and P = .0060).
`Effect sizes were calculated as the difference between
`the mean score in the placebo regimen and the mean
`score in the treatment conditions, divided by the stan-
`dard deviation for the placebo scores. For the dependent
`variables described above, the average effect size for the
`
`Page 7
`
`

`

`302 Swanson et al
`
`CLINICAL PHARMACOLOGY & THERAPEUTICS
`SEPTEMBER 1999
`
`B
`A
`Fig 5. Study II: Time-response (A) and concentration–effect plots (B) for the tid-PM treatment (n = 16).
`
`bid condition was about 1.0, which is in the range of the
`expected magnitude of effects of immediate-release
`methylphenidate and is statistically significant for the
`sample size of this study. In the afternoon (sessions 3
`and 4), the mean effect size for the ascending treatment
`was 97.9% of the mean effect size for bid treatment, and
`the mean effect size for the flat treatment was only
`59.1% of the effect size for the bid treatment.
`These results indicate that an ascending methyl-
`phenidate delivery profile can produce about the same
`magnitude of efficacy as a bid regimen in the afternoon,
`which suggests that a large bolus of methylphenidate
`is not required to elicit full clinical effects. These results
`also indicate that the flat regimen was significantly less
`efficacious than the bid regimen in the afternoon, which
`suggests that acute tolerance may be developing dur-
`ing the day to sustained concentrations of methyl-
`phenidate.
`Study II. Of the 32 children recruited and enrolled
`in this trial, 30 children completed all four treatments
`(ie, placebo, tid, and ascending conditions plus either
`the tid-AM or the tid-PM condition). Simulated plasma
`methylphenidate concentration–time curves for the
`standard tid and two experimental tid regimens (tid-AM
`and tid-PM) are shown in Fig 2, based on a 30-mg total
`daily dose. Points from these curves were paired with the
`efficacy measures selected for this study (attention and
`activity) and were used to define the pharmacokinetic-
`pharmacodynamic relationship. Example graphs of the
`
`concentration–effect relationship between plasma con-
`centration and the attention efficacy measure are pre-
`sented in Figs 4 and 5. These figures show the data for
`different cohorts of subjects randomly assigned to the
`two parallel conditions, so same overall differences
`across Figs 4 and 5 may be due to between-subject
`comparisons. Each data point reflects the mean
`placebo-corrected value for the efficacy measure at the
`simulated plasma methylphenidate concentration for
`the session. In the left panels of Figs 4 and 5, the
`abscissa is defined by time (7:30 AM to 7:30 PM) and
`the ordinates are defined by the pharmacodynamic
`(attention) and pharmacokinetic (simulated methyl-
`phenidate concentration) scales. In the right panels of
`Figs 4 and 5, the same data are presented as concentra-
`tion–effect plots, with the pharmacokinetic measure on
`the abscissa and the pharmacodynamic measure on the
`ordinate, and time coded by arrows to show the pro-
`gression from session 1 to 10 across the day. As
`described earlier, for these placebo-adjusted and trans-
`formed-efficacy measures, increases (positive values)
`are considered to reflect improvement and decreases
`are considered to reflect deterioration.
`In all conditions, by 1 hour after the morning bolus
`(ie, by session 2), large acute increases in efficacy were
`observed. However, the magnitude of the increase in
`the efficacy measures was not the same in the two con-
`ditions after the second bolus. In the tid-AM condition,
`only small increases in efficacy were observed after the
`
`Page 8
`
`

`

`CLINICAL PHARMACOLOGY & THERAPEUTICS
`VOLUME 66, NUMBER 3
`
`Swanson et al 303
`
`second bolus administered at 9:30 AM (Fig 4; sessions
`3, 4, and 5). In the tid-PM condition, large increases in
`efficacy were observed after the second bolus at 1:30
`PM (Fig 5; sessions 6, 7, and 8). Differences were also
`apparent after the third bolus administered at 3:30 PM
`in both conditions. The third bolus produced large
`increases in efficacy for the tid-AM regimen (Fig 4; ses-
`sions 8, 9, and 10), but only small increases were
`observed for the tid-PM regimen (Fig 5; sessions 9 and
`10). Although these patterns are consistent with multi-
`ple interpretations, they are highly consistent with the
`tolerance hypothesis, which predicts smaller effects at
`a given concentration after recent exposure to high
`methylphenidate concentrations.39,40
`For an evaluation of tolerance, the data were dis-
`played in concentration–effect plots, which are pre-
`sented in the right panels of Figs 4 and 5. If the rela-
`tionship between plasma concentration and efficacy
`were constant, then the points in these plots would fall
`on a straight line. As shown in the right panels of Figs
`4 and 5, this pattern did not emerge. Instead, both the
`tid-AM and tid-PM conditions showed counterclock-
`wise hysteresis loops in these concentration–effect
`plots. This supports the existence of acute tolerance
`(tachyphylaxis).
`To evaluate this further, a pharmacokinetic-pharma-
`codynamic model was used that was a simple extension
`of the Emax model with additional parameters to
`account for tolerance39,40 (equation 2). The estimated
`parameters and coefficients of variation were reason-
`able, and a c 2 test indicated that the tolerance model
`provided a better fit than the Emax model. Compared
`with the Emax model, the tolerance model reduced the
`residual sums of squares by almost 50% (eg, a reduc-
`tion of 557 units on the attention subscale of the
`SKAMP). This reduction (due to the three additional
`parameters of the tolerance model) was well above the
`eight- or nine-unit reduction per additional parameter
`required for statistical significance. The effect site con-
`centration and the effect (assuming that tolerance had
`not developed) were calculated and plotted with use of
`the estimated parameters of the tolerance model. The
`inclusion of the tolerance parameters significantly
`decreased the area