The neurophysiological effects of single-dose theophylline in patients with chronic stroke: A double-blind,placebo-controlled,randomized cross-over study

2.1 Design

Subjects were studied in a cross-over design, on 2 separate days separated by at least one week (>21 theophylline half-lives). Subjects were randomized to receive either 300 mg extended-release theophylline or placebo on the first day and the opposite on the second day. TMS neurophysiology and motor behavioral assessments occurred in morning and afternoon sessions on each day. Patients, caretakers, and investigators were blinded to the condition order. Subjects were tested in the Motor Performance Laboratory at Columbia University. They gave written informed consent to participate in this study, which was approved by the CUMC Institutional Review Board.

2.2 Subjects

Twenty subjects were enrolled, but two subjects dropped out midway due to an unrelated medical or transportation issue. The results of 18 subjects are thus reported. Subjects were included if they were ≥40 years old and had a single ischemic or hemorrhagic stroke with resulting hemiparesis ≥6 months previously. Brain MRI or CT radiology reports were reviewed to confirm the presence of an infarction. Because we sought to examine bilateral neurophysiology, subjects who could at least minimally abduct their paretic index finger (MRC ≥1) and who had a recordable TMS-evoked response were included. Subjects were excluded for: multiple stroke events or bilateral paresis; history of seizure, preexisting CNS pathology, major psychiatric disorder, active substance abuse, respiratory disease, congestive heart failure, or peptic ulcer disease; inability to comprehend study requirements due to language abnormalities; thoracic or intracranial metal objects, implants, or devices, except for dental work; active use of any pharmacological agents metabolized through the cytochrome P450 1A2 pathway; or active participation in other trials.

2.3 Medication preparation and administration

The medications were prepared and dispensed by the Columbia University Medical Center Research Pharmacy. Patients received a single 300 mg dose of extended-release theophylline (Heritage Pharmaceuticals, NDC# 23155-062-01) or placebo, identical to theophylline in appearance and packaging. A single dose, rather than a week-long course, was chosen for this initial investigation because of the lack of experience with this medication in patients with stroke. Because many patients with chronic stroke are intracortically disinhibited at baseline (Schambra et al., 2015), we were initially concerned that longer drug exposures could overshoot the intended effect.

The standard dose of theophylline for asthma monotherapy is 300 mg twice daily, with a target serum level of 10–20 μg/ml. In its extended-release formulation, a mean peak serum theophylline concentration of ∼4.4 μg/ml is expected 4–8 h after intake (Heritage, 2014). Serologic theophylline levels were not drawn.

Subjects were not required to fast, but were instructed not to take any form of caffeine on testing days. After capsule intake in the morning session, neurophysiological and clinical measures were immediately assessed. These were repeated 5 hours later in the afternoon session. Between sessions, subjects quietly read or napped.

2.4 TMS set-up

Patients were comfortably seated with the forearm resting on a pillow in ∼90° elbow flexion. Frameless stereotaxic equipment (Brainsight, Rogue Research, Canada), used to co-register the subject’s scalp positions with a phantom MRI brain image, ensured stimulation accuracy during and across sessions. Co-registration errors to the phantom’s surface landmarks were matched to ≤3 mm at each session.

Surface EMG was obtained from bilateral first dorsal interosseous (FDI) muscles, with electrodes taped in a belly-tendon orientation (SX230-100 and K800; Biometrics Ltd, UK). Electrodes were outlined with permanent ink on the skin to ensure consistency of electrode placement within day. The EMG signal was sampled at 1000 Hz, amplified 1000x, band-pass filtered at 15–450 Hz, and saved for offline analysis. All TMS assessments were taken at rest, and EMG activity was monitored online to ensure muscle relaxation.

TMS always began in the nonlesioned hemisphere. Stimuli were delivered in BiStim mode to the cortical hand representation of the motor cortex using Magstim BiStim². A 70-mm figure-of-eight remote control coil (Magstim Company Ltd, UK) was used for most measures, with the additionof a 50-mm figure-of-eight coil (Magstim Company Ltd, UK) for the measurement of interhemispheric inhibition. Pulses were generated using specialized software (Signal; Cambridge Electronic Devices, UK) and a 1401 microprocessor (Cambridge Electronic Devices, UK). The TMS coil was held tangentially to the skull with the coil handle pointed 45° posterior-laterally. The scalp location (“hotspot”) producing the largest amplitude motor evoked potential (MEP) for the contralateral first dorsal interosseous (FDI) muscle was identified and marked virtually on the phantom MRI.

2.5 TMS outcomes

2.6 Safety assessments

2.7 Blinding assessment

Following the second day of testing, patients and the main assessor (HMS) guessed which days they believed theophylline and placebo were administered. Reasons for their speculation were recorded.

2.8 Data analysis

The average change in neurophysiological and behavioral outcomes between morning and afternoon sessions was calculated for each condition. A paired nonparametric regression model was used to estimate differences in change scores associated with theophylline relative to placebo. With this hierarchical modeling approach, we accounted for repeated measurements within each individual. Considering that hemispheres may differentially respond to an intervention, we assessed outcomes in lesioned and nonlesioned hemispheres separately.

To evaluate differences in ordinal data, a non-parametric Wilcoxon Rank Sum was used. To evaluate differences in rate of categorical side effects, Fisher’s Exact Test was used. Inter-rater reliability of guessing with assessed with a kappa statistic. To examine the possibility of a ceiling effect in neurophysiological outcomes, correlations between the baseline (morning) value and the delta between morning and afternoon sessions (PM minus AM) were tested using linear regression analyses. Bonferroni adjustments were made for the assessment of multiple outcome measures. Significance was set at an alpha of 0.05.

2.9 Power calculation

We assumed that theophylline in patients with stroke would decrease SICI by 0.21 decimal percentage points with a standard deviation of 0.23 points, based on previously published results in normal subjects (Nardone et al., 2004). Further assuming zero correlation of the two changes within-subject, the paired regression model with n = 18 and 5% Type I error rate had 78% power to detect this magnitude of change in SICI.

3.1 Clinical characteristics

Age, gender, handedness, stroke type, lesion location, lesioned hemisphere, and time since stroke are given in Table 1. Subjects were tested at a median interval of 7.5 days (range 7–38 days) between the placebo and theophylline administrations. No subjects had active tobacco use.

3.2 TMS outcomes

For each condition and session, neurophysiological outcomes are shown for the lesioned and nonlesioned hemispheres (Table 2). For both hemispheres, change scores were not significantly different between theophylline and placebo conditions for RMT (Fig. 1), SICI_1ms (Fig. 2), SICI_2ms (Fig. 3), LICI (Fig. 4), or IHI (Fig. 5). Subjects with marked disinhibition behaved consistently across days, and removal of their data did not substantively alter comparisons between the conditions.

To evaluate the possibility of reporting a Type II error, we computed the smallest detectable group change (SDC_group) for two TMS outcome measures, SICI_2ms and LICI, using repeated-measures reliability estimates previously attained (Schambra et al., 2015). The SDC_group is the threshold amount of change necessary to be considered a true change for a group (Beckerman et al., 2001). For a group of 18 patients with chronic stroke, the SDC_group of SICI_2ms is 0.14 and 0.08 for the lesioned and nonlesioned hemisphere, respectively. The SDC_group of LICI is 1.2 and 1.0 for the lesioned and nonlesioned hemisphere, respectively. Our changes in SICI_2ms and LICI did not exceed these SDCs_group, signifying that the observed deltas occurred within the envelope of measurement noise.

To evaluate for the presence of a ceiling effect in the neurophysiological outcomes, we assessed the correlation between the normalized baseline (morning) value and the delta between morning and afternoon sessions (Table 3). We examined correlations within condition and hemisphere, and corrected for the performance of multiple correlations. We found moderate-to-strong inverse relationships between baseline values and deltas for SICI_1ms, SICI_2ms, LICI, and IHI, with most being highly significant. These correlations showed that higher baseline values were associated with negative deltas, and lower baseline values were associated with positive deltas (see correlations for SICI_2ms in Fig. 6 for example). In other words, less inhibition at baseline was associated with a strengthening of inhibition in that outcome, whereas more inhibition at baseline was associated with a weaknening of inhibition. This relationship was present in both the theophylline and placebo conditions. The theophylline condition did not change the regression line intercept or slope. There was no significant correlation between baseline RMT and its change in either hemisphere orcondition.

3.3 Behavioral outcomes

For the paretic and nonparetic hands, change scores were not significantly different between theophylline and placebo conditions for strength, pegboard speed, or pegboard accuracy (Fig. 7).

3.4 Cardiovascular and psychometric outcomes

Average heart rate and blood pressure, and change from morning to afternoon sessions, did not differ between conditions (Table 2). Likewise, there were no significant differences in average alertness and excitement, change in alertness and excitement, or total amount of exercise and sleep under theophylline or placebo. The frequency of side effects was low and not significantly different across conditions: 1 headache (placebo condition), 1 episode of jitteriness (theophylline condition), and no nausea or seizures.

3.5 Randomization and blinding

Eight (44%) of the first sessions were placebo and 10 (56%) were theophylline. Correct guesses about the administration order occurred for 56% of subject responses and 61% of investigator (HMS) responses, and the subjects and investigator showed low agreement in their guesses (K = 0.33, NS). Patients reported feeling more alert or more physically adept as often on placebo as on theophylline.

4.1 Neurophysiological effects of single-dose theophylline

Our results are in keeping with a lack of cortical excitability changes found with a single dose of caffeine, another methylxanthine drug with similar pharmacological properties (Orth, Amann, Ratnaraj, Patsalos, & Rothwell, 2005). However, they do not replicate those found in healthy controls receiving theophylline, which resulted in a reduction in SICI_2ms relative to placebo (Nardone et al., 2004). There are two reasons why we may have not have observed this effect: methodological differences, or a differential responsiveness of post-stroke neural circuitry.

Our approach differed from that of Nardone and colleagues in two ways. First, Nardone and colleagues administered theophylline daily for 7 days, whereas we administered theophylline in a single dose. Because theophylline had not been investigated before in patients with stroke, we chose a conservative dosing strategy to avoid inducing pathological hyperexcitability in patients who are already disinhibited. However, this single dose resulted in a concentration that was likely too low to modulate neurotransmission: we probed neurophysiology when peak serological concentration was estimated to be ∼4 μg/ml, whereas Nardone and colleagues report an effect at 8–21 μg/ml. In addition, a single dose may have resulted in a more variable cortical concentration of theophylline, as the steady state of a drug can be expected after ∼5 doses. Thus, a repeated exposure to theophylline may be required to allow for equilibrium across the blood brain barrier, a sufficiently high and constant cortical concentration, and receptor modulation. Second, we based our CS stimulation intensity on 80% RMT, whereas Nardone and colleages used 95% AMT, generating stimulation intensities that equate to 76–81% RMT. However, it is not expected that incremental differences around this CS stimulation intensity would contribute to large differences in inhibition (Kujirai et al.,1993).

We also studied theophylline in patients with stroke, whereas Nardone and colleagues studied healthy subjects. Reduced SICI after stroke, particularly in lesioned hemisphere, is widely observed and may be related to diminished GABA_AR expression (M. Qu et al., 1998; Que et al., 1999). In the present sample, SICI_2ms in the majority of subjects was more disinhibited at baseline than in the previously sampled healthy controls (Nardone et al., 2004). In these subjects, SICI_2ms is nearing the upper limit of disinhibition and thus may be at the upper end of its physiologically meaningful range. The addition of an excitatory/disinhibitory perturbation in conjunction with SICI disinhibition may therefore produce nonadditive results. First, responsiveness to further excitatory/disinhibitory interventions may be blunted in patients with stroke, resulting in no substantive change, as observed after motor training (Blicher et al., 2009). Alternatively, disinhibition may be down-regulated through homeostatic metaplasticity mechanisms, returning circuitry to a more physiologically responsive range (Bienenstock, Cooper, & Munro, 1982). This phenomenon has been observed with continuous theta-burst stimulation (Huang et al., 2005; Huang et al., 2008) and anodal transcranial direct current stimulation (Heise et al., 2014).

In our correlation analyses of the neurophysiologic outcomes, we did not find an asymptote centered on zero change, which would suggest a ceiling over some level of baseline disinhibition. We did, however, find strong inverse relationships between the amount of baseline inhibition and its change between morning and afternoon sessions, in both conditions and hemispheres. Subjects with strong baseline inhibition became less inhibited, those with normal inhibition or mild disinhibition had minimal change, and those with strong baseline disinhibition became more inhibited. This observation could be considered a regression toward the mean, where observations that are randomly extreme on the first measurement tend to be closer to the mean on the second. However, the extreme measurements were not random, but rather were consistent within subjects across days, making this phenomenon less likely. This observation therefore would suggest the action of a homeostatic process. Theophylline did not flatten the slopes of these relationships, but also did not increase their intercepts (i.e. right-shift the relationships). These findings suggest that theophylline neither modifies nor usurps this homeostatic process, at least in a single-dose administration.

Of note, we applied a conservative statistical correction for our exploration of multiple outcome measures. Controlling for Type I error when multiple outcomes are evaluated is a practice implemented in other clinical neuroscience methodologies (Bennett, Wolford, & Miller, 2009; Mensen & Khatami, 2013), but uncommonly adopted in multiple-outcome TMS studies. Although it is conceivable that we are reporting Type II errors (false negatives) for our TMS outcomes, the within-condition change scores of SICI_2ms and LICI are less than the smallest detectable change expected for their hemisphere and group size. This means that differences between conditions in change scores likely occurred within the envelope of measurement noise (Schambra et al., 2015).

4.2 Safety of single-dose theophylline

We found that theophylline was well-tolerated, with no adverse neurologic or cardiovascular effects. For both upper extremities, theophylline did not worsen finger strength or manual dexterity relative to placebo. These findings imply that, at least for a single dose, theophylline does not interfere with motor performance, allowing the subjects to safely proceed with assessments of motor skill learning.

Unlike a single assay of motor performance, motor skill learning models the activity-dependent plasticity brought about through physical therapy, which engages the robust plasticity milieu triggered by stroke (Zeiler & Krakauer, 2013). While the neurophysiologic effects of any pharmacological intervention are important to establish, justification for its translation to recovering patients requires a beneficial effect on motor skill learning. Given the present study was a single-day drug exposure with a cross-over design, we were limited in our ability to assess motor skill learning, a direction for future efforts.

4.3 Theophylline and GABA_AR

There is support for the hypothesis that theophylline can modulate neural circuitry to support recovery after stroke. Because Nardone and colleagues found an isolated reduction in SICI_2ms after theophylline, antagonism of GABA_AR by theophylline was inferred (Nardone et al., 2004; Ziemann et al., 2015). However, the neurophysiological underpinnings of SICI are complex, with contributions by glutaminergic neurotransmission and modulation by dopaminergic, serotonergic, adrenergic, and acetylcholinergic systems (reviewed in (Paulus et al., 2008). A decrease in SICI may therefore reflect direct GABA_AR antagonism, increased glutamatergic neurotransmission, or modulation of the motor cortex by other neurotransmitters.

Theophylline weakly binds to GABA_AR at concentrations within the clinical dosing range (Segev, 1988), but electrophysiological disinhibition only occurs at concentrations far exceeding clinically toxic levels (Scholfield, 1980, 1982). Theophylline is ∼100 times more potent at the adenosine than GABA_A receptor and nonselectively antagonizes the A₁ and A_2a subtypes (Fredholm, 1979; Snyder et al., 1981). A₁R blockade increases presynaptic neurotransmitter release without affecting basal GABA release, thus increasing post-synaptic excitability (Burke & Nadler, 1988; Dunwiddie & Diao, 1994; Hollins & Stone, 1980; Prince & Stevens, 1992). A_2aR antagonism, on the other hand, generally reduces post-synpatic transmission across multiple neurotransmitter systems (reviewed in (Chen & Chern, 2011).

It is thus conceivable that reduced SICI by theophylline could reflect A₁R antagonism, resulting in increased glutamatergic, adrenergic, serotoninergic, and acetylcholinergic signaling in/to the motor cortex. Similarly, SICI reduction may reflect A_2aR antagonism, resulting in decreased dopamine-mediated action on the motor cortex. Given theophylline’s weak affinity for GABA_AR at low concentrations, GABA_AR antagonism is less likely to explain SICI reduction. Increasing glutamatergic, adrenergic, and serotoninergic signaling in the motor cortex are approaches currently being explored in preclinical and clinical stroke recovery studies (Cherry, Lenze, & Lang, 2014; Dhawan et al., 2011; Grade et al., 1998; Siepmann et al., 2015).

4.4 Limitations

In the present sample of patients with chronic stroke, we note measurement variability that challenges the detection of group change, irrespective of condition. We were sufficiently powered to detect a significant effect of theophylline on SICI_2ms, had the previously reported effect size and variability been borne out in this sample. In our sample, SICI_2ms and LICI measurements were more variable than those previously observed in a group of patients with chronic stroke, despite phenotypic similarity, identical methodology, the same TMS operator, and several shared subjects (Schambra et al., 2015). It is possible that the difference lies in the number of trials averaged— 15 in the present study versus 40 previously. In the future, increasing the number of trials may help to reduce variability, especially in patients with stroke. In addition, adjusting individual stimulation intensities at baseline may help to standardize the degree of inhibition across individuals, which can then be assessed in the face of an intervention (Florian, Muller-Dahlhaus, Liu, & Ziemann, 2008). We also did not obtain serological levels of theophylline, and therefore can only report that neurophysiological changes and adverse effects did not occur at a single 300 mg dose of theophylline, rather than at a certain serological concentration. Particularly if clinical translation is intended, future investigations may consider delimiting the lower and upper boundaries of drug level that induce a neurophysiological or behavioral learning effect. Finally, we did not evaluate excitatory TMS outcomes given a lack of theophylline effect in healthy subjects (Nardone et al., 2004), although future studies may to wish to investigate differential responses in post-stroke neural circuitry.