Risk of Major Cardiovascular and Cerebrovascular Events in Users of Lisdexamfetamine and Other Medications for Attention-Deficit/Hyperactivity Disorder in Denmark and Sweden: A Population-Based Cohort Study

The clinical importance of this cohort study among adult new users of LDX and remote users of other ADHD medications is supported by the data. Long-term use of LDX in this study was fairly prevalent, with 35–40% of patients in the LDX cohort having current LDX use of longer than 12 months. We observed little to no increased risk of MACE among patients treated with LDX compared with patients previously treated with other ADHD medications (i.e. amphetamine, dexamphetamine, methylphenidate or atomoxetine). The adjusted IR of MACE in LDX users was 1.62 and 2.08 per 1000 persons-years in Denmark and Sweden, respectively. A weak association between the use of LDX and the risk of MACE, compared with previous users of other ADHD medications who were no longer treated with these medications, was observed, with adjusted IRRs (95% CI) of 1.01 (0.48–2.13) in Denmark, 1.13 (0.75–1.71) in Sweden and 1.10 (0.77–1.58) for the pooled population. No clinically meaningful association among LDX users compared with previous users was observed for any of the secondary outcomes. Results were generally consistent in exploratory and sensitivity analyses.

The prevalence of cardiovascular disease in adults with ADHD, regardless of whether and how they have been treated, has not been extensively researched [27]. Furthermore, evidence from previous evaluations of the cardiovascular and cerebrovascular risks associated with ADHD medications, mostly amphetamines (other than LDX) and methylphenidate, has been inconclusive and somewhat limited by potential unmeasured confounding and selection bias [9‐11, 14]. The results of this study are in accordance with several studies investigating other cardiovascular and cerebrovascular outcomes among patients treated with stimulants indicated for ADHD. The IRs for MACE observed among current LDX users in our study (1.62 and 2.08 per 1000 person-years in Denmark and Sweden, respectively) are similar to IRs reported in two studies conducted in the USA. In a study conducted among US adults (aged 18 years or older) using claims data between 1999 and 2006, Schelleman et al. [9] reported that the IR for a composite score of stroke and MI among new users of amphetamine was 2.62 per 1000 person-years, while the IR for age- and sex-matched non-users of ADHD medications was 3.03. The resulting adjusted hazard ratio (HR) was 0.78 (95% CI 0.51–1.19), and the authors concluded that no increased risk of serious cardiovascular events was observed in patients initiating treatment with amphetamines compared with patients not initiating an ADHD medication. In a study including more than 440,000 adults from four US study populations (1986–2005), Habel et al. [14] reported rates of serious cardiovascular and cerebrovascular events (including MI, SCD or stroke) of 1.88 per 1000 person-years among current users of amphetamines (other than LDX), 2.51 per 1000 person-years among current users of methylphenidate and 3.60 per 1000 person-years among previous users (patients with more than 364 days since the end of the last days’ supply). The authors of the latter study concluded that current or new use of ADHD medications was not associated with an increased risk of serious cardiovascular events: the adjusted risk ratios for serious cardiovascular and cerebrovascular events were 1.03 (95% CI 0.86–1.24) when comparing current users of methylphenidate, amphetamine or atomoxetine with previous users and 0.83 (95% CI 0.72–0.96) when comparing current users with non-users (i.e. a general population of the same age and sex with no current and no previous use).

Another US study by Holick et al. [10] showed an increased risk (HR 3.44; 95% CI 1.13–10.60) of transient ischaemic attack but not of cerebrovascular accident (HR 0.71; 95% CI 0.34–1.47) among patients treated with atomoxetine or ADHD stimulant medication compared with the general population. However, in this secondary analysis, the authors used an intention-to-treat approach and did not match the cohorts by propensity score as they did in their primary analysis, where they found no association [10]. Another study by Schelleman et al. [11] found an increased risk of sudden death or ventricular arrhythmia (adjusted HR 1.84; 95% CI 1.33–2.55) and all-cause death (adjusted HR 1.74; 95% CI 1.60–1.89) when comparing methylphenidate users with the general population. A more recent meta-analysis including three trials conducted in adults concluded that no difference was detected in the number of adverse cardiac events (including cerebrovascular accidents, MI, SCD and significant electrocardiogram changes) between the participants treated with methylphenidate and those treated with placebo [12]. Another recent meta-analysis evaluating use of methylphenidate both in adults and in children and adolescents (which included the studies by Habel et al. [14] and Schelleman et al. [11]) also demonstrated no increased risk of stroke or MI [7]. Another study investigated new-onset heart failure or cardiomyopathy among initiators of lisdexamfetamine, methylphenidate and atomoxetine [13]. The event rates (per 1000 person-years) of heart failure/cardiomyopathy by age categories among patients treated with amphetamines during the first 90 days of use were 1.61 (less than 22 years), 2.42 (22–44 years), 9.22 (45–64 years) and 47.11 (65 years or older). Longer duration of amphetamine use was not associated with an increased rate of cardiovascular or cerebrovascular events. Outcome rates were slightly higher for all age groups of at least 22 years for 90 days’ use of methylphenidate products than for 90 days’ use of amphetamine products [13]. Finally, a study of patients treated with methylphenidate who had had an MI in South Korea, Taiwan and Hong Kong isolated MI risk by time period—before, during and after methylphenidate exposure—relative to a baseline period of non-exposure [28]. The highest risk of MI occurred during the pre-exposure period, suggesting that there is not a causal relationship between methylphenidate and MI [28].

Some limitations of this study are acknowledged. By design, different inclusion criteria were applied to LDX users and previous users, who required a “washout” period of at least 180 days from the end of the previous use of any of the other ADHD medications (amphetamine, dexamphetamine, methylphenidate or atomoxetine), potentially leading to immortal-time bias. Nonetheless, in the analysis evaluating previous exposure, in which patients in the LDX cohort were subject to the same washout period as was applied for previous users, the results remained consistent with the primary findings, suggesting that immortal-time bias may be minimal. Further, some patients may not have taken the dispensed LDX as prescribed, and the exact dosing regimen of prescribed medication was unavailable in this study. Nevertheless, data on redeemed prescriptions is considered to more closely reflect actual use by patients than prescription data. We assumed the use of one LDX capsule per day regardless of the strength and therefore may have overestimated the duration of LDX use. While evaluation of duration of use was focused on long-term use, and short-term use (e.g. less than 3 or 6 months) was not evaluated, we do not anticipate that an increased risk of MACE for LDX would have been detected with short-term use. Because the IRR estimate for MACE was 1.10 (95% CI 0.77–1.58) in the overall analysis including short-term and long-term users and was 1.59 (0.86–2.95) in the analysis of long-term use, the IRR point estimate for short-term use (less than 6 months), in theory, would have been lower than 1.10. The precision of the IRRs for several subgroup analyses was low since they were based on a low number of events; in particular, the number of events in the subgroups aged 18–29 and 30–39 were small, and thus risk estimates were imprecise (i.e. had wide 95% CIs) and should be interpreted with caution. It also was not possible to support an exploratory analysis for users aged 65 years or older because of an insufficient number of users within those age groups. An additional limitation is possible unmeasured confounding, though a quantitative bias analysis showed that the main results were robust to unmeasured confounding, except in extreme scenarios, where an unmeasured confounder was strongly associated with MACE and much more prevalent in LDX users than in previous users of other ADHD medications. Finally, although the inclusion of a comparator group of previous users of other ADHD medications reduced the potential for selection bias or confounding due to a healthy-user effect among patients from the general population or with chronically treated ADHD [14], it cannot be discounted that patients in the previous-users cohort were those for whom the ADHD treatment did not work or who may have had fewer symptoms, leading to confounding by indication by the comparison of untreated versus treated patients. However, the risk was considered minimal because the baseline characteristics of both cohorts were balanced using matching and propensity scores.