We searched PubMed, Embase, and Cochrane databases, using terms such as “SUDEP,” “AED,” and “ASM.” The final search was conducted in February 2024. This search yielded 2917 articles, of which we only considered studies with SUDEP as an independent outcome and a control group as relevant. We present the ASM studies per prescribing scenario (starting, adding or discontinuing an ASM and using specific ASMs) and separately discuss the SUDEP studies evaluating other scenarios, including drugs other than ASMs and SDDs.

We found two retrospective studies evaluating the modulatory effect of starting ASM on SUDEP beyond TCS control (Table 1).[26, 27] One combined analysis found a lower SUDEP risk for those using monotherapy [odds ratio (OR) 0.5, 95% confidence interval (CI) 0.3–0.995] compared to those without ASM use.[26] A more recent and extensive study did not find a significant difference in SUDEP risk comparing no ASMs with monotherapy.[27] It is therefore unlikely that ASM monotherapy lowers SUDEP risk beyond the effect of TCS control.

Adding an extra ASM has an apparent protective effect on SUDEP in studies that did not control for seizure burden. Pooled data from 112 double-blind, placebo-controlled randomised trials investigating add-on ASMs in adults with refractory focal or generalised epilepsy showed that SUDEP incidence was seven times lower in those receiving an extra ASM compared with the placebo group.[28]

Studies that did not correct for TCS frequency have reported a higher SUDEP risk for those using multiple ASMs versus those on monotherapy or no ASMs. We identified several retrospective controlled studies evaluating the effect of ASM poly- versus monotherapy on SUDEP occurrence whilst correcting for TCS frequency (Table 1).[24,25,26,27] Only one case–control found an increased risk for SUDEP when comparing more than two ASMs with less than two ASMs. This effect did not remain in a combined analysis, including two other retrospective studies.[26] This would suggest that the initially reported higher SUDEP risk in those taking multiple ASMs reflected TCS severity.

A significant limitation of these studies is using medical records to determine the ASM regimen. This may not necessarily correspond to the prescription at the time of death and, more importantly, may not resemble the actual intake of the prescribed ASMs. A large-scale Swedish study addressed this concern by using dispensed prescription data from a national register, thus allowing the estimation of ASM usage over time.[27] While dispensing may be a closer estimation of actual ASM usage, it still serves as a proxy as it remains unknown whether the dispensed drugs were actually taken. The Swedish study determined ASM usage by the number of ASMs dispensed within 90 days of death or index date. This study found a lower SUDEP risk among those taking polytherapy (more than one ASM) when accounting for TCS frequency compared to no ASMs (OR 0.48, 95% CI 0.26-0.90). When further specifying polytherapy, risk reduction occurred when three or more ASMs were used (OR 0.31, 95% CI 0.14–0.67).[27] This would suggest a protective effect for polytherapy beyond TCS control.

We found no studies that evaluated the effect of ASM discontinuation on SUDEP rates.

We identified 17 papers that evaluated specific ASMs[25,26,27, 29,30,31,32,33,34,35,36,37,38,39,40,41,42], including four that controlled for TCS frequency (summarized in Table 2). Most ASMs had no effect on SUDEP occurrence, but two sodium-modulating ASMs (NaM–ASM) appeared to aggravate SUDEP risk: carbamazepine (CBZ) and lamotrigine (LTG). Most case–control studies did not investigate CBZ, LTG and SUDEP risk beyond TCS control.[30,31,32,33,34,35,36,37,38,39,40,41,42] We identified three studies that evaluated CBZ use accounting for TCS frequency, of which only one reported an increased risk of SUDEP occurrence (Table 2).[25] It was not specified whether CBZ was combined with other ASMs. A combined analysis selecting only users of CBZ as monotherapy, including the study that initially suggested an augmented risk, found no elevated SUDEP risk.[26] Similarly, the Swedish case-control study that evaluated prescription data over the past 90 days also reported no elevated risk for SUDEP when comparing CBZ monotherapy to no treatment.[27]

We found three studies on the association between lamotrigine (LTG) and SUDEP that corrected for TCS frequency.[26, 27, 29] All studies excluded LTG polytherapy users. No effect on SUDEP occurrence was found. These studies include the combined analysis of the Swedish case–control study, which used prescription data over the past 90 days.[26, 27] The other study was a large-scale study (101 SUDEP cases), which also found no effect for other sodium channel ASMs compared with commonly used non-NaM-ASMs (mostly levetiracetam, valproate and topiramate).[29] This study established LTG and other ASM use at admission in their Epilepsy Monitoring Unit (EMU) and did not validate usage at the end of follow-up, which could last up to 16 years. This could mean that LTG exposure might have been absent during follow-up. A 2021 US Food and Drug Administration warning revived the safety concerns for LTG use. The warning was issued on the basis of vitro data suggesting that LTG exhibits class IB antiarrhythmic activity. A rapid systematic review found that there is not enough evidence to support that LTG is associated with SCD or SUDEP or ECG changes in people with or without epilepsy when compared to other ASMs.[43] This was supported by a pharmacovigilance study that analysed adverse events associated with LTG reported to the FDA for people with epilepsy.[43]

We found one study describing a preventive effect for a specific ASM when correcting for TCS frequency with a limited number of users (n = 28). The Swedish drug prescription register study reported a protective effect on SUDEP (OR 0.10, 95% CI 0.02–0.61) for levetiracetam (LVT), monotherapy when compared with no ASM use.[27]

In the past, post-mortem studies have shown that SUDEP cases often show ASM levels at a sub-therapeutic level, which may reflect non-adherence.[44,45,46] These studies were, however, uncontrolled. We identified 14 controlled studies investigating the association between adherence and the occurrence of SUDEP (Online resource 1). Several methods for estimating non-adherence were used, all with limitations that may explain the conflicting results.

Most studies found no association between non-adherence and SUDEP occurrence. [25, 32, 40, 47, 48, 50,51,52,53] Some defined non-compliance as sub-therapeutic ASM levels at the last registered visit or after death.[32, 47, 48] The levels at the last visit may not accurately reflect compliance at the time of death, while redistribution may bias post-mortem results. This effect may diverge through varying post-mortem intervals.[55] Other studies determined adherence through monthly monitoring to count remaining pills, where more than two missed doses in the previous month were considered as non-adherence.[50, 51] Other methods used to define non-compliance were stated by treating physicians and family members or collected from medical records.[25, 32, 40, 52, 53]

We found four studies reporting an elevated risk of SUDEP in non-adherence.[27, 38, 49, 54] Two studies assessed variations in ASM hair strand levels, with a more significant variability indicating greater non-adherence.[38, 49] Adherence can be measured over multiple months, assuming hair grows approximately 1 cm a month. The Swedish case–control study used two methods to determine non-adherence while controlling for TCS frequency.[27] Firstly, adherence was based on documentation in medical records, which showed a significant risk compared to when no non-adherence was mentioned (OR 2.75, 95% CI 1.58–4.78). Documentation of adherence in the records was checked for 5 years before the date of death or index date, which, for controls, corresponded to the demise of cases. The second method that established non-adherence was with the prescription database, for which 91–180 days or 181–365 days since the last dispensed ASM were used as cut-offs. No association with SUDEP was found with a similar OR for 181–365 days (OR 2.96, 95% CI 0.46–18.89) compared with the previous method. With this latter method, it is, however, unknown whether ASM was stockpiled. The fourth study evaluated adherence patterns, using ASM prescription and dispensing data from the Australian Pharmaceutical Benefits Scheme (PBS) to calculate the daily polypharmacy possession ratio (DPPR), a measure for the proportion of prescribed pills available for a person each day.[54] This method can account for adherence to multiple drugs and stockpiling. Cluster analysis defined four patterns of adherence to the trajectory of three years: good, declining, poor and very poor. Declining adherence over time was associated with an increased risk for SUDEP.[54] Both hair strand levels and prescription databases allow for the evaluation of adherence over a more extended period, which might be more reliable than information from patient records or statements from the family. The Swedish study used medical records to determine compliance, but in contrast to prior studies, it applied a five-year observation period.

We found three studies investigating the effect of the number of dose changes (Online Resource 2).[31, 38, 48] Studies showed conflicting results as one found an increased risk for SUDEP with a higher number of changes (three to five) compared with few changes (zero to one), and others showed no or a near significant protective effect for frequent dose changes. TCS frequency was not controlled for, so changes might also reflect disease control in this case.

3.6 Medication other than ASMs

Only a few studies evaluated SUDEP risk for non-epilepsy concomitant drugs.[24, 27, 31, 32, 38, 56] We found six studies with a small number (30 or less) or an undefined number of users. They did not report an effect on SUDEP occurrence; the two small-scaled papers reported an association between SUDEP and anxiolytics but had conflicting results (Online Resource 3). One more extensive case–control study reported no significant effect for beta-blockers (n = 133), SSRIs (n = 114), neuroleptics (n = 93) or other antidepressants (n = 53), but lower risk among statin users (n = 87).[27] It is yet unclear whether statins may reduce SUDEP susceptibility through modulating the cascade or a mere effect of associated profiles with statin use (more frequent health visits or comorbidities associated with other mortality risks).

SDDs may alert caregivers to seizures that would have otherwise been unwitnessed. These devices can identify seizures by detecting autonomic alterations or movements caused by the seizure. We found no studies investigating the effect of SDDs on SUDEP occurrence. Prospective studies to assess the impact of these devices are challenging to set up due to the low incidence of SUDEP, thus resulting in lengthy study duration and ethical constraints in the case of an RCT. We only found studies with indirect proof that enhancing supervision could lower SUDEP risk.[3, 4, 25, 38, 57] Most studies evaluated the protective effect of the presence of a roommate. One study comparing SUDEP rates between two residential care centres evaluated additional measures.[