A mixed-methods study of patient perspectives on antiseizure medication discontinuation attempts after a period of seizure-freedom Authors: Samuel W Terman, Jordan M Silva, Max Kuster, Jasper Lee, Amanda Brand, Kara Manuel, Navya Kalia, Micaela Dugan, Marla Reid, Katherine Mortati, Alexandra Tolmasov, James F Burke, Chloé E Hill, Susanna S Okula, Palak S Patel Published: 2026 ABSTRACT AbstractBackgroundAntiseizure medications (ASMs) are the standard treatment for epilepsy. Yet because ASMs can have adverse effects and seizure risk decreases with longer seizure-free periods, guidelines have suggested considering ASM withdrawal after a period of seizure-freedom. However, work exploring patient experience with ASM discontinuation is scarce. We explored patient attitudes regarding past or possible future ASM discontinuation attempts.MethodsWe conducted a mixed-methods study of adults at least 1 year seizure-free, seen for epilepsy across three institutions. A semistructured interview script assessed views about past or hypothetical future ASM discontinuation attempts.ResultsWe interviewed 32 participants. Seizure-free duration ranged from 1.0 to 35.0 years (median 2.9, IQR 1.6–11.2). 21 (66%) participants reported having tried to discontinue an ASM in the past due to a period of seizure-freedom. 13 participants (43% of 30 who responded to the question) felt that they would still be glad to have tried discontinuation for the knowledge gained even if it resulted in a seizure. This corresponded to 52% (11/21) of patients who had previously attempted ASM discontinuation due to a period of seizure-freedom, compared with 22% (2/9) of patients who had not previously attempted ASM discontinuation (Fisher’s exact test p=0.23). Qualitative responses generally indicated that even a ‘failed’ discontinuation attempt was ‘worth trying’ and could support a patient’s future decision to remain on long-term ASMs without wondering any longer if they really ‘need it.’ConclusionsEven ‘failed’ discontinuation attempts may have value due to knowledge gained to support subsequent continuation. WHAT IS ALREADY KNOWN ON THIS TOPIC Guidelines suggest considering antiseizure medication withdrawal after a period of seizure-freedom for patients with epilepsy. Though limited research has evaluated patient attitudes about withdrawal. Guidelines suggest considering antiseizure medication withdrawal after a period of seizure-freedom for patients with epilepsy. Though limited research has evaluated patient attitudes about withdrawal. WHAT THIS STUDY ADDS 13 of 30 (43%) participants felt that they would still be glad to have tried discontinuation for the knowledge gained even if it resulted in a seizure. Qualitative responses generally indicated that even a ‘failed’ discontinuation attempt was ‘worth trying’ and could support a patient’s future decision to remain on long-term antiseizure medications without wondering any longer if they really ‘need it.’ 13 of 30 (43%) participants felt that they would still be glad to have tried discontinuation for the knowledge gained even if it resulted in a seizure. Qualitative responses generally indicated that even a ‘failed’ discontinuation attempt was ‘worth trying’ and could support a patient’s future decision to remain on long-term antiseizure medications without wondering any longer if they really ‘need it.’ HOW THIS STUDY MIGHT AFFECT RESEARCH, PRACTICE OR POLICY Even a ‘failed’ antiseizure medication discontinuation attempt may have value due to the patient gaining information about their medication indication. Even a ‘failed’ antiseizure medication discontinuation attempt may have value due to the patient gaining information about their medication indication. Introduction Seizure freedom is a primary goal in epilepsy treatment, as seizures may lead to injuries, jeopardise driving privileges and worsen quality of life.1 However, antiseizure medication (ASM) treatment can have downsides such as adverse effects2 that also reduce quality of life.1 Fortunately, two-thirds of patients with epilepsy experience a prolonged period of seizure-freedom on ASMs,3 up to 70% remain seizure-free even off ASMs4 and a post-withdrawal relapse does not worsen prognosis.5 Furthermore, deprescribing ASMs improves mood6 and cognition,7 and in existing data, no patients died of seizures or went into status epilepticus due to deprescribing.8 9 Accordingly, guidelines considering deprescribing.10 Prior work has focused on identifying patients at low seizure risk.11 12 However, little is known about attitudes towards deprescribing ASMs. The ideal scenario is for a patient to remain seizure-free after weaning. A patient who relapses after a weaning attempt could reflect on their attempt with regret, though alternatively they could still be glad they attempted weaning due to the information gained. One study13 found that attempting to withdraw ASMs was ‘worth a try’ to avoid lifelong ASMs, despite recurrence. However, that study was small (n=12), geographically narrow, from ~20 years ago. Since then, ASMs have changed,14 and they had almost no information from patients with ‘unsuccessful’ attempts. Additional investigation into how patients would feel about successful or unsuccessful withdrawal attempts would inform attitudes surrounding discontinuation. We interviewed patients or caretakers about past experiences with ASM discontinuation and potential future hypothetical ‘failed’ discontinuation attempts. Methods Participants and recruitment We included patients from three sites (John F Kennedy University Medical Center (JFK), Downstate State University of New York University Health Sciences University (SUNY) and University of Michigan (UM). We screened patients who had clinic visits at each site for a primary diagnosis of epilepsy within the last month, were ≥18 years old and were ≥1 year seizure-free. While past guidelines suggested considering ASM discontinuation after 2 years of seizure-freedom,15 16 we included patients who were ≥1 year seizure-free because these patients have stabilised and may soon be eligible for discontinuation, and restricting further would have reduced available samples. We made two telephone attempts, then email if no response.1719 Recruitment continued until thematic saturation, when new interviews appear to produce little useful information.2024 While this requires a subjective assessment, prior meta-analyses found that most concepts can be captured from 6 to 7 interviews in a homogeneous sample, and 30 interviews were highly sensitive with greater heterogeneity.2527 We also sampled to ensure representation across key demographics (age, race, seizure semiology, years seizure-free, brain surgery, surrogate responders). Semi-structured interview guide development and content Interview guide development occurred via repeated meetings by the research team to discuss core concepts and study objectives surrounding ASM discontinuation experiences. We then drafted content intended to address these core concepts. All members (including individuals with subspecialist board certification in epilepsy (SWT, KM, CEH, SSO, PSP) plus expertise in implementation science (CEH), neurological decision-making more generally (JAB) and qualitative methods (CEH, SSO) pretested the instrument, and questions were refined iteratively until all members agreed that questions were clear and met study objectives. (online supplemental appendix) Interviews were conducted virtually over Zoom at UM and JFK or over Teams at SUNY via screen sharing to display prompts, and a mixture of virtually and in-person at SUNY. Interviews were conducted by attending epileptologists at each site including SWT and CEH at UM and JFK and SSO at SUNY. All interviewers had previously conducted coursework in qualitative or mixed methods. Interviewers did not interview their own patients, other than two patients at SUNY who were under clinical care by the interviewer. Both audio and video were recorded for virtual sessions, whereas only audio was recorded for in-person sessions. Participants received $20 for the ~45 min interview. Results reported here represent one portion of our larger study.28 29 We asked participants whether they had previously attempted ASM discontinuation, and if so to reflect on that experience. We then asked if participants were ever to discontinue ASMs in the future due to seizure freedom, would they be glad that they tried versus regretted it if they had a recurrent seizure. Additional variables Data from the medical record included age, sex, race and seizure and epilepsy characteristics (table 1). Table 1Population descriptionVariableValuesN (%) or median (IQR); rangeSiteSUNY15 (47%)UM11 (34%)JFK6 (19%)Age46 (33–56); range 19–84Female15 (47%)RaceNon-Hispanic White14 (44%)Non-Hispanic Black14 (44%)Asian2 (6%)Hispanic/Latino2 (6%)Other0 (0%)Surrogate responder3 (9%)Age at first seizure21 (13–29); range 0–73Years between first and most recent seizure12 (4–23); range 0–65Years since most recent seizure3 (2–11); range 1–35Years since most recent seizure: at least 2 years23 (72%)Seizures impaired awareness31 (97%)Motor seizures30 (94%)Focal onset seizuresNo3 (9%)Yes26 (81%)Unknown3 (9%)Brain surgeryRefractory epilepsy surgery7 (22%)Any other4 (13%)First or second degree relative with epilepsyNo22 (69%)Yes7 (22%)Unknown3 (9%)History of febrile seizuresNo29 (91%)Unknown3 (9%)At least 10 seizures, lifetime19 (59%)Self-limited epilepsy syndrome0 (0%)History of convulsive status epilepticus4 (13%)Developmental delay3 (9%)Abnormal neurological examinationAny8 (25%)Cognitive/language4 (13%)Motor4 (13%)Sensory3 (9%)Epilepsy aetiologyUnknown18 (56%)Structural*9 (28%)Genetic3 (9%)Infectious2 (6%)Most recent EEG, epileptiform findingsNo previous EEG7 (22%)Epileptiform8 (25%)Not epileptiform17 (53%)Number of current ASMs04 (13%)119 (59%)27 (22%)32 (6%)Past attempt of ASM cessation due to seizure-freedom21 (66%)ASM co-indicationsMigraines9 (28%)Neuropathic pain3 (9%)Tremor3 (9%)Bipolar1 (3%)Current ASMs*Newer generation25 (78%)Older generation7 (22%)Number of previously tried ASMs06 (19%)111 (35%)21 (3%)34 (13%)46 (19%)53 (10%)Last provider seenMedical Doctor30 (94%)Doctor of Osteopathy1 (3%)Nurse Practitioner1 (3%)Last provider seen specifically for seizuresEpileptologist25 (78%)Non-epileptologist neurologist4 (13%)Non-neurologist3 (9%)Structural: traumatic brain injury (3), ischaemic stroke (2), tumour (2), mesial temporal sclerosis (1), cyst (1).N=32.*Antiseizure medications with more than 1 patient included levetiracetam (10), lamotrigine (9), carbamazepine (4), oxcarbazepine (3), lacosamide (3) and topiramate (2). Antiseizure medications taken by only 1 patient included: clonazepam, zonisamide, phenobarbital, clobazam, perampanel, primidone and valproate.ASM, antiseizure medication; JFK, John F Kennedy University Medical Center/Hackensack Meridian School of Medicine; SUNY, State University of New York Downstate Health Sciences; UM, University of Michigan. Population description Structural: traumatic brain injury (3), ischaemic stroke (2), tumour (2), mesial temporal sclerosis (1), cyst (1). N=32. Antiseizure medications with more than 1 patient included levetiracetam (10), lamotrigine (9), carbamazepine (4), oxcarbazepine (3), lacosamide (3) and topiramate (2). Antiseizure medications taken by only 1 patient included: clonazepam, zonisamide, phenobarbital, clobazam, perampanel, primidone and valproate. ASM, antiseizure medication; JFK, John F Kennedy University Medical Center/Hackensack Meridian School of Medicine; SUNY, State University of New York Downstate Health Sciences; UM, University of Michigan. Patients were also sent a post-interview survey reporting their health status, education, driving needs, general beliefs about medication harms30 31 or maximising their medical care (a 1-item Medical Maximizer-Minimizer scale),32 and self-reported numeracy33 to correlate with responses and further describe our population. Analysis We used a convergent parallel design whereby we collected both qualitative and quantitative data. We described data using medians with IQRs and frequencies. We compared participants with eligible non-participants to evaluate for non-response bias. We conducted two exploratory, prespecified quantitative analyses. Quantitative data were analysed using Statistical Analysis System 9.4 (Cary, NC) and Stata 17.0 (College Station, TX). First, because past experiences with ASM discontinuation may affect future views, we compared patients who had versus had not previously discontinued any ASMs due to seizure-freedom, regarding whether they would be glad to try in the future even if it resulted in seizure recurrence. With our sample size of 9 in one group and 22 in the second group, power to detect differences with a two-sided hypothesis at α=0.05 would be 7%, 16%, 31%, 52%, 75% and 92% to detect absolute differences of 10%, 20%, 30%, 40%, 50% and 60%, respectively. Second, because views regarding discontinuation could also be influenced by traits inherent to individuals rather than past experiences, we also stratified responses according to the Medical Maximizer-Minimizer scale (Fisher’s exact tests). Interviews were transcribed automatically by recording software, then reviewed and corrected by a study team member who listened to the audio for accuracy. Transcriptions were then uploaded to Dedoose version 9.0.17 (Los Angeles, CA; SocioCultural Research Consultants, LLC). Two coders familiar with study objectives and qualitative coding reviewed each transcript using deductive (researcher-driven) and inductive (response-based) reasoning to independently determine themes. Differences in themes were resolved by consensus then submitted to the entire study team for refinement. As we analysed text to explore the underlying broader narratives that recurred across interviews rather than the presence or frequency of specific words, we used thematic rather than content analysis. We labelled quotations by institution: J# refers to participants from JFK, S# from SUNY and M# from UM. Results Population We screened 189 adults with epilepsy, of whom 74 (40%) were seizure-free for at least 1 year, and 32/74 (43%) participated. Ages ranged 19–84 years (median 46, IQR 33–56), seizure-free duration ranged 1–35 years (median 3, IQR 2–11), 3 (9%) had a surrogate responder and 11 (35%) had prior intracranial surgery (table 1). There were no significant differences between responders and non-responders (online supplemental table 1). Viewpoints regarding ‘failed’ ASM discontinuation attempts Quantitative responses about past attempts 21 (66% of 32) had tried to discontinue any ASM in the past due to a period of seizure-freedom (table 2). 18 (86%) had discontinued ASMs at their doctor’s advice, 2 (10%) had done so on their own and 1 (5%) was unknown. Four were no longer on any ASMs at the time of our interview, whereas 17 had resumed ASMs due to a seizure recurrence. Table 2Anticipated future decision regret, stratified by whether patients had previously attempted antiseizure medication discontinuation due to a period of seizure-freedomHave you ever attempted ASM discontinuation?Would you be glad you tried to discontinue antiseizure medications due to the knowledge gained, even you had a post-discontinuation seizure recurrence?RegretUnsure/mixedGladNot askedNo5 (45%)2 (18%)2 (18%)2 (18%)Yes5 (24%)5 (24%)11 (52%)0 (0%)Total10 (31%)7 (22%)13 (41%)2 (6%)Percentages are displayed within each row. Note that the text presents percentages with the denominator of non-missing responses (excluding the ‘Not asked’ column presented in this table; Fisher’s exact test p=0.21).ASM, antiseizure medication. Anticipated future decision regret, stratified by whether patients had previously attempted antiseizure medication discontinuation due to a period of seizure-freedom Percentages are displayed within each row. Note that the text presents percentages with the denominator of non-missing responses (excluding the ‘Not asked’ column presented in this table; Fisher’s exact test p=0.21). ASM, antiseizure medication. Qualitative responses about past attempts Many patients with a past attempt indicated that it was worth trying Many reflected favourably on their experience due to the knowledge gained increasing confidence in their treatment and indicated that attempting was reasonable: J24: I don’t regret it. It was probably the perfect time to try it (medication weaning) out because I wouldn’t get my driver’s license until later. M15: I like to understand the trade-offs, and often in neurology they can be tricky to anticipate. (The neurologist) suggested going off the Keppra. We learned it didn’t work, but that seemed to me like an appropriate thing at the time to suggest. M50: I did stop medication for a short time and then, it didn’t turn out so well and then we had to go back on. That was worth a try. Cause I’d rather not have to dedicate part of my day just to carrying something around all the time being tethered to this thing (medication bottle). J31: …it was worth a shot. Others described a past unsuccessful discontinuation attempt with regret One patient expressed some degree of regret due to a seizure recurrence after trying to wean off phenytoin. M52: I was glad to be off the meds…especially because of the side effects of the Dilantin, and I wish that he had not taken me off so soon. … That would be my advice to anyone who’s thinking about it, unless there’s a certainty, which probably there’s not, that you’re not gonna have another seizure, I don’t think it’s worth it. Another patient (J52) with previous acute symptomatic seizures after past intracranial surgery weaned medication at 6 months. He was “curious if I was going to need it for my entire life.” However, he experienced recurrent seizure days into weaning and thus has been back on levetiracetam for the last 10 years. His initial reflection during the interview expressed that in hindsight he might have made a different decision: J52: I was dumb and didn’t understand the severity of seizures. However, with further reflection, he elaborated that he nonetheless gained important information from the attempt: J52: I think I’m probably happy that I did it now knowing what happened, and I wasn’t injured and no one else was injured because of it. Now knowing that I need to be on the medication that I’m happy having that knowledge and experience. Quantitative responses about hypothetical future attempts 30 out of 32 (94%) responded during the interview regarding how they might feel about a future discontinuation attempt. Two patients were not asked this question due to interview time constraints, neither of whom had previously attempted ASM discontinuation (M57, S12). 13 (43% of 30 respondents) felt that if they were to attempt ASM discontinuation in the future even if it resulted in a seizure, they would still be glad to have done so due to the knowledge gained from the attempt. This corresponded to 11/21 patients (52%; 95% CI 31% to 73%) who had previously attempted ASM discontinuation due to a period of seizure-freedom, compared with 2/9 patients (22%–95% CI 5% to 60%) who had not previously attempted ASM discontinuation (Fisher’s exact test p=0.23). We then analysed responses according to the Medical Maximizer-Minimizer scale (table 3). 26 out of 32 (81%) completed the survey including responses to this scale. Responses ranged from 1 (prefer to do the minimum necessary) to 7 (prefer to do everything possible), with a median of 7 (IQR 6–7). Among the four patients whose response to the scale was 1 (n=1) or 2 (n=3) (ie, prefer less intensive medical care in general), all four (100%) indicated they would be glad if they attempted ASM discontinuation even if it resulted in a seizure. Among the 21 patients whose response to the scale was 6 (n=6) or 7 (n=15) (ie, prefer more intensive medical care), 8 (38%) would be glad (Fisher’s exact test comparing a response of ‘glad’ for all non-missing scale responses p=0.06). Table 3Anticipated decision regret, stratified by preference for medical treatment intensity“In general, how do you approach your healthcare? I prefer to do…”Would you be glad you tried?% glad, among responders in each row (excluding the ‘Not asked’ column)RegretUnsure/mixedGladNot asked*1—the minimum necessary00101/1=100%200303/3=100%30001–40000–50000–632101/6=17%7—everything possible53707/15=47%Did not respond22111/5=20%Total10713213/30=43%Fisher’s exact test p=0.06 for the 25 patients who responded to both questions, comparing ‘Glad’ versus ‘Regret or unsure/mixed.’*Although this question was included in the semistructured interview script for all patients, interviewers did not ask this question to two patients due to interview time constraints. Anticipated decision regret, stratified by preference for medical treatment intensity Fisher’s exact test p=0.06 for the 25 patients who responded to both questions, comparing ‘Glad’ versus ‘Regret or unsure/mixed.’ Although this question was included in the semistructured interview script for all patients, interviewers did not ask this question to two patients due to interview time constraints. Qualitative responses about hypothetical future attempts Patients provided a wide range of attitudes regarding how they might feel about future attempts. Among those with a past discontinuation attempt, some remained favourable about the possibility of a future discontinuation attempt: S11: I would be glad that I tried because I’m confident in how I feel and that fear of having another seizure has gone away. So, I’d be pretty glad that I did it. Others with a past discontinuation attempt indicated they would be unlikely to try again and anticipate regret if they did so and had a recurrent seizure: S6: I would regret it because having the seizure was more problematic, more debilitating than taking the pills. M55: I think that it would have been worth going off because the side effects were intolerable. I would be very disappointed. That would be like the worst day of my life, I think. That wouldn’t be the end of the world. But it would be really difficult for me. I think that I would be worried about the probability of future seizures. And I think I would just be disappointed because I’ve gone so far, so long without seizures, and I’ve been so confident in the effectiveness of surgery. And so, it would be very difficult. Another example of a patient with a nuanced view, J22 described his past discontinuation attempt after seizure-freedom following intracranial surgery. At the time your doctors say, ‘Hey, you might be able to come off of this’. Yeah. Okay. And I was happy about that. He was thereafter seizure-free off ASMs for 5 years. Though after resuming treatment after a relapse, he now feels ‘safe’ when he takes his ASM and at this time would be unlikely to reconsider discontinuation. He also described his rationale based on the understanding from his doctor that any more seizures might lead to him undergoing more brain surgery. The likeability (sic) of me having a seizure (if I were to continue ASMs) is zero to none. I don’t wanna have another seizure in front of my boss. I know because I’ve experienced it firsthand that if I’m not on the medication, I’m at risk of having a seizure because it’s happened to me many times now. However, he would still ‘for sure’ consider coming off ASMs again in the future after several more years of seizure-freedom if his doctor thought his risk was low: Among those without a past discontinuation attempt, opinions were likewise mixed regarding the degree of anticipated decisional regret if they were to discontinue in the future with a resultant recurrent seizure. Some expressed that they would be glad to have tried due to the knowledge gained from the experience. S2: I would say I’m 50-50 on that. No, maybe not. Maybe 80-20 (80 glad). I think that I would be glad because I would have a definite answer that I need this medication, but then I would like just be upset with my body. Like, why is this happening? I guess come into terms that this is something that is just with me forever. Because I know that I have my definite answer. S8: Well, I mean, I don’t think I’d regret trying. I just wanna be like, “Basically, cool that we tried. Let’s not do that again.” S14: I’d be glad that we tried because you can’t live on it forever. You never know what’s gonna happen. Others expressed more anticipated regret and caution: M60: I don’t want to go having seizures for science. S13: I think I would regret, because it seems as if I would have to start all over again. I’m gonna have to start taking the medication and get it back into my system. That’s why I wouldn’t (attempt to discontinue ASMs). Discussion We explored attitudes towards both past and possible future ‘failed’ ASM discontinuation attempts, that is, how patients felt or might feel if they were to attempt ASM discontinuation followed by a seizure. Prior guidelines have focused on predicting ‘successful’ drug withdrawal,16 referring to patients with a low post-withdrawal seizure risk. However, our study highlights that a decision should be evaluated not only by seizure recurrence, as a poor decision can result in a ‘good’ outcome (a high-risk patient who fortunately did not seize) just as a good decision can result in a ‘bad’ outcome (a low-risk patient who nonetheless seized). Broader decision-making literature emphasises that ‘decision quality’ is determined by whether the patient made an informed choice (knowledge) that was consistent with their values (value concordance). We demonstrated that patients could see a ‘failed’ attempt either way—either being glad they tried even if they had a recurrent seizure thereafter, or else regretting having tried. Slightly under half of our sample indicated they would be glad they tried to discontinue in the future, even if it resulted in another seizure, due to the information gained about ASM necessity. Patients with a prior attempt viewed a hypothetical ‘failed’ future discontinuation more favourably than those without a prior attempt. This was not statistically significant, but our sample size was insufficiently powered and thus this can only be viewed as hypothesis-generating. Still, this could make sense. Our data suggest that those patients’ baseline beliefs are less favourable towards medications, separate from experience with weaning ASMs. Another reason could be that ‘anticipated’ regret is typically perceived as being more severe than ‘experienced’ regret.3436 A patient who might be conflicted regarding whether to continue ASMs could still have benefited from a ‘failed’ discontinuation attempt by boosting confidence in their long-term ASM indication. Yet another reason could be hindsight rationalisation or recall bias, whereby patients retrospectively justify decisions. Regardless of the predominant explanation, our results nonetheless illustrate that many patients ultimately later feel glad they tried to withdraw even if a seizure occurs. Some participants would not consider discontinuation attempts unless there was certainty of seizure freedom. This rationale is concerning for several reasons. First, this view fails to acknowledge that certainty cannot exist even if they continue ASMs indefinitely. The largest randomised trial of seizure-free patients with epilepsy found approximately a 40% vs 20% chance of another seizure among arms assigned to discontinue versus continue ASMs, respectively. This potentially overstated sense of security on ASMs may understate the non-zero chance of another seizure regardless, for example, in this population that is typically permitted to drive. This view also lacks consideration of likely the key input—the magnitude of the absolute risk difference comparing continuation versus discontinuation, which is not 100%. Possible decision regret is an inherent part of making decisions under uncertainty when the consequences of a decision emerge only much later. Optimal decisions require that patients and clinicians alike are equipped with adequate knowledge regarding seizure chances on versus off treatment, being cautious regarding overattributing causality too quickly and being aware that while seizure reduction is an important goal, it is critical to have realistic expectations regarding imperfect treatment effects. It would be fascinating to conduct analogous interviews in future work regarding whether patients would voice the need for certainty regarding seizure-freedom before agreeing to start ASMs. Limitations With a sample size of 32, it is impossible to generalise to all patients or all centres, with predominantly motor seizures seen mostly by epileptologists, and statistical comparisons across patient subgroups are not possible. We report a limited number of prespecified quantitative comparisons of greatest interest, awaiting confirmation in future larger studies. However, the three centres represented a wide range of socioeconomic diversity and level of complexity. We also feel that heterogeneity is a strength—excluding, for example, those three surrogate responders or those with shorter than 2 years seizure-freedom, etc would have been a limitation due to omitting potentially important voices, and it is simply impossible to compare how responses differed according to all potentially clinically important features. Additionally, while categorical responses such as whether patients would be glad they attempted discontinuation may not capture complex decision-making reflections, that was the purpose of our rich qualitative data. Ultimately, the purpose of qualitative work is depth—obtaining a detailed understanding in a smaller sample of key informants to uncover dominant themes. Future large-scale quantitative work can statistically compare preferences across subgroups. Non-response bias is possible. Those who responded could have had more (un)favourable views about ASMs in ways not captured by the electronic medical record. However, at the two sites where non-respondent characteristics were available, we did not find any significant differences between responders versus non-responders even after analysing important dimensions such as seizure-free duration and seizure severity metrics. Interviews about past decisions can be subject to recall bias or hindsight rationalisation. However, when studying subjective experiences, these features of decision-making nonetheless are real, integral components when reflecting on past decisions. We did not conduct member checking. Member checking is the process of soliciting feedback from participants about data and the researchers’ interpretation. While member checking can serve as a useful validity check to ensure that participants’ responses are accurately reflected as they intended, others have questioned the degree to which this process improves scientific quality and may burden participants.37 Conclusions We interviewed seizure-free patients with epilepsy or caregivers about whether an ASM discontinuation attempt would be ‘worth a try’ even if a seizure reoccurred. Our data support that while a discontinuation attempt may be inappropriate for many patients, we provide critical insight that even a ‘failed’ discontinuation attempt can have value due to the knowledge gained from the experience. Patients also sometimes voiced that they might consider discontinuation only if remaining seizure-free was certain. This suggests that people with epilepsy could be better counselled that some seizure recurrence risk is present regardless of continued treatment and regardless of seizure-free duration. Supplementary material