In this study, we demonstrated the effectiveness and safety aspects of treatment with everolimus in 64 patients with TSC in Norway and Denmark. This is one of few unselected, population-based studies from clinical practice in countries with similar health care systems, characterised by high-quality follow-up and equal access to health care. In the following sections, we compare the outcome measures with results from the randomised, double-blinded, placebo-controlled EXIST studies [22,23,24,25,26] and other clinical studies.

The effectiveness of epilepsy treatment with everolimus, defined as at least 50% seizure reduction, was observed in one-third of patients with epilepsy in this study. This proportion is similar to the patients treated with a low dose of everolimus in EXIST-3 [24], and similar to another randomised, placebo-controlled trial [48].

This finding might be lower than expected, since most patients also changed ASMs during the study period. Interactions due to concomitant use of enzyme inducers could explain some of the changes, as CYP3A4-mediated metabolism of everolimus is affected by drugs such as carbamazepine, phenytoin, and phenobarbital [49, 50]. However, only six patients changed an enzyme-inducing drug.

In contrast to RCTs studies, it is difficult to know if the seizure frequency reduction is caused by the intervention alone in real world studies [31, 34].

In our study, the proportion of patients with ≥ 50% seizure reduction was higher in Denmark than in Norway.

The use of everolimus for epilepsy is restricted in Denmark, to ensure sufficient effect of everolimus, which is an expensive drug with potentially serious adverse effects. Thirty-three and fifty percent seizure reduction, good cooperation, and tolerable adverse effects are required after 4 and 12 months, respectively, to continue treatment. To fulfil these requirements, Danish patients submit seizure diaries for review, and the treating physician submits an evaluation to the Danish Medicines Authorities for documentation after the first treatment year. This is not required in Norway.

The observed variability in effectiveness between Norway and Denmark may not reflect a real discrepancy in efficacy, but may be influenced by the requirement to fulfill efficacy criteria in Denmark.

Another possible explanation is that patients treated for epilepsy indication were younger at start of treatment in Denmark than in Norway and although the difference were not significant, it is possible that also influenced the difference. In addition, there was not difference in patients with less than 50% reduction between Norway and Denmark.

The proportions of patients who reported ≥ 50% seizure reduction varied between 33 and 78% [39, 41, 47] in other clinical studies, possibly due to differences in inclusion and selection criteria. The study with the best effectiveness included participants without refractory epilepsy; all were children, and almost half of them were under 2 years [47]. This study reported > 50% seizure reduction in > 90% of 47 children aged less than 2 years treated with sirolimus [47].

Patients who started treatment before 2 years did not report a better effect in our study, although seizure reduction was associated with younger age, as also reported previously [27].

The number of ASMs did not decrease under everolimus treatment in our study. The different safety profiles of everolimus compared to other ASMs, gives a higher and different adverse effect load for patients treated with everolimus for epilepsy.

In contrast to the results from EXIST-3, which reported higher odds for response in patients treated with high exposure [24], seizure reduction was not associated with the calculated C/D-ratio in our study. This might be due to quite few participants, lack of data in some patients, less use of concomitant enzyme inducers among other antiseizure medications and extensive physiological and pharmacokinetic variability and tolerability between patients.

However, if the efficacy of everolimus is not dose dependent, as it seems in some other studies [38, 39], this could suggest that treatment with lower doses could be sufficiently efficacious and associated with fewer adverse effects. More studies are needed to investigate this.

No growth or a small reduction in rAML and SEGA lesions might be sufficient to prevent symptoms, whereas a small reduction in seizure frequency might not be as clinically relevant. However, a slight reduction in seizure frequency may make a difference in everyday life for persons with epilepsy [51], indicating that outcome measures ≤ 50% seizure frequency reduction could be clinical relevant, although a slight reduction in seizure frequency is a less reliable outcome measure. Seizure free days are a novel outcome measure [52] that could be considered in future studies.

A reasonable effect with regard to seizure reduction and tolerable adverse effects, as requested in Denmark, could, in general, be recommended to continue treatment for the epilepsy indication.

About one-third of our patients were responders (> 30% reduction of LD) versus 42% in EXIST-2 [23] and 58% in the final results of EXIST-2 [25]. These results are only partly comparable since the outcome measures in EXIST-2 differed from our study. EXIST-2 measured proportions of patients with ≥ 50% reduction in sum of volumes of all target angiomyolipomas (≥ 1 cm in LD) and measured change from baseline to best percentage change during treatment [53]. Our study measured proportions of patients with ≥ 30% reduction of LD of largest rAML and mean change of largest LD in both kidneys, and measured change from baseline to last imaging.

Volume measurement was not feasible for various reasons; it was too time-consuming, the automated method used for volume measurement did not capture accurate volume assessment, and some images were only taken in single sections. The change in size of rAML was measured according to Response Evaluation Criteria in Solid Tumors (RECIST) [54] because it is equivalent to volume measurement. Overall, 76% of the patients with rAML had at least a 10% reduction in the rAML size. In EXIST-2, volume reduction was described in 97% of the cohort [25]. In other clinical studies, this varied between 64 and 98% [41, 43, 46]. No renal bleeding was reported in our study, although a renal haemorrhage of older date was described on routine magnetic resonance imaging in one patient. No renal bleedings ≥ grade 2 were described during everolimus treatment in the final results of EXIST-1 and 2 [25, 26] or in the TOSCA Pass sub-study [12]. Embolisation because of flank pain was described in one patient under treatment in EXIST-2 [25].

SEGA volume decreased in all three patients. Our observation (43–71% volume reduction) is in line with or better than results from the EXIST-1 trial, in which 35% [22] and 58% [26] of the patients had at least a 50% reduction in SEGA volume. This should, however, be interpreted with caution, as only three patients were included.

The frequency of adverse effects was in line with EXIST-1–3 [24, 26, 53] but higher than reported in other studies [37, 39, 41, 47, 55]. Adverse effects ≥ grade 3 were reported in 35% of the patients, and were quite similar to those of EXIST-1 and 3 [22, 24], and varied between 0 and 35% in other studies [37, 46, 47, 55, 56].

Notably, patients with adverse effects ≥ grade 3 were younger than patients without adverse effects ≥ grade 3, this is in line with results in the extension study of EXIST-3 [27], making it especially important to be aware of management and follow up of adverse effects in younger patients.

The frequency of stomatitis/oral ulceration was slightly lower than those reported in EXIST-1–3 [22,23,24], and varied between 14 and 91% in other studies [37, 41, 46]. The study with the highest frequency reported a higher proportion of patients with intellectual disabilities and suggested that oral care might be insufficient [46].

The upper respiratory tract infections in our study were quite similar to those in EXIST-1–3 [22,23,24], but our frequency was higher than those reported in other studies [37, 55, 56]. Hypercholesterolemia, hypertriglyceridemia, and leukopenia were reported more frequently than in EXIST-1–3 [22,23,24, 26], and varies in other studies [29, 37, 46, 56]. Inequality in patient selection, serum concentration, management, education, and follow-up might explain some of the variation in the frequency and severity of adverse effects.

Dose reductions and interruptions were reported in 50%, as reported in EXIST-2 [53], and were higher than those reported in other studies (22–31%) [37, 41, 55]. Dose reductions or interruptions should be considered with adverse events ≥ grade 2 [57], and the high frequency in our study might reflect that management recommendations were followed [57]. However, dose reductions and interruptions occurred in only 25% of patients with stomatitis/oral ulcerations. For almost half of patients with symptomatic stomatitis/oral ulcerations everolimus treatment were not modified indicating that monitoring and management of adverse effects still needs to be better implemented.

Immunodeficiency disorder was described in one patient. This is a serious condition and is as far as we know not previously described as an everolimus related adverse event. Immunoglobulins were not measured before initiation of everolimus in this patient, and consequently it was not possible to determine for certain whether the patient that developed immunodeficiency disorder has a primary immunodeficiency, or if this is drug related. Clinically infections increased significantly after treatment initiation, making everolimus treatment a probable cause.

Data collection was finished for most patients before the outbreak of covid-19, only 14 had follow-up for days up to a month after the outbreak. For two of the nine patients that discontinued everolimus treatment, the COVID-19 pandemic was part of the reason, but apart from that it is not likely that the study result was influenced by the pandemic.

The study included unselected patients from two countries with similar health care systems and follow-up from specialists. The patients were recruited from the Norwegian TSC population and from two regions in Denmark. It is possible that some treated patients invited through the National Centre for Rare Epilepsy-Related Disorders in Norway did not respond to the invitation. Only two invited patients treated with everolimus did not want to participate. Due to equal access to health care services and follow-up from specialists, it is likely that most patients with indications for treatment are known and included. In Denmark, everolimus treatment is centralised, and all treated adult patients from Aarhus University Hospital and all paediatric patients from University Hospital Rigshospitalet were invited. The total response rate in Denmark was high (88%), indicating that the included patients were representative and the risk of selection bias low.

Due to the observational design, missing data in medical records and data unconformity were important limitations. Patients’ and parents’ interviews were carried out to reduce this limitation, but potential recall bias could not be excluded. Imaging data were not available for all patients and reduced the sample size. Imaging was reevaluated by experienced radiologists, ensuring data conformity.

Due to the irregular (multilobular) shape of most of the SEGAs, the simplified method of volume measurement by multiplying diameters in three orthogonal directions and divided the result by 2 was considered too inaccurate. Despite the chosen method, some inaccuracy in volume measurements may persist due to differences in imaging quality across multiple MRI exams from different centres and over the actual time period.

Other ASMs were changed in a majority of the patients, and some underwent epilepsy surgery, which made it difficult to draw conclusions on the cause of the change in seizure frequency. Different conditions for epilepsy treatment between Norway and Denmark were also limitations.