Direct and indirect costs and cost-driving factors in adults with tuberous sclerosis complex: a multicenter cohort study and a review of the literature

The present study was designed as a cross-sectional, multicenter survey that enrolled individuals with TSC throughout Germany (Berlin, Bochum, Dresden [Radeberg], Frankfurt, Greifswald, Homburg, Kempten, Marburg, München, Münster [Lingen], Rostock, Stuttgart, and Wiesbaden) and through the German TSC patient advocacy group (Tuberöse Sklerose Deutschland e.V., Wiesbaden, Germany).

After receiving written informed consent from the patients or their legal guardians (if applicable), all individuals with TSC were deemed eligible for study inclusion. We based the diagnostic criteria for TSC on the latest recommendations established by the 2012 international TSC consensus conference [19]. We identified seven primary manifestation categories affected by TSC, including epilepsy, structural brain defects, psychiatric, heart/circulatory system disorders, kidney and urinary tract disorders, dermatological system manifestations, respiratory system manifestations, and other manifestations [16]. The seizure and epilepsy syndrome classifications were adapted to the latest definitions established by the International League against Epilepsy (ILAE) [20, 21]. This study received ethical approval and was registered with the German Clinical Trials Register (DRKS00016045; Universal Trial Number: U1111-1229-4714). We closely followed the STROBE guidelines (Strengthening The Reporting of Observational Studies in Epidemiology) [22].

We asked individuals with TSC to complete a retrospective questionnaire based on their experiences during the previous three months. The questionnaire was validated in earlier studies [23‐26] and we adapted it for use in individuals with TSC. The questionnaire included 36 questions relating to disease characteristics (e.g., genetics, affected organ systems, seizures, medications, and additional symptoms), healthcare resource use (e.g., healthcare visits, accidents, and emergency care), and social conditions. Paper questionnaires in German were sent to individuals with TSC between February and July 2019.

The aim of this study was to calculate the specific genuine costs associated with TSC, rather than those associated with conditions unrelated to TSC. Therefore, we asked individuals in detail whether the medications, services, and other resources that were consumed were associated with particular organ manifestations of TSC. We evaluated costs using a bottom-up approach from the perspective of the statutory health insurer (“Gesetzliche Krankenversicherung” [GKV]) and society as a whole. The cost categories that were included in the analysis were direct health service costs, patients’ out-of-pocket (OOP) expenses, care grade allowances as approximation of informal care costs, and indirect costs. We evaluated these costs according to the German recommendations for performing health economic evaluations [27].

We conducted statistical analysis using IBM SPSS Statistics, version 26 (IBM Corp., Armonk, NY, USA). We summarized the variables of interest using the mean, median, and standard deviation (SD). For cost data, we calculated 95% confidence intervals (CI), using the bootstrap-corrected and accelerated (BCa) method with n = 2000 repetitions to estimate parameters robust to skewed distributions and outliers [33, 34]. Due to the small population of TSC and related statistical challenges [35, 36], we refrained from a power calculation or a predefined number of participants, and aimed to include all potential patients with TSC in Germany. We compared groups using adequate parametric and nonparametric tests after testing for the normality of distribution. The significance level was assumed at p < 0.05. We investigated the relationships between an individual’s clinical characteristics and TSC-related costs using multivariate linear regression using the BCa method with 2000 repetitions. Total direct, total indirect, and nursing care-level costs were regressed against a set of clinical variables, which we selected following univariate analysis and according to evidence presented by previous cost-of-illness studies examining TSC [16, 37, 38]. We tested all variables for interactions and collinearity. To identify independent predictors of costs, we performed standard multiple linear regression analysis using the bootstrapping technique and applied a Bonferroni correction for multiple testing.

One hundred and ninety-two adults with TSC completed the questionnaire. The mean participant age was 33.4 years (SD: 12.7 years; median: 31.0 years; range: 18.0–78.0 years), 51.6% (n = 99) were women. Among the respondents, TSC was diagnosed at a mean age of 10.4 years (SD: 14.9 years; median: 2.0 years; range: 0–66.0 years), and the first symptoms of TSC were noted at a mean age of 5.7 years (SD: 12.0 years; median: 0 years; range: 0–66.0 years). In three individuals (1.6%), a diagnosis of TSC was suspected before birth, based on ultrasound examinations. Pathogenic variants in TSC1 were reported by 32 individuals (16.7%), and pathogenic variants in TSC2 were reported by 48 individuals (25.0%, ratio 1:1.5). Three individuals (1.6%) suffered from a polycystic kidney disease with tuberous sclerosis (PKDTS) contiguous gene deletion syndrome.

Most individuals lived with others. Of those, 44 (22.9%) were married or in a relationship, and 84 (43.8%) lived with relatives. Less than half of individuals were either employed (n = 71, 37.0%) or participated in vocational training (n = 21, 10.9%). Further sociodemographic and clinical characteristics, including information on affected family members, are presented in Table 1. The majority of individuals suffered from a range of TSC organ manifestations. Disorders of the central nervous system were commonly reported, with 140 (72.9%) individuals reporting a diagnosis of epilepsy, 129 (67.2%) describing various structural brain disorders, and 97 (50.5%) indicating psychiatric disorders. Furthermore, 182 (94.8%) individuals reported skin manifestations, 142 (74%) described kidney and urinary tract disorders, and 97 (50.5%) indicated heart and circulatory system disorders. Additional details can be found in Table 2.

Mean total direct costs were calculated at EUR 6452 (median EUR 1920; 95% CI EUR 5533–7422) per study participant for the 3-month study period, and details are presented in Table 3 and Fig. 1. Direct medical costs were primarily associated with the costs of drug treatments (76.8% of total direct costs; mean EUR 4953 per 3 months; median EUR 573; 95% CI EUR 4087–5876), and hospitalization (8.0% of total direct costs; mean EUR 518; median EUR 0; 95% CI EUR 312–750).

The largest medication costs were those due to mTOR inhibitors (everolimus, n = 69; sirolimus, n = 2), with a mean of EUR 4358 per 3 months (67.5% of total direct costs; median EUR 0; 95% CI EUR 3448–5342). mTOR inhibitor costs were higher than those associated with ASDs, which were on average EUR 415 (6.4% of total direct costs; median EUR 104; 95% CI EUR 239–706). Individuals used on average 1.9 ASDs (SD: 0.8; median 2: range 1–4). The five most frequently prescribed ASDs were lamotrigine (n = 51; 26.6%), valproate (n = 46, 24.0%), oxcarbazepine (n = 32; 16.7%), levetiracetam (n = 25; 13.0%), and lacosamide (n = 12; 6.3%). Monotherapy with ASDs was prescribed to 24.0% (n = 46) of all participants, which was associated with significantly lower costs than polytherapy with two, three, or more ASDs (each p < 0.001). The detailed costs and daily dosages reported for different ASDs are listed in Table 4.

In total, 23 (12.0%) individuals reported at least one TSC-related hospital admission during the 3-month study period. Overall, 29 admissions were reported, with a mean length of stay of 5.5 days (SD: 3.6; median: 5 days; range: 1–14 days). Epilepsy and seizures resulted in eight admissions, whereas six admissions were associated with diagnostic procedures, four admissions were due to pneumothorax, three were related to operations concerning AML in the kidneys and other organs. A further two admissions were due to adverse reactions to everolimus intake and two were associated with facial skin treatments. Four admissions had other TSC-related reasons.

Ancillary treatments, such as occupational therapy, physiotherapy, and speech therapy, were prescribed to 54 participants (28%), and were associated with an average cost of EUR 125 per 3 months, which comprised 1.9% of total direct costs (median: EUR 0; 95% CI: EUR 84–174). In addition, families directly paid EUR 39 in therapy-related costs during the 3-month study period.

Fifty percent (n = 97) of individuals were categorized as requiring care grade levels I to V, based on the “Pflegebedürftigkeit” scale: 2.6% as level I (‘low impairment of independence’); 9.9% as level II (‘significant need for care’); 11.5% as level III (‘heavy need for care’); 13.5% as level IV (‘most difficult to care for’); and 13% as level V (‘most difficult to care for and special demands regarding nursing care’). Two individuals did not meet the level I–V criteria but were still in need of care according to their caregivers, and 48.4% of individuals denied being in need of care. The mean approximate costs for nursing care were EUR 929 (median: EUR 0; 95% CI: EUR 780–1083) over each 3-month period, or EUR 3716 annually, assuming that care is provided by family members. Patient’s caregivers reported that they had paid additional costs for care, with a mean of EUR 24 (median; EUR 0; 95% CI: EUR 10–41). Furthermore, they paid for supervision, with a mean of EUR 48 (median: EUR 0; 95% CI: EUR 22–79) per 3-month period. Further informal care costs that were neither reflected in the care grade allowance nor perceived by caregivers are inevitably not represented in our approximation of informal care costs. In total, 124 individuals (64.6%) had a handicapped ID, indicating a degree of disability between 70 and 100%.

The estimation of mean indirect costs was based only on questionnaire responses from patients of working age, younger than 67 years (n = 190). The mean total indirect costs were EUR 3174 (median: EUR 0; 95% CI: 2503–3840) over three months or EUR 12,696 annually. The main contributor to indirect costs (n = 30) was the inability to work due to intellectual disability, epilepsy, or kidney disorders (mean: EUR 1775; median: EUR 0; 95% CI: EUR 1183–2367). Furthermore, 19 individuals were only able to work part-time, which was associated with a mean estimated cost of EUR 514 ± 1762 per 3 months (median: EUR 0; 95% CI: EUR 283–792). Twenty-eight individuals reported missing days from work during the last three months due to TSC-related causes (mean: EUR 234; median: EUR 0; 95% CI: EUR 115–382), and six individuals were unemployed (mean: EUR 355; median: EUR 0; 95% CI: EUR 118–651). Five individuals reported retiring prematurely (mean: EUR 296; median: EUR 0; 95% CI: EUR 59–592). The details of indirect productivity costs can be found in Table 5 and Fig. 1. The mean duration of work absenteeism was 7.6 ± 10.7 days (range: 1–50 days) per 3 months.

To identify potential cost drivers, we performed univariate analyses for total direct, total indirect, and nursing care costs and a number of demographic and clinical characteristics. For details, please refer to Table 6. In the univariate analyses, younger age, the use of mTOR inhibitors, polytherapy with two or more ASDs, recurrent seizures, all TSC manifestation categories, the total number of TSC manifestations, and the level of disability were associated with higher direct costs. Lung manifestations (lymphangioleiomyomatosis), the total number of TSC manifestations, and disability were associated with higher indirect costs, whereas younger age, polytherapy with two or more ASDs, recurrent seizures, the TSC manifestations of epilepsy, structural brain disorders, psychiatric disorders, and skin manifestations, the total number of manifestations, and disability were associated with increased nursing care costs. Overall, total direct, indirect, and nursing costs increased with the number of affected organ systems (Table 6).

Multiple linear regression analyses revealed that the use of mTOR inhibitors independently predicted a 3-month direct cost increase of 12,069 Euro (BCa-corrected B 12,068.85, BCa-corrected standard error [SE] B 836.28, β 0.770, p < 0.001), and persistent seizures predicted a 3-month direct cost increase of 2113 Euro (B 2113.29, SE B 651.13, β 0.137, p < 0.001). Applying a Bonferroni correction for twelve comparisons, the threshold for the p-value was set at 0.00417, and the mTOR inhibitor use and persistent seizures were able to explain 71% (R², F (12, 179) = 37.09, F sig.) of the total direct cost variance. Older age and disability were independent cost-driving factors for total indirect costs, with disability predicting a 3-month indirect cost increase of 2131 Euro (B 2131.48, SE B 693.90, β 0.216, p = 0.004) and older age of 1220 Euro per 3 months (B 1220.05, SE B 391.20, β 0.212, p = 0.003). The significant factors together explained 13% of the indirect cost variance (corrected p < 0.0125; R² = 13%, F (5, 184) = 5.67, F sig.). Epilepsy, psychiatric diseases, and disability were independent cost drivers for approximate informal (nursing) care costs, with relatively similar cost-driving effects. Disability predicted a 3-month nursing-cost increase of 622 Euro (B 622.02, SE B 160.86, β 0.280, p < 0.001), psychiatric diseases of 599 Euro (B 599.45, SE B 149.67, β 0.282, p < 0.001), and epilepsy of 528 Euro (B 528.27, SE B 138.67, β 0.221, p = 0.001). Together, these significant variables (corrected p < 0.00625) were able to explain 51% of the 3-month nursing care cost variance (R² = 51%, F (9, 182) = 20.98, F sig.).

Our study highlights the substantial direct costs incurred by individuals with TSC. Medication was the largest single component of direct costs and associated with an estimated annual direct cost of EUR 25,808. The highest medication costs were due to mTOR inhibitors (annual costs of EUR 17,432), which were used by 37% (71/192) of the individuals in this study. mTOR inhibitor use was identified as an independent cost-driver in the multivariate analysis. This finding is somewhat expected, given the currently high price of mTOR inhibitors. Everolimus, the mTOR inhibitor overwhelmingly used by individuals in this study, was first given conditional marketing authorization as an orphan drug by the European Medicines Agency (EMA) in 2011 for TSC-associated SEGA and in 2012 for TSC-associated renal AML [46], followed by an extension of indication to epilepsy refractory to ASD in 2017. Due to the time point of this study, the long-term cost-effectiveness of this currently costly drug remains to be evaluated [47]. The possibility of avoiding potentially costly TSC consequences, such as resection surgery for SEGA or epilepsy, cerebral shunt placement, and AML-associated renal bleeding following everolimus may balance favorably against the unwanted treatment effects of mTOR inhibitors, and everolimus may emerge as a cost-effective treatment option. The costs associated with the use of mTOR inhibitors will also likely decrease in the future due to the availability of generic formulations.

Interestingly, gender differences were not associated with differences in TSC-related costs. Women with TSC are known to be more likely to develop AML, and AML in women tends to be larger and require more interventions [10]. Women are also almost exclusively affected by pulmonary manifestations of TSC, which have been demonstrated to incur high direct costs [11, 15]. The findings in the present study may be due to the relative rarity of severe LAM complications, which typically occur only with increasing age. However, we did record four hospital admissions among our cohort for pneumothorax, a known complication of LAM, within the short evaluation period of three months. An alternative explanation is that the salient contributions of mTOR inhibitor therapy to overall costs, which exceeded inpatient treatment costs, may have masked gender differences in our cohort. Young adults (18–29 years) with TSC incurred higher costs than older individuals did, which is likely associated with the performance of a larger proportion of diagnostic procedures and the increasing use of mTOR inhibitors among younger patients.

In adults, renal manifestations of TSC are more common than in children. The most common types of renal manifestations, including AML and renal cysts, tend to appear first during adolescence and grow during adulthood. In our study, a similar proportion of individuals reported AML as in the TOSCA cohort [10], and renal manifestations were a significant factor for direct costs in the univariate analysis. Interestingly, in the multivariate analysis, only recurring seizures remained an independent clinical cost driver, and no single other TSC manifestation category was identified as an independent cost driver, which may be due to the higher direct costs associated with recurrent seizures compared with all of the other seven clinical categories, which were all associated with similar direct costs. This finding supports the known severe burden of illness of ongoing epilepsy in TSC, particularly because a relevant share of individuals experience pharmacorefractory epilepsy [48]. The high direct costs are in line with those reported for other rare developmental and epileptic encephalopathies like Dravet syndrome [49] or Lennox-Gastaut-syndrome [50]. This finding further highlights the need for ongoing identification of epilepsy surgical candidates among those with TSC and epilepsy, new emerging therapies such like the MR-guided laser interstitial thermal therapy might help to increase the suitability of patients for a surgical treatment [51]. Generally, the results of this study indicate that the management of TSC might result in high direct costs that exceed the costs incurred by all-cause epilepsy patients [26]. Our multivariate analysis model was able to account for 71% of the total variance in directs costs, suggesting that mTOR inhibitor use together with recurring seizures can explain a relevant share of the direct cost components among TSC patients in Germany. Similar to results reported for the United Kingdom [15], direct costs increased with the number of TSC manifestations. Although this finding was expected due to the complexity of TSC, which necessitates surveillance and treatment for most manifestations, this finding further demonstrated the need for systemic causal treatment and integrated, streamlined care, such as by specialized TSC centers.

Cannabidiol (CBD) is a new treatment option for drug-resistant seizures associated with TSC, it has recently been approved by the United States Federal Drug Administration (FDA) and the EMA. CBD has shown promising results in a randomized-controlled study published recently [52]. Since our study preceded the approval of CBD for TSC in the European Union, only two patients were treated with CBD in our cohort and the influence of CBD on the direct cost is negligible in our results.

Most cost categories were heavily skewed due to the clinical heterogeneity among adult individuals with TSC. While a few patients required significant health care resources, many were only mildly affected. In our study, the ratio of patients diagnosed with TSC1 vs. TSC2 mutations was slightly higher than anticipated from past study findings [12, 53]. Because individuals with TSC2 mutations tend to be more severely affected, especially by neuropsychiatric manifestations and epilepsy, an even higher COI can be presumed among populations with higher shares of TSC2 mutations. However, we caution that a large proportion of the patients in this study did not report any genetic test results.

To our knowledge, this is the first study to report the indirect costs of TSC. Nearly half of individuals (86/192, 46%) who participated in this study reported productivity losses. Adult individuals with TSC in our study incurred substantial indirect annual costs, equal to EUR 12,696 per year, with the largest share due to the inability to work (EUR 7100). Approximately two-thirds (132/192, 67%) of individuals in our study were able to work without impairment (excluding extra days off due to TSC). The indirect costs reported in our study are broadly similar to those reported for other rare neurological diseases, such as spinal muscular atrophy [54] or Becker’s muscular dystrophy [55], but were substantially higher than those reported among individuals with all-cause epilepsy [24]. Unlike direct costs, only a few clinical or demographic categories were associated with increased indirect costs on univariate analysis. We identified both the manifestation burden and the level of disability as associated with increased indirect costs. LAM also emerged as a variable significantly associated with higher indirect costs, indicating that while rare overall, LAM may play an outsize role in affected individuals due to its associated severe impairments.

Incomplete participation due to part-time work, unemployment, and early retirement was common even among those who were able to participate in the primary work market. Additionally, TSC impairs participation in education, as demonstrated by the high proportion of individuals that did not report any job qualifications (42.2%) and that did not graduate from school (30.2%). In addition, half of patients required care on the care grade allowance scale. Most clinical variables were not independently associated with higher approximated informal (nursing) care costs in the multivariate analysis, e.g. heart manifestations (which was primarily arterial hypertension among this adult cohort), renal manifestations, and lung manifestations. These types of clinical manifestations are typically intermittent, as with AML bleeding, or mostly do not necessitate nursing care, such as renal or pulmonary insufficiency. In contrast, epilepsy, neuropsychiatric manifestations, and disability were independent cost drivers that frequently required nursing care. We must caution that our approach of assessing informal care costs from allowances according to care grades necessarily misses all intangible informal costs that were not perceived by the caregiver as further costs and thus not specifically reported. This includes e.g. losses of caregiver productivity due to providing informal (= not by a health professional) care. True informal care costs are thus most likely higher than in our approximation.

The direct costs identified in our study are broadly comparable to those reported by other recent studies (see Table 7 for a comparison of costs). However, costs in our cohort were both higher than the annual direct costs reported by studies from the UK (GBP 4227–5054 per year) [15, 16, 41] and lower than the costs reported by two studies from the US (USD 8543–85,397 per year) [37, 38]. The direct costs associated with ASD were lower in our study than those reported in previous studies, despite the specific drugs that were reported being similar [44]. This finding likely reflects lower medication costs in Germany due to price negotiations between statutory health insurers and drug manufacturers in recent years [56]. Importantly, this finding of lower costs was not true for mTOR inhibitors, which incurred higher costs than were reported by another study from the US [44]. The costs associated with hospitalizations and outpatient treatment were lower in our study than in earlier studies. Interestingly, ancillary therapies represented a smaller share of total direct costs than were reported in another study that explicitly reported this variable [43], which is likely due to differences in the reimbursement policies between varying healthcare systems. In general, direct comparisons to studies from different settings or countries are difficult because due to differing definitions, policies, measurements and other factors. To date, only one other study has analyzed productivity losses in adults with TSC [43]. Interestingly, although this study was performed in a different country, activity impairment was similar, with approximately one-third of adult individuals with TSC reporting not fully participating in the work market.

In terms of the use of healthcare resources, our results appear to be in line with other COI studies, particularly hospitalization frequency and the use of ASDs [9, 11, 15, 16, 37, 42, 44, 45]. Most studies could not evaluate mTOR inhibitor use because the periods of data inclusion preceded their authorization. In our study, more patients (37%) used mTOR inhibitors than in three other recent studies (8–16.5%) [37, 44, 45].

A particular contribution of this study is the collection of data regarding the indirect costs and the nursing requirements, which were measured by the care grade allowances. Nursing-associated costs were identified as an important cost component, associated with annual expenditures of EUR 3716, reinforcing the importance of different organ manifestations and seizure-related costs, as were reported by Skalicky et al. [43].

The limitations associated with the questionnaire used in this study include the potential for recall bias regarding three-month-old events, which might result in incomplete and underestimated costs. Furthermore, although the sample consisted of individuals recruited from a variety of sources (multiple clinics and centers across Germany and through the patient advocacy group), we do not know whether the included sample is representative of individuals with TSC in Germany owing to the rarity of TSC. Only two individuals older than 67 years of age were included in this study, which may indicate the limited access to specialized care among this vulnerable group. In addition, the analysis of cost drivers should be interpreted with caution given the limited sample size. However, the significance that the number of organ manifestations had on COI in the current study suggests some common ground with earlier studies, which also identified the number of organ manifestations as cost driver [16]. In addition, the skewness identified in the cost calculations should be noted, as disparities were noted between the mean and median costs. A limitation is the incomplete capturing of informal care costs, as we based these only on allowances paid to the individual for support of caregivers, and services that were paid out-of-pocket by the patients or caregivers. Further intangible and likely substantial informal care costs such as work productivity loss of caregivers is explicitly exempt from our analysis. Another limitation of the study was due to the calculation of indirect costs using the human capital approach. The study was performed in 2019 in a situation of nearly full employment in the general population before the onset of the Corona virus disease (COVID-19) pandemic, however indirect costs may not exactly reflect the burden on society and may be overestimated [57]. However, due to the limitations of the friction cost approach [58], we retained the human capital approach, which is in accordance with the German and international recommendations for performing health economic evaluations [59‐61]. The major strength of the study remains its large sample size of 192 individuals and caregivers, which is significant given the rarity of TSC, and the timing of this analysis after mTOR inhibitors were licensed for the treatment of various disease manifestations.