Introduction
Idiopathic normal pressure hydrocephalus (iNPH) is a treatable gait disorder and one of very few treatable causes of dementia, most often also causing balance and urinary disturbances [
25]. Treatment by shunt surgery is effective with substantial clinical improvement in up to 80% of the patients [
4,
25,
43]. Health-related quality of life (HRQoL) is lower in iNPH patients than in age-matched healthy individuals [
19,
31], but improves in 86% of patients after surgery to almost the same level as in the normal population [
31].
The prevalence of iNPH is high. Between 0.5 and 2.9% of people older than 65 years of age suffer from iNPH [
5,
18,
40]. The disorder is underdiagnosed as well as undertreated; possibly only about 20% of affected patients undergo shunt surgery [
6,
40,
42].
Taken together with improved diagnostic methods and increased knowledge about the disorder, the number of patients in need of shunt surgery for iNPH will most likely increase. This will challenge the allocation of healthcare resources.
The cost benefit of shunt surgery in iNPH has been addressed in only a few studies. In a retrospective study of patients with hydrocephalus, the Medicare expenditure for treated patients was lower than for untreated patients [
41]. Based on data from the literature, the average 65-year-old patient receiving a shunt would gain an additional 1.7 QALYs [
36]. The caregiver burden (Zarit burden interview score) also decreases after a shunt operation [
22]. Kameda et al. [
20] recently reported that shunt surgery yields a positive return on investment within less than 2 years.
The cost effectiveness of treating iNPH has not been investigated in a prospective study with patient experienced QoL and cost data, and the aim of the present study was to investigate this from a societal perspective, using a decision-analytic Markov model adapted to Swedish circumstances.
Discussion
This study shows that shunt surgery is an inexpensive and cost-effective treatment in iNPH. The average iNPH patient gains 2.2 life years and 1.7 QALY, a substantial gain for a patient aged around 70 years. In comparison, Donezepil treatment in Alzheimer’s disease gives an estimated gain of 0.11 QALYs at a considerably higher incremental cost effectiveness ratio (ICER) [
12], whereas endovascular thrombectomy in acute stroke adds 0.4 life years and 0.99 QALY [
3].
The ICER for shunt surgery was €7,500/QALY and the cost per life year saved was €6,000. Even though we used a societal perspective and included other costs than healthcare costs, it is well below the UK National Institute for Health and Care Excellence (NICE) nominal healthcare cost-per-QALY threshold for new interventions. Thus, shunt surgery would most probably have been strongly recommended if introduced as a new intervention. Compared to other established surgical procedures, shunt surgery has a low extra cost per QALY gained, matching, for example, hip replacement surgery (Table
4). Our results corroborate with a recent report by Kameda et al. [
20] who also concluded that shunt surgery is cost effective, even though they found a higher cost/QALY (US$30,000–41,000/QALY equalling €28,000–39,000/QALY), probably due to a different study design with indirect estimations of costs and utility values.
Table 4
Extra cost per QALY gained for different established surgical procedures
Hip replacement, 1 year | 6,750 | |
Shunt surgery for iNPH | 7,500 | |
Knee replacement, 1 year | 14,000 | |
PCI for MI | 17,000 | |
DBS for Parkinson’s disease | 23,100 | |
Intractable partial epilepsy | 24,500 | |
This study was performed in a Swedish setting. We are, however, convinced that the results reported here are valid in developed countries throughout the world. The surgical procedure and the effects of shunting are comparable in most specialised centres. Also, costs for the “Deteriorated” state used here are within the range of the societal costs presented for other countries. The societal cost per person with dementia in Europe was estimated to US$24,565 (€22,168) and in North America to US$36,603 (€33,301) in 2009 years prices [
46]. A recent study from Sweden, estimated the annual cost per person with dementia to €43,259 [
1]. The yearly monetary cost in the United States per person with dementia was estimated to US$56,290 (€50,800) or US$41,689 (€37,620), depending on which method was used to value informal care [
17].
The patients in our study were consecutively included, presented typical clinical features, and 67% were improved on the iNPH scale, a slightly lower figure than in many contemporary studies [
43]. However, 83% reported improved HRQOL. The overall complication rate of 23% was somewhat high, but the revision rate of 13% is comparable with earlier studies [
11]. All patients were diagnosed based on typical clinical symptoms and MRI findings according to international guidelines [
34]. No supplementary tests, such as a cerebrospinal fluid (CSF) drainage test or lumbar infusion test, were used for inclusion. Inclusion of supplementary tests for the selection of patients for surgery has been discussed and recommended in several recent papers [
14,
23,
27‐
29]. These tests are widely used due to their ability to identify shunt responders and adding these tests probably increase response rate to shunt surgery. The lower success rate in this study could be an effect of not using these tests. However, the European multicentre study [
25] showed that when no supplementary tests are used, the number of patients improved by shunt surgery is 82%, which is similar to studies using such tests for inclusion. The dilemma of using or not using supplementary tests is further strengthened by the fact that iNPH patients that would benefit from surgery would be excluded by the use of supplementary tests as selection criteria [
25]. Altogether, we consider the patient sample to be representative and the reported results valid. It was also established that the shunts worked in all the participants.
The input parameters used in the model were derived from the recent literature. Results from the European multicentre iNPH study [
25], a large prospective longitudinal study with results comparable with those of many previous reports, and a recent systematic review of the outcome of shunt surgery in iNPH [
20] formed the basis for the calculation of model outcome probabilities. A 73% probability of being improved after shunt surgery was used, corresponding to outcomes after more than 3 years reported in studies since 2006 [
20], a figure that also corresponds to our experience and previous reports from our centre. Even if this figure is higher than recently reported in a large quality registry based study [
37], we believe it is reasonable. For complication probabilities, data from a previous study at our centre, showing a rate of 13% for complications in need of intervention during the first year and 2% thereafter, were used [
11]. These figures are comparable with other reports [
43] and higher than in a recent report on cost-effectiveness [
20] why we do not consider the complication rate underestimated. The mortality for the improved state was estimated according to the age-adjusted mortality rate for the normal Swedish population with an increase of 2.5 times. The same applies to the mortality for the deteriorated state, but with an additional 10% risk of mortality. We consider these input parameters to be the best possible estimates available. Also, the sensitivity analysis showed the robustness of the model—the results were not sensitive to major changes in long-term costs or utility weights—and, although some of the input parameters may be slightly uncertain, the main results will remain, even if the input parameters are varied.
The annual cost for the state “Deteriorated” was derived from the report by the National Board of Health and Welfare [
35], where 78% of the cost is related to home care and the cost of living in a nursing home. According to the report, 42% of patients with dementia live in nursing homes, which is higher than the rate reported here for patients with iNPH. Shunt surgery will reduce patients’ need for assisted living; furthermore, in our experience, arranging accommodation in a residential home is often neglected for patients with suspected iNPH at the time of diagnosis, probably due to poor knowledge of the disorder and its prognosis. The natural course of iNPH is sparsely studied, with few reports in the literature, and, to our knowledge, there are no reports of the care burden and the cost of untreated patients. Also, for ethical reasons, a prospective study of the natural course would hardly be acceptable, considering the effective treatment available. We consider the reported cost of dementia disorders in Sweden a good estimate of the cost of untreated iNPH patients.
No supplementary tests, such as the CSF tap-test or lumbar infusion test, were used when including patients in this study. Costs for these tests are very low compared to the surgical procedure and are probably compensated by the extensive clinical testing performed on all of our patients. Again, sensitivity analysis showed that the results were not sensitive to major changes in parameters. We consider the results reported here valid also for centres using supplementary tests.
We used a societal perspective in the analysis and included the cost of informal care provided by relatives. However, we did not include the caregivers’ utility weight in the model. A previous review has shown that self-reported health was strongly related to the time spent caring and to the carer’s perceived burden [
16]. Thus, since the cost of informal care is similar for the states “Improved” and “Deteriorated”, we assume that the burden for the caregivers is equal.
In general, simulation models have limitations when estimating cost effectiveness, often due to the lack of long-term data, resulting in major uncertainties of the long-term effects of the treatment in terms of costs, HRQoL and mortality. One weakness in our study is the small sample, mainly caused by the difficulty of undertaking such a detailed survey with personal interviews in larger samples. We chose to exclude seven patients aged <65 years and included only retired patients ≥65 years in order to get a patient group most representative of the target population. As stated earlier, we believe that the sample is representative, but further studies with larger samples are warranted. For this reason, results from other published studies have been included in the model, as probabilities of moving between the health states and the annual mortality rate. The uncertainties should be analysed with sensitivity analyses, as in this study.
One strength of the study is the rigorous recording of related costs after shunt surgery. All examinations were performed by a skilled occupational therapist (J.B.), either in connection with patient visits to the hospital or in the patient’s home. Relatives were always interviewed. Costs were also estimated through a meticulous evaluation of medical records. We believe that this objective procedure, evaluating qualitative and quantitative aspects of care, yields the most correct data possible. Further, HRQoL was measured using a reliable and widely used instrument.
Comments
The authors provide an important analysis on the cost effectiveness of shunt surgery in idiopathic NPH. There is a definite need for such studies and thus far little has been done to obtain more insights regarding NPH. Idiopathic NPH remains to be an enigmatic disorder even today. The most relevant paradigm shift in the concept of NPH over the last four decades, however, was that gait disorders are seen now as the primary symptom of NPH and not dementia. Therefore, it might be considered questionable to use reported costs of dementia disorders in Sweden as an estimate for the costs of untreated NPH patients. Furthermore, we do not know if nowadays NPH is still underdiagnosed and undertreated. It would be worthwhile to scrutinise this issue and to explore whether there was a change in awareness—or whether clinicians sometimes even tend to overdiagnose NPH now.
Joachim K. Krauss
Hannover, Germany