It has often been stated that the common methodology for clinical and cost-effectiveness studies is unsuitable for orphan diseases, due to the limited number of patients [
1,
39]. One recurring problem stemming from the low disease frequency is limited statistical power. A disease model can be used when long-term follow-up is unavailable. The concept of a disease model incorporates disease progression based on a combination of clinical plausibility and statistical analysis. Here we present a disease model for Pompe disease. Empirical data was used to test the relationships in this model. The model forms a starting point, to which information about treatment effects can be added in the future.
Purpose and validity of the model
We set out to describe the natural course of Pompe disease using an application of the Wilson-Cleary model. The appropriate structure of the model is disease dependent. For some disease progression models, classification of disease states is feasible (e.g. cancer, arthritis, multiple sclerosis, heart disease). Pompe disease is characterized by a continuous range of phenotypes, which makes identification of disease states difficult. Furthermore, accurately estimating parameter values for disease states is hampered by small sample sizes. Therefore, we applied the Wilson Cleary conceptual model rather than a model containing health states. Content validity (i.e. the validity of the associations that were modelled) of this model is ensured by the clinical experts that were involved in the development of the model and evidence from the literature. By comparing the results of the final model to findings reported in the literature construct validity can be assessed. Such comparisons reveal that the conceptual model is largely supported by the data since many of the results correspond with findings from previous studies. Disease duration is negatively associated with MRC sum score, in line with earlier findings [
21,
35,
36]. The rate of MRC sum score deterioration (0.6 % points per year) is comparable to our previous publications [
22,
38]. No significant association between disease duration and FVC was found (
p = 0.190), in contrast to findings described earlier [
20,
35,
36]. However, the relatively small though nonsignificant p-value we found in our analysis may be an indication of limited statistical power, a reduced ability to detect an existing association. Female patients have higher FVC than male patients, which is similar to the findings of our earlier study [
20]. Enzyme activity is not associated with the MRC sum score or FVC, which agrees with the findings of earlier studies that did not find any relationship between enzyme activity and severity in adult patients [
8,
36]. Being the underlying biological explanation of the disease, enzyme activity is retained in the conceptual model, despite the absence of a statistical significant association with other variables in the conceptual model.
The genetic mutation is the underlying cause of reduced enzyme activity. However, in our population most of the patients have the same mutation. Hence, mutation is left out of the model.
FSS was not associated with MRC sum score or FVC, albeit that FSS was significantly related to MRC at the 10 % level (
p = 0.070). Our previous studies did not find a relationship between fatigue and muscle strength and pulmonary function for adult patients with Pompe disease. While more severely affected patients (i.e. those who are in a wheelchair and/or ventilator dependent) do report higher FSS scores, almost all patients report fatigue [
24,
40]. For example, 71 % of patients who did not use respiratory support and/or wheelchair reported a FSS score of more than 4 (reflecting fatigue), while 59 % had an FSS score of more than 5 (reflecting severe fatigue). Also, even when patients are still in the pre-clinical stage of the disease they can report fatigue.
RHS is significantly affected by MRC sum score and FVC and better performance on these physiological variables are reflected by lower handicap scores. These findings resemble earlier results that showed that severely affected patients (that is, those requiring ambulatory and respiratory support) reported lower scores on RHS than other patients [
29]. The significant relationships indicate that functional status is not fully explained by fatigue, but by other problems related to reduced skeletal muscle strength and pulmonary function as well. One of these problems could be shortness of breath, which was absent in the model due to lack of data.
The conceptual model we present here embodies the current state of knowledge on Pompe disease in adult patients. Data that becomes available in the future can be used to fine-tune the relationships in the model. Moreover, new insights in the measurement of respiratory function might lead to the replacement of FVC by maximal inspiratory pressure (MIP) or maximal expiratory pressure (MEP). Inclusion of shortness of breath in new versions of the model might be reconsidered if new insights or analyses become available in future. Similarly, a newer version of the model can include once such data are available.
Strengths and applicability of the model
The availability of a (large) cohort of Pompe patients with several outcome measures enabled statistical testing of the conceptual model. Moreover, the dataset is relatively large compared to other orphan diseases, with available data on 79 patients and multiple observations for most patients. Although the current statistical analyses were based on Dutch patients only, there are no indications that the disease processes differ between the Dutch population and patients in other countries. Therefore, the underlying conceptual relationships described in the model will also be valid in other settings. The results described in this paper are also transferable for use in other countries since the measurement techniques (e.g., to describe muscle function) are widely used and not country-specific.
For many other orphan diseases, the number of patients is too small for a single institution or even a single country to assess a drug’s effectiveness. Data from international disease registries can potentially be used to explore relationships between clinically relevant levels of health concepts of an orphan disease. The testing of longitudinal relationships requires systematic follow-up, which might not yet be available in orphan disease registries.
The use of random effects models appears particularly beneficial in orphan diseases since it can compensate for the low number of patients to a certain extent. For this purpose, the availability of multiple observations per patient is indispensable.
Treatment
The assessment of a treatment effect is outside the scope of the current study. The analyses were deliberately restricted to untreated patients because we aimed to develop a model that describes the conceptual intercorrelations between the various levels of health concepts for Pompe disease. However, the conceptual model could be seen as a starting point for developing a full cost-effectiveness model to evaluate the cost-effectiveness of enzyme replacement therapy, assuming that the causal pathway of the disease will not change. Enzyme replacement therapy supplements the low levels of enzyme and has been shown to influence various levels of health concepts in the model, i.e. muscle strength and respiratory function, fatigue, functional health and quality of life, presumably through the supplementation of enzyme [
13,
22,
40‐
42]. If the model would in the future be used to estimate effects of ERT on health outcomes, this could be done in several ways. One way is to apply relative risk reductions to the outcomes as predicted by the equations in the model. These risk reductions would reflect the effect of ERT. Another approach is to re-estimate the equations with ERT as a predictor in the regression models. In both cases we would need longitudinal data on enzyme activity to estimate the impact of changes in enzyme activity on changes in health outcomes.
Limitations of the study
The number of observations and study period are relatively limited given the slow disease progression seen in Pompe disease. However, these limitations are common for studies of rare chronic diseases. This limitation could be rectified by international studies that collect relevant patient data in a standardized manner with a minimum loss to follow-up. Another limitation of our study was lack of data on shortness of breath, which could therefore not be included in the model.
Transferability of the model
This is the first study in which the Wilson-Cleary model is applied to an orphan disease. Although our model is specifically developed to describe and quantify the relationships between specific levels of health concepts of Pompe disease in adult patients, the approach used in this study can also be applied to other Pompe patient populations, i.e. children and infantile patients, or for other orphan diseases. Obviously, it is necessary to adapt the model to ensure that it contains only relevant disease-specific information and knowledge, although parts of the model developed here for Pompe disease in adult patients may still be relevant. To model Mucopolysaccharidosis type II for instance, “physiological factors” might be expressed using joint angle range and mental retardation can be used as a measure for “functional health”. As another example, Duchenne muscular dystrophy might be adequately represented using the current model if a factor for heart performance were to be included. However, in all cases, the application of the Wilson-Cleary model should be studied for each disease separately.