Rhenus:Validation 2008 (Rhenus-A-0.11-GER)
From Prosit
This page contains development contents which may not yet represent final PROSIT Content.
Contents |
[edit] External Validation
[edit] Comparison of the model structure with other models
In this section we compare structures of published diabetes models with the structure of Rhenus (incl. purpose of a model).
[edit] Zhou et al, Diabetes Care, 2005
The numbers at the transition arrows are the references to the publications used in the Zhou et al model. The transition probabilities in the Zhou et al model were derived from the results described in these publications. Please have a look at these publications for further details.
Obviously, there are no differences in the model structure by Zhou et al compared to the structure of Rhenus A.
The retinopathy model by Zhou et al consists of exactly the same disease states. Even the transitions from one state into another are realized in the same way as this is done in Rhenus A.
The only marginal difference is the mortality model. In the Zhou et al model exists a separate mortality model, which concerns mortality from non-diabetic reasons. In the retinopathy part of the Zhou et al model there are no transitions to a death state.
[edit] Marshall et al, Stat Med, 2005
Reference:
The multi-state model consists of three transient states that represent the early stages of retinopathy, and one final absorbing state that represents the irreversible stage of retinopathy. The values at each arrow are the related values from the appropriate transition matrix.
The transitions from one state to another are according to a Markow Chain, so only the transition from one state j to the adjoining states j-1 or j+1 is possible, apart from the last
In comparison to the Rhenus model the Marshall G, Jones RH model consists of four states standing for six different stages of Retinopathy. The stages are in dependence on the Airlie House classification. The difficulty is based on the difference between these two models. The Marshall G, Jones RH model is a multistate, technical based model where the transition probabilities were not calculated from fixed time periods. But that is indispensable for a validation with the Rhenus model. Finally an analogy to the Rhenus model is hardly detectable.
[edit] Sally C. Brailsford, Health Care Manag Sci., 1998
The model simulates the progress of Retinopathy in a population of non-insulin-dependent patients. Only the transition probabilities of the two types of retinopathy - background diabetic retinopathy (BDR) via the proliferative form (PDR) to blindness and diabetic macular oedema (DMO) via the clinically significant stage (CSMO) to blindness - are given in cause of the assumed independence. But these two types aren't exclusive processes so there are also other stages possible as shown in the bubble diagram.
The differences between this model and the Rhenus model are: The split of the two types of Retinopathy and the existence of the mortality status.
[edit] CDC Diabetes Cost-effectiveness Group, JAMA, 2002
This Markov model structure placed greater emphasis on macrovascular complications and introduced interdependencies among diabetes progression paths.
Three states of the Rhenus A Model are summarized as the element Photocoagulation in the CDC Model, which is without Death state. There are two types of photocoagulation:
- The first type is called focal photocoagulation. This procedure is usually effective for Non-proliferative retinopathy and macular edema.
- The second procedure is called scatter or panretinal photocoagulation. It is used to treat proliferative retinopathy.
[edit] The CDC Diabetes Cost-Effectiveness Study Group, JAMA, 1998
There are only 2 states,which will cause the blindness in the Health States Model,but one more state of Rhenus A Model can cause the blindness and all states outside death state can possibly keep the particular status.
The PR can cause the DME too in the Rhenus A Model,but not in the Health States Model.
[edit] Vijan et al, Jama, 2000
The intention of designing this Markov cost-effectiveness Model was to examine the marginal cost-effectiveness of various screening intervals for eye disease (especially Diabetic Retinopathy) in patients with type 2 diabetes, stratified by age and level of glycemic control.
This Model is based on hypothetical patients in the US population of diabetic patients older than 40 years from the Third National Health and Nutrition Examination Survey.
Mortality rates A, B, C, D and E were based on life tables published by the US government [see reference 37 in the paper] and modified to reflect the increased mortality rates observed in patients with diabetes in general [see references 26-30 in the paper].
[edit] Comparison of transition probabilites with other models
This section provides a literature comparison with methods and transition probabilities of other published models.
| Model | p1 | p2 | p3 | p4 | p5 | p6 | p7 | p8 | p9 | p10 | p11 | p12 | p13 | p14 | p15 | p16 | p17 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Rhenus A | n.a. | Klein R et al (1989) | n.a. | Klein R et al (1998) | Klein R et al (1995) | Moss SE et al (1998) | Roper NA et al (2002), Federal Statistic Office (2005) | n.a. | Klein R et al (1995) | Moss SE et al (1998) | Roper NA et al (2002), Federal Statistic Office (2005) | n.a. | Moss SE et al (1998) | Roper NA et al (2002), Federal Statistic Office (2005) | n.a. | Roper NA et al (2002), Federal Statistic Office (2005) | Roper NA et al (2002), Federal Statistic Office (2005) |
| Zhou et al | n.a. | Klein R et al (1989), Klein R et al (1994) | n.a. | Klein R et al (1989), Klein R et al (1994) | Klein R et al (1995) | Moss SE et al (1988), Moss SE et al (1994) | n.a. | n.a. | Klein R et al (1995) | Moss SE et al (1988), Moss SE et al (1994) | n.a | n.a. | Moss SE et al (1988), Moss SE et al (1994) | n.a. | n.a. | n.a. | n.a. |
| Sally C. Brailsford | n.a. | E. Agardh et al., (1993), R. Klein, B.E.K. Klein and S.E. Moss, (1990) | n.a. | R. Klein et al., (1995) | n.a. | n.a. | n.a. | M.S. Dwyer et al., (1985) | n.a. | n.a. | n.a | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. |
| Vijan et al, Jama, 2000 | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. | n.a. | Plan and Operation of the Third Nation Health and Nutrition Examination Survey 1988-94. Series 1:Programs and Collection Procedures Morrish NJ et al (1990) Panzram G. (1987) Muggeo M (1995) Panzram G et al (1981) Krolewski AS et al (1981) | Plan and Operation of the Third Nation Health and Nutrition Examination Survey 1988-94. Series 1:Programs and Collection Procedures Morrish NJ et al (1990) Panzram G. (1987) Muggeo M (1995) Panzram G et al (1981) Krolewski AS et al (1981) | n.a. | n.a. | n.a. | n.a. | n.a. |
| The CDC Diabetes Cost-Effectiveness Study Group | n.a. | Eastman et al (1997) | n.a. | Eastman et al (1997) | n.a. | Eastman et al (1997) | n.a. | Eastman et al (1997) | n.a. | n.a. | n.a | n.a. | Eastman et al (1997) | n.a. | n.a. | n.a. | n.a. |
[edit] Cross-Testing
- Cross-testing of diabetes disease models
[edit] Model outcomes
Target: Compare the results of the Rhenus model to those of clinical long-term studies.
[edit] Comparison 1: Rhenus A with ROSSO 4
Comparison of the the model prediction of the Rhenus A model with the results of the Retrolective Study Self- Monitoring of Blood Glucose and Outcome in Patients with Type 2 Diabetes (ROSSO 4 [1]). In the ROSSO 4 study the further development of diabetes was monitored retrospectively from diagnosis of type 2 diabetes. Data for more than 3000 patient files of 192 GP practices in Germany were collected.
Patient characteristics
- Cohort: 3142 persons
- Sex: 51,4% female
- Initial age: 62.5 years
- Blood pressure: 149.20 mmHg (systolic) / 87.00 mmHg (diastolic)
- HbA1c: 7.6%
- Initial amount of blind patients: 0.2%
- No Retinopathy: 99.8%
- BMI: 29,8
- Duration of diabetes: unknown [DEFAULT value 10 years was used in the model]
Table 1: Blindness in patient cohort over time
| Year | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|
| ROSSO 4 | 0.2% | 0.4% | 0.4% | 0.5% | 0.5% | 0.6% | 0.6% | 0.6% |
| Rhenus A | 0.2% | 0.2% | 0.2% | 0.2% | 0.3% | 0.4% | 0.5% | 0.6% |
Figure 1: Blindness in patient cohort over time
Table 2: Death in patient cohort over time
| Year | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 |
|---|---|---|---|---|---|---|---|---|
| ROSSO 4 | 0.0% | 0.3% | 0.7% | 1.3% | 2.1% | 2.7% | 3.1% | 3.5% |
| Rhenus A | 0.0% | 2.2% | 4.5% | 7.0% | 9.8% | 12.6% | 15.7% | 19.0% |
Figure 2: Death in patient cohort over time
[edit] Comparison 2: Rhenus A with Osaka, Japan
Comparison of Rhenus A with the following study: "A 15 year follow-up study of patients with NIDDM in Osaka, Japan. Long Term prognosis and causes of death" [2]
Patient characteristics
- Cohort: 1939 persons
- Sex: 34.16% female
- Initial age: 60.98 years
- Blood pressure: 146.90 mmHg (systolic) / 84.40 mmHg (diastolic)
- HbA1c: 8.5% [DEFAULT]
- Initial amount of non-proliferative retinopathy: 29.95%
- No Retinopathy: 70.05%
- BMI: 22.5 [DEFAULT]
- Duration of diabetes: 10 years [DEFAULT]
For all other parameters the default values of the model werd used.
Table 3: Death in patient cohort over time
| Year | 0 | 1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | 14 | 15 |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Osaka, Japan | 0.0% | - | - | - | - | - | - | - | - | - | - | - | - | - | - | 45,4% |
| Rhenus A | 0.0% | 1.7% | 3.6% | 5.6% | 7.8% | 10.1% | 12.6% | 15.4% | 18.2% | 21.3% | 24.6% | 28.1% | 31.7% | 35.5% | 39.5% | 43.6% |
Figure 3: Death in patient cohort over time
[edit] Comparison 3: Rhenus A with BES
Comparison of Rhenus A with the study: "Incidence of Diabetic Retinopathy in the Barbados Eye Studies" [3]
To examine the 4-year incidence and risk factors for diabetic retinopathy(DR) among black participants with diabetes in the Barbados Eye Studies(BES).
Patient characteristics
- Cohort: 410 persons
- Sex: 63.7% female
- Initial age: 60.98 years
- Blood pressure: 139.90 mmHg (systolic) / 81.8mmHg (diastolic)
- HbA1c: 10.8%
- No Retinopathy: 74.63%
- BMI: 22.5 [DEFAULT]
- Duration of diabetes: 6.5 years
For all other parameters the default values of the model werd used.
Table 3:
| Year | 0 | 1 | 2 | 3 | 4 |
|---|---|---|---|---|---|
| BES | 74.6% | - | - | - | 47.8% |
| Rhenus A | 74.6% | 68.8% | 63.3% | 58.0% | 53.1% |
Figure 4: No Retinopathy in patient cohort over time
[edit] Internal Validation
[edit] Input
Medical evidence:
[edit] Medicine
The Rhenus model includes the costs for the following drugs ...
- Calcium dobesilat
- Somatostatinanalogon, Octreotid
- Proteinkinase-C-Inhibitor
- Triamcinolonacetonid
[edit] Cohort characteristics
The model offers the following cohort characteristics input parameters:
- initial age (in years) - must be between 45 and 75 years!
- distribution between the sexes (in %)
- duration of diabetes (in years)
- smoker/ non-smoker
- HbA1c level (in %)
- blood pressure (in mmgH)
- proteinuria/ no proteinuria (in %)
- BMI (in kg/sqm)
Economic evidence: cost data set, discount factors, inflation rates
[edit] Model
[edit] Output
[edit] Face Validity
Behaviour of the finished model with regard to model outcomes:
- Extrem value analysis
- Sensitivity Analysis
- Interfaces & model handling: * Workflow logic, independence of user, reproducibility of results
[edit] Sensitivity Analysis
For the sensitivity analysis the values, given in the two graphics, were modified about +/- 10 percent. Only the model's input variables were changed in cause of the sensitivity analysis' first step in the Rhenus Beta project. The effect of each value on "Disounted costs" and "Quality adjusted life expectancy" was registered and evaluated in a Tornado Diagram shown below.
