Purpose: Evaluation of the cost-effectiveness of treatment for chronic disease requires development and validation of a model of the progression of disease validated using “real world” data. We constructed a Markov model using three large prospective studies of glaucoma treatment and evaluated the model’s predictive validity as a first step towards calibration.
Method: Glaucoma severity and disease progression were defined clinically in terms of visual field loss expressed as mean deviation (MD) measured in decibels (dB) quantified by automated perimetry. Patient level data for the model came from the Collaborative Initial Glaucoma Treatment Study (n=607), the Ocular Hypertension Treatment Study (n=1,636), and the Advanced Glaucoma Intervention Study (n=560). Due to the length of these studies our initial model was limited to the pattern of progression over seven years. Transition probabilities for this Markov model were calculated for each combination of year and MD. The model was estimated with TreeAge software using a microsimulation approach. Validation was conducted by comparing the predicted value of hypothetical participants to that of the actual study participants. For this purpose, a clinically significant difference was considered to be 2 decibels (dB) of MD.
Result: At year 7, the predicted value for the right eye of 79.8% (n=1,152) of hypothetical participants were within 2 dB of their “real world” counterpart, as was the left eye in 82.7% (n=1,194). Of those that were outside of the 2 dB band, over 85% had worse disease at year 7 than was predicted by the model. On average these participants were older at baseline (i.e., 64.0 versus 57.7, p < 0.05), and had more advanced disease than those within the 2 dB band.
Conclusion: We have developed a model for glaucoma progression that is based upon individual level data gathered from three large clinical trials. Our initial results indicate that the model properly predicts the result of disease progression in over 80% of “participants”. This suggests that our modeling approach provides a reasonable reflection of real world progression and provides a useful tool for researchers and policy makers. Further refinement and calibration of the model will include development of age specific transition probabilities, as well as consideration of the acceleration of disease progression among participants with more advanced disease at the time of diagnosis.
Candidate for the Lee B. Lusted Student Prize Competition