scispace - formally typeset
Search or ask a question
Journal ArticleDOI

Indirect estimation of a discrete-state discrete-time model using secondary data analysis of regression data.

20 Jul 2009-Statistics in Medicine (John Wiley & Sons, Ltd.)-Vol. 28, Iss: 16, pp 2095-2115
TL;DR: This paper presents an approach that allows the use of published regression data in a multi- state model when the published study may have ignored intermediary states in the multi-state model, called the Lemonade Method.
Abstract: Multi-state models of chronic disease are becoming increasingly important in medical research to describe the progression of complicated diseases. However, studies seldom observe health outcomes over long time periods. Therefore, current clinical research focuses on the secondary data analysis of the published literature to estimate a single transition probability within the entire model. Unfortunately, there are many difficulties when using secondary data, especially since the states and transitions of published studies may not be consistent with the proposed multi-state model. Early approaches to reconciling published studies with the theoretical framework of a multi-state model have been limited to data available as cumulative counts of progression. This paper presents an approach that allows the use of published regression data in a multi-state model when the published study may have ignored intermediary states in the multi-state model. Colloquially, we call this approach the Lemonade Method since when study data give you lemons, make lemonade. The approach uses maximum likelihood estimation. An example is provided for the progression of heart disease in people with diabetes.

Content maybe subject to copyright    Report

Citations
More filters
Journal ArticleDOI
01 Oct 2019
TL;DR: This paper demonstrates through several abstract examples how microsimulation disease models can be encoded using the SBML Arrays package, enabling reproducible disease modeling.
Abstract: Disease modelers have been modeling progression of diseases for several decades using such tools as Markov models or microsimulation. However, they need to address a serious challenge; many models ...

5 citations

Journal ArticleDOI
01 Nov 2016
TL;DR: The previously published maximum likelihood estimation technique to estimate model parameters from indirect secondary data is extended, and a selection method is demonstrated for picking a preferred model according to likelihood and structure criteria.
Abstract: The progression of a disease may be affected by many risk factors, such as gender, age, and current disease state. Such information is collected and made publically available by published clinical studies, yet combining this information into a disease model remains a challenge. This paper extends the previously published maximum likelihood estimation technique to estimate model parameters from indirect secondary data. Such information is available in the scientific literature so the modeler can access more data when estimating model parameters. The extension to the estimation procedure allows model transitions that depend on different sets of covariates for which secondary data are available. This extension uses a Markov model with transition probabilities stored in multi-dimensional tables accessed by covariate values. The paper uses a set of cases, including a case of cardiovascular disease in diabetes. The cases demonstrate the proposed method with various model variations. To help cope with model multiplicity, a selection method is demonstrated for picking a preferred model according to likelihood and structure criteria.

2 citations


Cites background or methods from "Indirect estimation of a discrete-s..."

  • ...It is equivalent to the example published in Section 3.1 of Isaman et al.2 It was presented here again to deal with the multi-dimensional aspect that was not previously discussed, for the sake of completeness....

    [...]

  • ...The details of how the expected values x and the associated covariance matrix C are calculated is beyond the scope of this paper and are addressed in detail by Isaman et al.(2) In 01 p = 01 p (No table – implemented as a single cell table)...

    [...]

  • ...1 of Isaman et al.(2) It was presented here again to deal with the multi-dimensional aspect that was not previously discussed, for the sake of completeness....

    [...]

  • ...The data for study B2 can be used as incident counts or as the UKPDS 5618 risk engine equation, as described in more detail by Isaman et al.2 The tables that appear in the categorical studies are based on gender in study E, and on gender and age in studies F1 and G. Multiple time outcomes are provided by study F1....

    [...]

  • ...Scenario #10 uses the simplest base model with incident counts equivalent to the results published by Ye et al.,3 whereas scenario #20 uses the simplest base model with the regression estimates equivalent to results published by Isaman et al.2 Also note that models #14–#17 and #34– #37 (or #24–#27 and #44–#47) are distinguished by the use of gender versus age as a decision covariate in p34....

    [...]

Journal ArticleDOI
TL;DR: A robust, long-term clinical prediction model to predict conditions leading to early diabetes using laboratory values other than blood glucose and insulin levels is developed and HDL-C levels were useful for prediction of increases in HbA1c.
Abstract: Objectives: We developed a robust, long-term clinical prediction model to predict conditions leading to early diabetes using laboratory values other than blood glucose and insulin levels. Our model protects against missing data and noise that occur during long-term analysis. Methods: Results of a 75-g oral glucose tolerance test (OGTT) were divided into three groups: diabetes, impaired glucose tolerance (IGT), and normal (n = 114, 235, and 325, respectively). For glucose metabolic and lipid metabolic parameters, near 30-day mean values and 10-year integrated values were compared. The relation between high-density lipoprotein cholesterol (HDL-C) and variations in HbA1c was analyzed in 158 patients. We also constructed a state space model consisting of an observation model (HDL-C and HbA1c) and an internal model (disorders of lipid metabolism and glucose metabolism) and applied this model to 116 cases. Results: The root mean square error between the observed HbA1c and predicted HbA1c was 0.25. Conclusions: In the observation model, HDL-C levels were useful for prediction of increases in HbA1c. Even with numerous missing values over time, as occurs in clinical practice, clinically valid predictions can be made using this state space model.

1 citations

Dissertation
01 Jan 2018
References
More filters
Journal ArticleDOI
TL;DR: This paper examines eight published reviews each reporting results from several related trials in order to evaluate the efficacy of a certain treatment for a specified medical condition and suggests a simple noniterative procedure for characterizing the distribution of treatment effects in a series of studies.

33,234 citations


"Indirect estimation of a discrete-s..." refers background in this paper

  • ...likelihood function of unknown parameters k(k) is Normal(k̂(k),R(k)) [17, 18]....

    [...]

Journal Article
TL;DR: In this article, the effects of intensive blood-glucose control with either sulphonylurea or insulin and conventional treatment on the risk of microvascular and macrovascular complications in patients with type 2 diabetes in a randomised controlled trial were compared.

17,108 citations

Journal Article
TL;DR: The effects of intensive blood-glucose control with either sulphonylurea or insulin and conventional treatment on the risk of microvascular and macrovascular complications in patients with type 2 diabetes in a randomised controlled trial were compared.

8,546 citations

22 Sep 1998
TL;DR: The effects of intensive blood-glucose control with either sulphonylurea or insulin and conventional treatment on the risk of microvascular and macrovascular complications in patients with type 2 diabetes in a randomised controlled trial were compared.
Abstract: BACKGROUND Improved blood-glucose control decreases the progression of diabetic microvascular disease, but the effect on macrovascular complications is unknown. There is concern that sulphonylureas may increase cardiovascular mortality in patients with type 2 diabetes and that high insulin concentrations may enhance atheroma formation. We compared the effects of intensive blood-glucose control with either sulphonylurea or insulin and conventional treatment on the risk of microvascular and macrovascular complications in patients with type 2 diabetes in a randomised controlled trial. METHODS 3867 newly diagnosed patients with type 2 diabetes, median age 54 years (IQR 48-60 years), who after 3 months' diet treatment had a mean of two fasting plasma glucose (FPG) concentrations of 6.1-15.0 mmol/L were randomly assigned intensive policy with a sulphonylurea (chlorpropamide, glibenclamide, or glipizide) or with insulin, or conventional policy with diet. The aim in the intensive group was FPG less than 6 mmol/L. In the conventional group, the aim was the best achievable FPG with diet alone; drugs were added only if there were hyperglycaemic symptoms or FPG greater than 15 mmol/L. Three aggregate endpoints were used to assess differences between conventional and intensive treatment: any diabetes-related endpoint (sudden death, death from hyperglycaemia or hypoglycaemia, fatal or non-fatal myocardial infarction, angina, heart failure, stroke, renal failure, amputation [of at least one digit], vitreous haemorrhage, retinopathy requiring photocoagulation, blindness in one eye, or cataract extraction); diabetes-related death (death from myocardial infarction, stroke, peripheral vascular disease, renal disease, hyperglycaemia or hypoglycaemia, and sudden death); all-cause mortality. Single clinical endpoints and surrogate subclinical endpoints were also assessed. All analyses were by intention to treat and frequency of hypoglycaemia was also analysed by actual therapy. FINDINGS Over 10 years, haemoglobin A1c (HbA1c) was 7.0% (6.2-8.2) in the intensive group compared with 7.9% (6.9-8.8) in the conventional group--an 11% reduction. There was no difference in HbA1c among agents in the intensive group. Compared with the conventional group, the risk in the intensive group was 12% lower (95% CI 1-21, p=0.029) for any diabetes-related endpoint; 10% lower (-11 to 27, p=0.34) for any diabetes-related death; and 6% lower (-10 to 20, p=0.44) for all-cause mortality. Most of the risk reduction in the any diabetes-related aggregate endpoint was due to a 25% risk reduction (7-40, p=0.0099) in microvascular endpoints, including the need for retinal photocoagulation. There was no difference for any of the three aggregate endpoints between the three intensive agents (chlorpropamide, glibenclamide, or insulin). Patients in the intensive group had more hypoglycaemic episodes than those in the conventional group on both types of analysis (both p<0.0001). The rates of major hypoglycaemic episodes per year were 0.7% with conventional treatment, 1.0% with chlorpropamide, 1.4% with glibenclamide, and 1.8% with insulin. Weight gain was significantly higher in the intensive group (mean 2.9 kg) than in the conventional group (p<0.001), and patients assigned insulin had a greater gain in weight (4.0 kg) than those assigned chlorpropamide (2.6 kg) or glibenclamide (1.7 kg). INTERPRETATION Intensive blood-glucose control by either sulphonylureas or insulin substantially decreases the risk of microvascular complications, but not macrovascular disease, in patients with type 2 diabetes.(ABSTRACT TRUNCATED)

7,252 citations