Prediction Of Type Diabetes A Hypothetical Example

Numerous recent studies have illustrated how combined marker information enhances the prediction of Type 1 diabetes (58), particularly in connection with preventive trials (48). The approach is most conveniently illustrated by a numerical example. In this section we use a positive family history of Type 1 diabetes (FH) as the first step marker which is combined with the presence of a genetic susceptibility marker (GM) as the second step marker. We apply hypothetical data which we, however, consider to be fairly representative for a country such as England with well-established traditions in epidemiological and clinical diabetology (48,54,55, 56,58).

First of all, we consider a population of unaffected children aged 0-14 years. For convenience, the population size will be fixed at 1 000 000 subjects which we assume to follow during a period of 5 years. The incidence of Type 1 diabetes (ignoring marker status) may be set at 0.00016 per person-year at risk, corresponding with an absolute disease risk at 0.08% over 5 years. Secondly, we assume that 1.0% of the subjects in this population has a positive family history (FH) and that subjects with positive FH have a risk of 0.88%; this is plausible considering the population incidence level. A given genetic susceptibility marker (GM) has a population prevalence of 0.02 (2.0%), which is of the order of magnitude of the frequency of one of the currently defined genotypes conferring high risk of Type 1 diabetes. However, within the FH-positive subjects the prevalence of GM is assumed to be 25% due to the prior probability of sharing any given genotypes with a sibling. The presence of GM is assumed to confer an absolute risk of 1.0% over 5 years in FH-negative subjects; this agrees with estimated absolute risks at 5 -10% over an extended period for unrelated subjects with a high-risk genetic marker of Type 1 diabetes (as mentioned before). However, in FH-positive subjects it is assumed that the disease risk over 5 years is 2% due to the effects of sharing additional risk factors with affected relatives. Under these assumptions the population distribution according to this dual marker system will be realized as shown in Figure 7B.2. The absolute risks as obtained from assumptions and implications are also shown, together with the corresponding estimated number of new cases of Type 1 diabetes over a period of 5 years.

The most important measures of test performance are presented in Table 7B.2. The positive predictive values (and, thus, the marker-specific absolute risks of developing Type 1 diabetes) are given by assumptions or implications in this

Figure 7B.2 Distribution according to positive family history (FH) and presence of a genetic marker (GM) in a population of 1 000 000 children. Assumptions: see text

example, but are realistic for reasons given before. Their low level, even for the combined markerpositive category, indicates that the large majority of marker-positive subjects will remain disease-free for a substantial period of time. Nevertheless, these subjects represent the target group for potential prevention of Type 1 diabetes. Most importantly, this dual screening strategy, which in the first step is based on anamnestic information only, has allowed for the classification of more than 97% of the whole population as belonging to a very low-risk category. In spite of this, the estimated values of sensitivity (SENS) in Table 7B.2 indicate that the majority of new cases of Type 1 diabetes will not be predicted by these markers, neither when applied as single markers nor when combined.

It must be stressed that hypothetical examples like the one above are very sensitive to changes in the underlying assumptions which, for reasons mentioned earlier, most likely will differ from population to population. Until more precise population-based assessments of various strategies

Table 7B.2 Positive predictive value (PPV), sensitivity (SENS), specificity (SPEC) and relative risks (RR) for markers in the prediction of Type 1 diabetes

Marker

PPV

SENS

SPEC

RR

(%)

(%)

(%)

Single markers:

FH alone

0.88 *

10.94

99.01

12.16

GM alone

1.13 *

28.13

98.02

19.17

Markers combined:

FH and GM

2.00 *

6.25

99.75

36.19**

FH: Positive family history GM: Genetic marker Hypothetical data.

* Assumed values, expressed as estimated cumulative risk of Type 1 diabetes over 5 years.

* * Expressed relative to being negative for both markers.

FH: Positive family history GM: Genetic marker Hypothetical data.

* Assumed values, expressed as estimated cumulative risk of Type 1 diabetes over 5 years.

* * Expressed relative to being negative for both markers.

in the prediction of Type 1 diabetes are available, different sets of assumptions should be explored in corresponding scenarios. The principles and methods illustrated above may be useful for such purposes. The methods may be further refined by stratification within a marker category, e.g. distinguishing between subjects at low versus high ICA-titer.

Diabetes Sustenance

Diabetes Sustenance

Get All The Support And Guidance You Need To Be A Success At Dealing With Diabetes The Healthy Way. This Book Is One Of The Most Valuable Resources In The World When It Comes To Learning How Nutritional Supplements Can Control Sugar Levels.

Get My Free Ebook


Post a comment