Pages within the "support" section:

14.9 Severity of effect (disease weights)

DALY or QALY weights are available in the literature (e.g. WHO). In many cases a single weight is given for each disease or health state, although in reality the severity of a disease may differ depending whether the individual eventually recovers, dies from the disease, or continues with the disease until their normal life expectancy. The Qalibra framework provides the option to represent this, using different weights for different stages of the disease, if these are available in the published literature or if the user wishes to explore the effect of different assumptions. It is also possible that disease severity may depend on personal attributes such as age or sex, or on the intake of the substance causing the effect. The weights should represent the average severity of the effect for individuals with the relevant set of attributes.

It is conceivable that disease severity may depend on the cause of the disease, in which case the weights published for the general population may differ from those that would apply when the disease is caused by the dietary change under assessment. Another important source of uncertainty is introduced if the dose-response for the effect is derived from animal studies, because it is then necessary to make an assumption about which human diseases the effect corresponds to, in order to make use of human disease weights. For continuous effects, published weights may be presented as step functions of the effect (e.g. IQ), whereas in reality the severity of the effect is more likely to be a continuous function of effect size. The impact of all identified uncertainties should be considered when interpreting the results of the assessment (see section 15.1).

Disease weights for QALYs are usually measured with standardised questionnaires like EQ-5D, SF-6D, HUI3 etc. in which generic health states are valued. As is to be expected different measurement instruments (questionnaires) will show different values (e.g. (Fryback et al.   2007) . Kopec and Willison (2003) show that depending on the measurement method disease weights (QALY) can differ substantially for the same disease e.g. from 0.26 to 0.71 (depression) or from 0.58 to 0.72 for arthritis. Also different populations value health states differently (Havranek and Steiner, 2005) .

Schwarzinger et al. ( 2003) show differences in disability weights (DALY) depending on the population and on the method with which weights are measured. Disability weights for stroke e.g. differ between 0.17-0.68. Differences also exist when patients, the general public or health practitioners are asked (De Wit et al.   2000, Hoekstra et al.   200 8 , Schwarzinger et al.   2003) . Essink-Bot et al. ( 2007) conclude that health state valuations may be sensitive to individual response patterns that could not be explained by characteristics such as age, sex or educational level. (Fryback et al.   2007) points out that the disease weights of the one who takes the decision should be used and that disease weights and consequently the valuation of health states is only a part of the general societal utility function.

The use of DALYs and especially age weighting and discounting is an area of open debate (e.g.   Arnesen and Kapiriri, 2004; Murray and Lopez, 2000; Paalman et al.   1998; Murray and Acharya, 1997) . Murray and Lopez (1996) published discounted as well as non-discounted values because of the difficulties of choosing a discount rate.

In the light of these issues, it is clearly important to maximise transparency when common health measures used (Arnesen and Kapiriri, 2004) . Care should be taken that disease weights are relevant for the population of interest. Furthermore, each disease should be measured with the same instrument and the group valuing the disease should be similar also, not the general public for one disease and patients for the other.

When animal experiments are performed with toxic substances, endpoints are measured that have no direct clinical endpoint. Sometimes this is also true for human epidemiological studies. Examples are sperm counts in rats that are exposed to dioxins or IQ tests for children whose mother has been exposed to methyl mercury via fish consumption. In order to compare these effects with other (beneficial) health effects they need to be converted in a common health measure such as DALYs. Obviously, this cannot be done without crude assumptions and the introduction of large uncertainties that would be difficult to quantify. Presumably, the measured endpoint will have some relation with human health (if it does not, there is no need to include it in the assessment). Because the conversion to a clinical endpoint for which DALY or QALY weights exist is not straightforward the conversion needs to be explained and sensitivity analysis must be performed to investigate the influence of the assumptions on the final results. The assumptions have to be made clear and the argumentation must be given in a narrative.