Read on Twitter
In response to my presentation at #SMDM20, @TMSnowsill has raised the issue of
treatments which extend life but provide zero health-related quality-of-life (HRQoL). How do the #QALY, #LY, #evLYG and #HYT deal with these? [Thread] @icer_review @Basucally
treatments which extend life but provide zero health-related quality-of-life (HRQoL). How do the #QALY, #LY, #evLYG and #HYT deal with these? [Thread] @icer_review @Basucally
In this thread I will also consider how each of these four approaches deals with treatments that extend a patient's life but with a *negative* HRQoL (i.e. in a state 'worse than dead'), and offer some thoughts on the implications of this for decision making.
As an illustrative example, suppose we initially have just two treatments, A and B. Treatment A provides good HRQoL (0.8) for 1 year, while treatment B provides slightly worse HRQoL (0.6) for 2 years. Let's begin by considering how each approach values treatments A and B.
The #QALY assigns greater value to treatment B than to treatment A, since 2 years valued at 0.6 each (1.2 QALYs) is greater than 1 year valued at 0.8 (0.8 QALYs).
The #LY approach also assigns greater value to treatment B, since it offers longer life expectancy (2 years) than treatment A (1 year). Note that HRQoL is irrelevant under the #LY approach.
The #evLYG approach similarly assigns greater value to treatment B. When compared to 'current treatment' A, the additional year of life expectancy with B is valued in full, while the first year is valued at the HRQoL (0.6), for a total of 1.6 evLYG. The value of A is 0.8 evLYG.
Finally, the #HYT also assigns greater value to treatment B. For B, 2 life years are added to 1.2 QALYs, for a total of 3.2 HYT. For A, 1 life year is added to 0.8 QALYs, and then 0.8 'post-death QALYs' are added (since life expectancy with B is 2 years), for a total of 2.6 HYT.
As a result, when we compare A and B alone, all four approaches favour treatment B. Now let's consider what happens if we introduce treatment C, which provides 6 years of life expectancy but with a HRQoL of zero.
Before we continue, it's worth bearing in mind that there are several different ways that a HRQoL of zero might arise in someone still alive. One would be if a patient is not sufficiently conscious to experience any 'health-related quality' to their life, such as being in a coma.
Another would be if a patient's baseline HRQoL was 'worse than death' and treatment then improved this to zero. A third would be if a patient's baseline HRQoL was positive but treatment toxicity reduced this to zero. These are just a few examples - I'm sure there are many more.
So, how does each of these four approaches respond to us adding treatment C into the comparison?
Let's start with the #QALY. Since treatment C provides zero HRQoL, it's assigned no value regardless of its life expectancy. Treatment B remains the most desirable treatment, followed by treatment A.
The #LY favours switching to treatment C. Each year with C is valued the same as each year with treatment A or B, even though A and B provide positive HRQoL to patients and C does not. Treatment C provides for the longest life, which is all that matters under the #LY approach.
The #evLYG also favours switching to treatment C. The 4 additional years of life expectancy with treatment C compared to treatment B are valued in full, while the first 2 years are valued using QALYs (but are assigned zero value, since HRQoL is zero), resulting in 4.0 evLYG.
However, the #evLYG is dynamically inconsistent. If this comparison were repeated in future after treatment C has been established as 'current care' (perhaps in response to a new treatment D), treatment C would be valued using QALYs only and the #evLYG would rank B higher than C.
Finally, the #HYT favours treatment C. Although its HYT comprise LYs only, these are sufficient to outweigh the total LYs, QALYs and 'post-death QALYs' assigned to A and B. Also, treatment A is now preferred to B, due to the change in the 'reference' strategy to treatment C.
In summary, only the #QALY necessarily favours a treatment that provides positive HRQoL to patients over those that do not. As this example shows, the other three approaches ( #LY, #evLYG and #HYT) can favour a treatment with a HRQoL of zero over treatments with positive HRQoL.
A related issue is that of treatments with *negative* HRQoL. This is a controversial subject: see @ChrisSampson87's research on 'states worse than death' and @waq0r's recent paper on issues that can arise with the #QALY https://bitowaqr.github.io/files/qalyisableist.pdf
Yet it is worth noting that two of these approaches ( #LY and #evLYG) would favour treatment C even if it had a substantially *negative* HRQoL. The #HYT can also favour treatments with negative HRQoL if life expectancy is sufficiently long (since the LYs overwhelm the QALY loss).
This issue does not arise with the #QALY. For all its faults, there are no circumstances under which the #QALY favours prolonging a patient's suffering in a state of extreme pain or other unbearably poor health over a treatment which provides for a shorter life of good health.
The reality is, some states of health are considered unbearably poor by many people. Yet the #LY, #evLYG and #HYT impose a value of 1 on these states somewhere within their construction (the #LY by default, the #evLYG during life extension, and the #HYT within the LY component).
