A cost effectiveness of the PACE trial

[Bob]

They found the average improvement for the SMC group over baseline was 0.02 QALYs. This is based on taking measurements from a few points. Basically if one had a graph with a line across at y=0.48, one marks in the values in table 3, join the dots using straight lines (they say that as far as I recall), calculate the area above the y=0.48 line and get the average value over the 52 weeks and that is the QALYs accrued.

So do you mean that there should be an ICER cost per QALY for SMC?

 

[Dolphin]

So do you mean that there should be an ICER cost per QALY for SMC?

If one assumes that if somebody hadn't done SMC and costs for such an individual were exactly the same, the healthcare costs of SMC was -£116. So as the average scores improved, SMC would be dominant overdoing nothing.

Similarly, the societal cost of SMC was -£1914. So as the average scores improved, SMC would be dominant over doing nothing.

 

[Bob]

If one assumes that if somebody hadn't done SMC and costs for such an individual were exactly the same, the healthcare costs of SMC was -£116. So as the average scores improved, SMC would be dominant overdoing nothing.

Similarly, the societal cost of SMC was -£1914. So as the average scores improved, SMC would be dominant over doing nothing.

Ah, OK, I think I understand that now.

Thanks for all your help today Dolphin.

I think it's time for bed now.

I'll probably dream that I'm being attacked by a mass of evil looking QALYs.

 

[Dolphin]

They found the average improvement for the SMC group over baseline was 0.02 QALYs. This is based on taking measurements from a few points. Basically if one had a graph with a line across at y=0.50, one marks in the values in table 3, join the dots using straight lines (they say that as far as I recall), calculate the area above the y=0.50 line and get the average value over the 52 weeks and that is the QALYs accrued.

Bob: I wrote an error when I first did this (I was looking at the wrong column). The baseline level is 0.50, not 0.48.

 

[Dolphin]

If one assumes that if somebody hadn't done SMC and costs for such an individual were exactly the same, the healthcare costs of SMC was -£116. So as the average scores improved, SMC would be dominant overdoing nothing.

Similarly, the societal cost of SMC was -£1914. So as the average scores improved, SMC would be dominant over doing nothing.

Imagine, instead, that the healthcare costs of SMC was £500 (over doing nothing). Then, the ICER (healthcare) would be £500 / 0.02 = £25,000 per QALY.*

*Dividing by a number less than one may be confusing. Another way of thinking of it is that one would need 50 of 0.02 changes to make 1 QALY. So it costs 50 x £500 = £25,000 per QALY.

 

[Bob]

Dolphin I'm sure that this is a stupid question, but if they don't know the ICER costs per QALY for SMC then how can they compare the cost effectiveness per QALY of CBT/GET with SMC?

 

[Bob]

If one assumes that if somebody hadn't done SMC and costs for such an individual were exactly the same, the healthcare costs of SMC was -£116. So as the average scores improved, SMC would be dominant overdoing nothing.

I've been thinking about it further...
QALY for SMC was measured at baseline and 52 weeks, and costs can be compared pre-randomisation and post-randomisation for SMC.
So costs per QALY could be calculated for SMC, using these sets of data, couldn't they?

 

[Dolphin]

Dolphin I'm sure that this is a stupid question, but if they don't know the costs per QALY for SMC then how can they compare the cost effectiveness of CBT/GET with SMC?

One can make any comparison one likes. One could have a standard regime than involves 10 parts and then have a regime which has the same standard 10 parts plus one more and one coud make calculations passed on whether it is seen as cost effective to add on that extra part. So what they are assuming is everybody should get SMC, is it cost effective to add on CBT or GET. One doesn't need to know the cost of SMC to make such a calculation.

 

[Dolphin]

I've been thinking about it further...
QALY for SMC was measured at baseline and 52 weeks, and costs can be compared pre-randomisation and post-randomisation for SMC.
So costs per QALY could be calculated for SMC, using these sets of data, couldn't they?

Yes, that's what I did to get -£116 and -£1914 (the data in the last column of Table 3)

Dominant means on average it doesn't cost anything to get the QALY.

But I then gave an example of how one could calculate it if it was a different situation.

 

[Bob]

One can make any comparison one likes. One could have a regime than involves 10 parts and then have a standard regime which has the same 10 parts plus one more and one coud make calculations passed on whether it is seen as cost effective to add on that extra part. So what they are assuming is everybody should get SMC, is it cost effective to add on CBT or GET. One doesn't need to know the cost of SMC to make such a calculation.

I'm not sure that's right Dolphin. The main thrust of the paper is a comparison between SMC and GET/CBT/APT, saying that GET/CBT are more cost effective than SMC.
In the abstract, it seems to suggest that they know the cost per QALY for SMC:
"If society is willing to value a QALY at £30,000 there is a ... 7.9% that SMC alone is most cost-effective."

But I don't understand it all anyway. I'll leave it there for the night.

 

[Dolphin]

I'm not sure that's right Dolphin. The main thrust of the paper is a comparison between SMC and GET/CBT/APT, saying that GET/CBT are more cost effective than SMC.
In the abstract, it seems to suggest that they know the cost per QALY for SMC:
"If society is willing to value a QALY at £30,000 there is a ... 7.9% that SMC alone is most cost-effective."

But I don't understand it all anyway. I'll leave it there for the night.

You're forgetting that CBT and GET means CBT+SMC and GET+SMC, they are just using shorthand. So if CBT is seen as cost effective, what they are actually saying is CBT+SMC is cost effective (over SMC). There was no group that had CBT but no SMC so they can't make comment on such an option from this data.

 

[Bob]

You're forgetting that CBT and GET means CBT+SMC and GET+SMC, they are just using shorthand. So if CBT is seen as cost effective, what they are actually saying is CBT+SMC is cost effective (over SMC). There was no group that had CBT but no SMC so they can't make comment on such an option from this data.

It is really confusing, but I don't think that's how they've done it.
I don't think that's how Table 6 is set out.
They talk about 'incremental' costs, and I think that means additional costs to SMC costs, not including SMC costs.
My own calculations were for additional costs (separate from SMC costs), and they worked out very similar to their Table 6 figures, so I think i'm right.
So I imagined that QALYs were looked at in the same way, but I haven't seen any data for QALYs yet, to work it out.

 

[Dolphin]

They talk about 'incremental' costs, and I think that means additional costs to SMC costs.

I'm not sure how that is different to what I have been saying.

 

[Bob]

I'm not sure how that is different to what I have been saying.

OK, I'm too tired to understand anything now anyway...
All I really wanted to know is what the costs per QALY for SMC are, to make a comparison with CBT and GET.
I'll read the paper again tomorrow to see if I can make any more sense of it.

Thanks again for all your help Dolphin.

 

[Dolphin]

OK, I'm too tired to understand anything now anyway...
All I really wanted to know is what the costs per QALY for SMC are, to make a comparison with CBT and GET.
I'll read the paper again tomorrow to see if I can make any more sense of it.

Ok. But remember, all you have figures for is CBT+SMC and GET+SMC (rather than (i) CBT, but no SMC and (ii)GET, but no SMC) so that is what you will be comparing with SMC.

 

[Bob]

Ok. But remember, all you have figures for is CBT+SMC and GET+SMC (rather than (i) CBT, but no SMC and (ii)GE, but no SMC) so that is what you will be comparing with SMC.

What I meant about Table 6, in my previous post, is that the figures are for CBT only and GET only, not CBT+SMC etc.
I'm pretty sure about that for all the figures apart from QALYs, but I assume that they are the same.

 

[Dolphin]

What I meant about Table 6, in my previous post, is that the figures are for CBT only and GET only, not CBT+SMC etc.
I'm pretty sure about that for all the figures apart from QALYs, but I assume that they are the same.

The figures in Table 6 are for comparisons between CBT+SMC vs SMC, GET+SMC vs SMC and APT+SMC vs SMC. I'm pretty sure of this.
If one looks at the conclusion of the paper, it says:

At a conventional cost-per-QALY threshold considering a one year outcome CBT has the highest probability of being the most cost-effective treatment option for CFS when given as a supplement to SMC and compared to SMC alone.

(my underlining)

 

[Bob]

The figures in Table 6 are for comparisons between CBT+SMC vs SMC, GET+SMC vs SMC and APT+SMC vs SMC. I'm pretty sure of this.
If one looks at the conclusion of the paper, it says:

At a conventional cost-per-QALY threshold considering a one year outcome CBT has the highest probability of being the most cost-effective treatment option for CFS when given as a supplement to SMC and compared to SMC alone.

(my underlining)

I did the calculations earlier, and found them to be for CBT only and GET only, compared with SMC.
I might be wrong but I'll double-check it tomorrow and send you the details.

Edit: To clarify, this applies to Table 6. Table 3 is for CBT+SMC and GET+SMC vs SMC alone.

 

[Firestormm]

New comment from Tom: http://www.plosone.org/annotation/listThread.action?root=52637

 

[Bob]

Edit:
Please ignore this post. I got very confused about the comparisons that they are making in the paper.
As Dolphin says, they are comparing CBT+SMC with SMC.


Table 6 shows the incremental changes for APT/CBT/GET, so the sums in Table 6 do not include changes seen for SMC, but they are a comparison of CBT+SMC vs SMC, and GET+SMC vs SMC, and APT+SMC vs SMC.


Even if they've used ICER, my rough calcuations don't agree with their analysis...

By my calculations, SMC is dominant over GET for physical function scores (I haven't looked at fatigue scores yet), so I don't know how they worked out that GET is dominant.
(My figures aren't exactly accurate, because they've made changes to the number of participants used in Table 6., but I can't see how that could account for the discrepancy between their figures and mine.)


Costs per person societal for physical function.

SMC
Per person societal costs (-1914 x 148) / (148 x 58%) = -£3300 savings per person improved

GET
Per person societal costs (-197 x 140) / (140 x 12.6%) = -£1563 savings per person improved, on top of SMC savings

CBT
Per person societal costs (-464 x 145) / (145 x 13.4%) = -£3463 savings per person improved, on top of SMC savings.


Or using Simon's unadjusted figures (to make sure I'm not making major errors):

SMC
Per person societal costs (-1933 x 148) / (148 x 58%) = -£3332 savings per person improved

GET
Per person societal costs (-244 x 140) / (140 x 12.6%) = -£1937 savings per person improved, on top of SMC savings.

CBT
Per person societal costs (-532 x 145) / (145 x 13.4%) = -£3970 savings per person improved, on top of SMC savings.