NCBI Bookshelf. A service of the National Library of Medicine, National Institutes of Health.

Rogers G, Hoyle M, Thompson Coon J, et al. Dasatinib and Nilotinib for Imatinib-Resistant or -Intolerant Chronic Myeloid Leukaemia: A Systematic Review and Economic Evaluation. Southampton (UK): NIHR Journals Library; 2012 Apr. (Health Technology Assessment, No. 16.22.)

Cover of Dasatinib and Nilotinib for Imatinib-Resistant or -Intolerant Chronic Myeloid Leukaemia: A Systematic Review and Economic Evaluation

Dasatinib and Nilotinib for Imatinib-Resistant or -Intolerant Chronic Myeloid Leukaemia: A Systematic Review and Economic Evaluation.

Show details

Appendix 6Estimation of overall survival by treatment in the PenTAG model

Overall survival curves for each treatment were generated in the following four-stage process.

Stage 1. estimation of survival owing to chronic myeloid leukaemia-related deaths for responders and non-responders

Cytogenetic response has been shown to be a useful predictor of OS and has been used in other models of cost-effectiveness for CML.274 In particular, MCyR, as opposed to CCyR, has been shown to provide a good surrogate outcome for OS.271 Further discussion of this surrogate relationship is provided in Appendix 5.

We define ‘responders’ as those patients who achieve an MCyR while on treatment, and ‘non-responders’ as those patients who do not achieve an MCyR. Two separate Weibull curves, SRESP CML(t) for responders and SNON-RESP CML(t) for non-responders, are used to fit OS owing to CML-related deaths, where OS is better for responders than for non-responders.

In common with another cost-effectiveness model for CML,274 we assumed a constant hazard ratio (λ) between the OS because of CML-related deaths for responders versus non-responders, i.e.:

[Equation 1]

A literature search was undertaken to find those studies that reported OS for patients on normal-dose imatinib according to whether they were responders or non-responders. Trials of imatinib (rather than IFNα were preferred because, according to our clinical experts, OS given a certain response for dasatinib and nilotinib is most likely to be similar to OS for imatinib given its mode of action. A meta-analysis of studies of imatinib concluded that λ = 0.370.

Stage 2. estimation of overall survival (chronic myeloid leukaemia- and non-chronic myeloid leukaemia-related deaths) for responders and non-responders

Next, the OS curves for responders and non-responders, SRESP overall(t) and SNON-RESP overall(t), allowing for CML- and non-CML-related mortality combined, were calculated as follows. First, the rate of CML-related mortality at time t was calculated as:

[Equation 2]

for responders and non-responders, respectively.

Next, the general background mortality rate at time t, rgeneral(t) was taken from UK life tables.232 Then, the OS for responders and non-responders was calculated allowing for the sum of the rates of CML-related and general mortality;

[Equation 3]
[Equation 4]



For consistency with the studies of the drugs for second-line treatment, the male : female ratio was assumed to be 50 : 50 and in the base case, all patients in the model were assumed to be aged 56 years at the start of second-line therapy.

Stage 3. estimation of overall survival for all treatments

If the proportion of responders for a given treatment is denoted by MCyR%, then the OS for that treatment at time t, OS(t), owing to all deaths is:

[Equation 5]

Stage 4. calibration to empirical overall survival

So far, we have not specified the parameters of the CML-related survival for responders and non-responders, SRESP CML(t) and SNON-RESP CML(t), except that they are related by Equation 1. These parameters were estimated by calibration by regressing expected OS for HDI, OS(t), to the empirical OS for HDI from Jabbour et al.182 This empirical data were chosen for two reasons. First, this is the most mature OS for treatment after normal-dose imatinib failure that we are aware of. Second, the response rate is quoted in this study, which allows us to specify precisely the expected OS curve, OS(t).

© 2012, Crown Copyright.

Included under terms of UK Non-commercial Government License.

Bookshelf ID: NBK98826
PubReader format: click here to try


  • PubReader
  • Print View
  • Cite this Page
  • PDF version of this title (3.2M)

Other titles in this collection

Recent Activity

Your browsing activity is empty.

Activity recording is turned off.

Turn recording back on

See more...