Comparing tapentadol to oxycodone/naloxone combination: imagining castles in the air … while building sandcastles! Re: Coluzzi F, Ruggeri M. Clinical and economic evaluation of tapentadol extended release and oxycodone/naloxone extended release in comparison with controlled release oxycodone in musculoskeletal pain. Curr Med Res Opin 2014;30:1139-51

Dear Editor,

Professor Scarpignato and Dr. Gatta raised a number of methodological concerns over our work. In the title of their letter to the editor they specifically used the idiom ‘castles in the air’. We are very grateful for this metaphor, which allows us to clarify some aspects of our work.

1. In our paper we clearly compared tapentadol (TAP) against CR oxycodone (CR OXY), and oxycodone/naloxone (OXY/NAL) against CR oxycodone. In other words, we presented two direct, head-to-head comparisons. A misunderstanding may have arisen from the term ‘indirect’, which in the context of our work should be interpreted in a non-statistical sense, i.e. as meaning there was no direct comparison between TAP and OXY/NAL.

2. Referring to the consideration made for case histories and for the non-homogeneous purposes of the analyzed studies, this aspect has been clearly evaluated and considered by the authors, as already clearly stated in our article. The principal aim in evaluating studies for inclusion in a meta-analysis is the combinability of treatments, patients and outcomes. A meta-analysis is a statistical procedure that integrates the results of several independent studies considered to be ‘combinable’1.

   • (A) Considering the first variable (i.e. treatments), the selected studies are clearly combinable as we presented direct comparisons between TAP and CR OXY, and between OXY/NAL and CR OXY.

   • (B) For the second variable (i.e. patients), there is sufficient overlap between patient populations to allow comparisons to be made. The three OXY/NAL studies all included patients with moderate to severe chronic non-cancer pain, and the trial by Vondrackova et al.2 specified patients with chronic low back pain (LBP). Similarly the three TAP studies have been conducted on patients with chronic pain related to osteoarthritis, low back, or both.

   • (C) Finally, regarding the third variable (i.e. outcomes), both the TAP and OXY/NAL studies measured the impact of the drugs on bowel function and reported the number of discontinuations due to lack of efficacy or side effects. Skin symptoms were not available in the OXY/NAL studies; therefore the authors could not include this item in their final comparison. As outlined under the previous three points, the studies included in our meta-analysis have the same outcomes and are, therefore, combinable.

3. As clearly stated in our article, some considerations may come from the different percentage of opioid-naïve patients respectively included in the TAP and the OXY/NAL studies. This bias, however, leads to two fundamental consequences: the first is that it reinforces, even more, the tolerability data coming from studies with TAP, since the majority of the patients enrolled were opioid naïve and therefore more likely to develop side effects; the second is that patients already on a stable treatment with opioids (as required by inclusion criteria in the OXY/NAL studies) have a lower rate of side effects (in absolute). The only method to overcome this bias is to use the risk ratio, which showed that both treatments significantly reduced the risk of developing side effects, particularly constipation, which was reduced by about 40% compared to CR OXY in patients treated with TAP and in those treated with OXY/NAL.

4. The first aim of any kind of model, even economical models, is not to describe reality as it is in all its aspects, but to outline the key aspects of a problem in order to define and analyze how reality could be. This creates important opportunities for making informed and responsible decisions. Economic models do not build castles in the air. Rather they build a strong architecture that can systematize clinical problems and guide decisions.

5. The real problem, coming from the use of decision models, which determines their usefulness and proper application to reality, is their robustness: structural (ability to describe the problem) and analytical (correctness of the calculations and internal and external consistency of the parameters used to settle the decision problem). Let’s try to apply these criteria to our paper.

   • (A) Structural robustness: our model answers, for the purpose of an economic evaluation, the currently recognized treatment of chronic pain with particular reference to musculoskeletal pain. Any reference to the literature, as well as international guidelines, confirms that, in the clinical field of analysis, there are two innovative alternatives: tapentadol and oxycodone/naloxone. On the other hand the standard of care is represented by oxycodone. The trials for the two innovative treatments are drawn representing this comparison. At the time of the drafting of our article, a direct comparison between oxycodone/naloxone and tapentadol was not available in literature. Our model reflects the exact nature of the scientific evidence that was available at the time, thereby avoiding aprioristically the direct comparison between the two innovative treatments. Thanks to the probabilistic assumptions that we made in conducting the sensitivity analysis, it was possible to rebuild, orient and integrate information from different sources, only formulating hypotheses about specific stochastic distributions, while maintaining a high level of scientific exactness. Singular ‘alpha’ and ‘beta’ parameters used to fit every single stochastic distribution, which we did not include in the manuscript for editing reasons, are available by writing to the corresponding author. Moreover, the probabilistic sensitivity analysis is considered one of the pillars of health economic evaluations used to address and solve problems of a complex nature, in terms of uncertainty, in many areas of decision making. Agencies for health technology assessment (HTA) of most of the developed countries, such as the National Institute for Health and Clinical Excellence (NICE) and the Italian Medicines Agency (AIFA), make extensive use of probabilistic sensitivity analyses and consider them a real operational tool to achieve the ‘bridge between science and decisions’, which is the true nature of HTA3–8.

   • (B) Consistency of the parameters: in this case too the probabilistic sensitivity analysis helps us. The ability of the model to reproduce important parts of reality necessarily passes through the assignment of a probabilistic distribution of the parameters used. Experts in economics and outcomes research follow ISPOR’s methodological guidelines for dealing with economic evaluations to ensure the reliability of their work. Yet, verification of compliance with these guidelines is imposed by the above-mentioned HTA agencies, and is an important part of the evaluation process. Therefore, compliance with the ISPOR guidelines has been stated in our work and made clear by precise references. Intuitively, the potential to assign probabilistic values instead of deterministic ones to the used parameters, is, on the one hand, the ability to incorporate, in a consistent manner, evidence coming from different sources (six trials in our paper) and, on the other hand, mapping the uncertainty about a specific decision problem. Decision models achieve their goal when they do not return a value, but a degree of reliability of the same. This is the most valuable information models leave to decision-makers, a wealth of knowledge about a problem, rather than a single solution.

6. In actual fact, we applied probabilistic distributions to efficacy data, adverse events, and costs of our model. If the trials return statistically significant values in the differences between adverse events in favor of tapentadol compared to oxycodone, the dominance of the first drug compared to oxycodone follows, since the higher price is offset by lower costs for adverse events. Even if the same happens for OXY/NAL, since the price of TAP is lower than OXY/NAL, it follows that the former drug is superior when compared with the latter. Referring to the objection on ‘discontinuation’ as a statement of the proposed Markov model, we would like to draw the readers’ attention to Table 1 in our original manuscript. Here, two different utility coefficients have been clearly reported respectively for ‘discontinuation due to lack of efficacy’ and ‘discontinuation after adverse events’. Therefore it is evident from our Table 1 that different inputs for discontinuation were utilized in the model and that these were associated with different probability values. Conversely, we do not consider Figure 1 of Professor Scarpignato’s and Dr. Gatta’s letter appropriate here, since the inclusion criteria for the trials on OXY/NAL entered only patients who had already obtained analgesic benefits by oxycodone or by opioids in general. This was not true for TAP, which had a higher percentage of opioid-naive patients in general. Therefore, also regarding the ‘discontinuation due to lack of efficacy’, it may indeed be inappropriate to compare studies in which some patients are ‘opioid naïve’ (i.e. as in TAP trials), while the others are all in chronic treatment with opioids (i.e. as in OXY/NAL trials). However, in this case, as previously discussed, the disadvantage is for TAP, because the opioid-naive patients are not already tolerant to some of the most frequent and immediate side effects, such as nausea and vomiting. It is also true that these results would be reversed if superiority in efficacy of OXY/NAL compared with TAP were proved. In our meta-analysis we assumed equal efficacy between the two treatments remaining; however, we were very conservative on this figure. Recent findings, as we are going to discuss in point 8 of our reply, showed exactly the opposite result: the efficacy of TAP in patients with chronic low back pain with a neuropathic component is significantly higher than that of OXY/NAL9–11.

7. The alternative to our meta-analysis, as proposed by the authors of the letter, is network meta-analysis. This approach, although used in many cases, presents some issues that prevented us from considering it a complete tool for the decision-making process. Just to mention a few of these issues: the deterministic view instead of the probabilistic one, non-inclusiveness of cost parameters, the incapacity to consider all the data in an integrated manner, and the inability to project the results over several periods.

8. Finally, the results of our model seem to be confirmed by new interim results, already published, which provided supportive evidence in line with our analysis and showed a direct superiority in terms of efficacy and tolerability of TAP with respect of OXY/NAL8–10. A recent multicenter randomized study compared TAP and the association OXY/NAL in patients with severe chronic LBP, with a proven neuropathic component. The total study length was 12 weeks (3w titration phase + 9w maintenance phase). The primary endpoint of efficacy (variation of basal NRS -3 at the end of the study, 12 weeks) has been reached; in fact, TAP caused a significant reduction in the value of NRS -3, showing itself to be more effective than OXY/NAL. This result has a clinical value; in fact, during the entire study period, TAP showed better analgesic results compared to OXY/NAL. TAP allowed the attainment of an improvement in quality of life significantly superior to OXY/NAL, as well as a better control of all the major symptoms of neuropathic pain. Consequently, it has been reported that there was greater adherence to the treatment with TAP. During the study, tolerability was generally better in patients receiving TAP compared to those treated with OXY/NAL. Moreover a significantly lower incidence of constipation has been observed in the TAP group compared to the OXY/NAL group. If patients were pre-treated with oxycodone (e.g. in an OXY/NAL trial) and had constipation or nausea, this would be captured at trial baseline as part of medical history (as pre-existing) and not as an adverse event. Under the new treatment, even if the events were ongoing at the same severity, they would not be recorded as adverse events, unless they were worsening. These data confirm in a conclusive manner that the model we adopted is absolutely valid and that it concretely approached the clinical and economic reality.

We therefore agree on avoiding castles in the air, but even more we warn against facing complex problems by building castles of friable sand, which could go into the eyes and confuse readers. We believe in building structurally sound models, which reflect the complex vision of modern society, and that can approach complex decision-making problems in an analytical way.

Sincerely,

Flaminia Coluzzi

Dept. Medical and Surgical Sciences and Biotechnologies

Sapienza University of Rome

Corso della Repubblica 64, 04100 Latina, Italy Tel: [phone]

Fax: [phone];

[email protected]

Matteo Ruggeri

ALTEMS – Postgraduate School in Health Economics and Management Catholic University of Sacred Heart of Rome Italy

Transparency

Declaration of funding

This letter was not funded.

Declaration of financial/other relationships

The authors have no relevant financial relationships to disclose other than those already reported in their original article.

Acknowledgments

The authors would like to thank Prof. Rossella Di Bidino PhD MSc, health economist, external consultant of the Pricing and Reimbursement Committee, AIFA (Italian Medicines Agency) and Dr. Adam B. Smith PhD and Dr. Rob Hodgson PhD, statisticians at York Health Economics Consortium, University of York, for their valuable comments and suggestions in evaluating their results.