1The CRC PET Oncology Group, Hammersmith Hospital, London 2Christie Hospital, Manchester, UK
Positron emission tomography (PET) with 18F-fluorodeoxyglucose (FDG) exploits the abnormal energy metabolism of tumours to enable functional as opposed to structural imaging. The last decade has seen more than 100 reports of the sensitivity and specificity of FDG-PET for the detection and staging of cancer in various settings [1]. These show that a combination of PET and conventional imaging mostly outperforms conventional imaging on its own, in its ability to predict histopathological reference findings. The progression of research in clinical PET to address combined modality therapies is welcome, and it is encouraging to see the results of groups represented by Flamen and colleagues published in dedicated oncological journals such as the Annals [2]. The increasing predictive power of diagnostic technologies is bound to change the way we practice medicine in coming years. In fact, the technologies themselves are already here; the real challenge is to identify when they will improve patient outcomes enough to justify their use.
In their study of oesophageal cancer, Flamen and colleagues have demonstrated that FDG-PET can predict "major chemoradiation response". This was broadly defined as pT02 N0 M0 status after treatment (pT3 N0 M0 in selected cases). The authors protocol included baseline PET, chemoradiotherapy (CRT; cisplatinum, fluorouracil and 40 Gy in 20 fractions) until day 29, follow-up PET at mean day 64 and surgery at mean day 87. Serial FDG-PET had a predictive accuracy of 0.78 for major response, with a sensitivity of 0.71 (10/14) and a specificity of 0.82 (18/22). Imaging results were verified by histopathology in at least 30 of 36 patients. The small sample size does introduce statistical uncertainty, with 95% confidence intervals (CI) for accuracy of 0.620.88, for sensitivity of 0.450.88 and for specificity of 0.610.93. (Even where measured accuracy is 1, a denominator of n = 35 is required for the lower boundary of the 95% CI to exceed 0.90). However, a Cox regression model confirmed greater predictive power for survival from the reduction in FDG uptake in follow-up scans (P = 0.002) than from the presence and extent of lymph node involvement in baseline studies (P = 0.087). Ultimately, there are no real surprises for oncology in the authors summary of their results:
Result 1 restates the rule that bulk disease resists therapy. Result 3 confirms the known insensitivity of FDG-PET to microscopic disease. Result 4 confirms increased prognostic accuracy after the fact of therapy. Results 2 and 4 however, are important for PET, as they were not guaranteed in the setting of CRT, which can cause significant inflammation. FDG is taken up by inflammatory cells, which could lead to false posi-tives [3]. Results 2 and 4 show this is not an issue in the authors CRT protocol, at least by 5 weeks from completion of treatment. In summary, FDG-PET does not achieve anything different from standard diagnostic investigations, but it does increase predictive power. The question of exactly what to do with this increased predictive power remains unaddressed.
The authors are interested in developing FDG-PET for very early response assessments, within a few days of starting CRT. However, extrapolations of their data to different CRT regimens or different intervals between FDG-PET would have to be made with caution. Whether or not inflammation might be a problem in a very early window of time is an open question. Although a cited report of follow-up FDG-PET at 2 weeks into treatment looked promising, the patients in that study were receiving chemotherapy only [4]. It should be noted that PET markers other than FDG could be used. The status of FDG as the workhorse of oncological PET is mainly due to the 90 min half-life of 18F. This has facilitated distribution of tracer from central host cyclotrons to peripheral satellite scanners in a financially optimal model of service provision [5]. It has also allowed relatively simple data acquisition in whole body scans for detection of unsuspected advanced disease in pre-surgical staging. However, there is promise in the developmental ligand 18F-fluorothymidine (FLT), which may become an agent of choice for PET response assessment [6]. Although the FLT technique is still under development and may require metabolite correction, it promises a more direct index of cellular proliferation than glucose metabolism. Preliminary work with its 11C-labelled analogue suggests it will have greater accuracy for tumour response assessment [7]. In conclusion, anyone contemplating a study of the utility of serial PET for very early response assessments would be advised to go through the clinical and scientific background in some depth.
Data on response prediction with FDG, FLT and other forms of PET, single-photon emission computed tomography, magnetic resonance imaging, modified computed tomography, molecular methods, special stains, bedside assessment, artificial intelligence and probably other technologies, before, during and after treatment should appear in the literature over the next few years. The looming issue is how cancer clinicians will identify the best from a growing range of choice. This requires a change of perspective of the sort outlined to a European audience by Fryback and Thornbury in 1992 [8]. Their hierarchical model of efficacy climbs through six levels, each a necessary but not sufficient condition for the next:
This model, which effectively translates physical science into economics, is of great value for understanding reports of diagnostic technology. Flamen and colleagues study, in keeping with most published work, has measured diagnostic accuracy [2]. They did not aim to assess therapeutic or patient outcome efficacy, or the cost-benefits of possible strategies for the routine use of serial FDGPET in oesophageal CRT. An assessment of cost-benefits, with key input from cancer clinicians, is now needed. There exist robust decision model methodologies that will not only do this but will also prioritise the next steps for research [9]. We think that clinical scientists will increasingly want to make the case for new practices and research in economic terms. Although there has been resistance to the idea of putting a price on life, it should be remembered that the currency units used in most economic settings are ultimately opportunity costs: a tool for summarising what is foregone when one commits to a given course of action in settings of scarcity. Over and above this, in most scenarios the main determinant of expenditure is not unit cost but the number needed to treat (NNT). Rising NNTs remind us that the scarce commodity does not have to be health scheme Euros, Pounds or Dollars; it can equally well be expert time and care. In other words, the expense of inefficacious tests and treatments is an index of missed opportunities for all concerned, not just the taxpayer. Insofar as public sector physicians have an ideal of achieving the greatest good for the greatest number, arguments for how we do things need to be based on cost-benefit.
A systematic model for defining practice and prioritising research on CRT response assessment in oesophageal carcinoma could begin with three decision nodes; the options to direct management away from futile therapy in non-responders before, during or after CRT. It would include data on life expectancy and morbidity associated with various schedules of CRT, surgery and palliative care. The diagnostic accuracy of imaging and other technologies would be compared with the incidence of non-response within predefined subsets of stage and histology. The costs of diagnostic tests and therapies would be offset against willingness-to-pay for health outcomes; typically set at i50 000 per quality adjusted life year. Second order Monte Carlo simulation can be used to re-express the uncertainty around each variable as the expected value of further information from research. In an age of information overload clinicians and clinician-scientists will have to adopt these kinds of health informatic and economic techniques. Our own experience in building a model for the use of FDG-PET to predict outcome in non-small-cell lung cancer has convinced us that it is both feasible and useful [10]. We think it entirely appropriate for the discussion section of primary reports in medical journals to reference similar analysis as to why their data should change practice or research. The aim for oncologists is to identify questions relevant to the increasing number of answers on offer from our radiological and laboratory colleagues. In the end, it is the clinical and research questions that will define the optimal role of technologies such as FDG-PET.
Acknowledgements
Grant to P.M.P. from Cancer Research UK.
References
1. Bar-Shalom R, Valdivia AY, Blaufox MD. PET imaging in oncology. Sem Nucl Med 2000; 30: 150185.[ISI][Medline]
2.
Flamen P, Van Cutsem E, Lerut A et al. Positron emission tomography for assessment of the response to induction radiochemotherapy in locally advanced oesophageal cancer. Ann Oncol 2002; 13: 361368.
3. Hautzel H, Muller-Gartner HW. Early changes in fluorine-18-FDG uptake during radiotherapy. J Nucl Med 1997; 38: 13841386.[Abstract]
4.
Weber WA, Ott K, Becker K et al. Prediction of response to preoperative chemotherapy in adenocarcinomas of the esophagogastric junction by metabolic imaging. J Clin Oncol 2001; 19: 30583065.
5.
Keppler JS, Conti PS. A cost analysis of positron emission tomography. Am J Roentgenol 2001; 177: 3140.
6. Shields AF, Grierson JR, Dohmen BM et al. Imaging proliferation in vivo with [F-18]FLT and positron emission tomography. Nat Med 1998; 4: 13341336.[ISI][Medline]
7. Shields AF, Mankoff DA, Link JM et al. Carbon-11-thymidine and FDG to measure therapy response. J Nucl Med 1998; 39: 17571762.[Abstract]
8. Thornbury JR, Fryback DG. Technology assessmentan American view. Eur J Radiol 1992; 14: 147156.[ISI][Medline]
9. Fenwick E, Claxton K, Sculpher M, Briggs A. Improving the efficiency and relevance of health technology assessment: the role of iterative decision modelling. In: Centre for Health Economics Discussion Paper Series. York, UK: University of York 2000; 135.
10. Laking GR, Price P, Sculpher MJ. Expected value of perfect information (EVPI) on 18F-fluorodeoxyglucose positron emission tomography (FDG-PET)-based therapy selection in non-small-cell lung cancer (NSCLC). Proc Am Soc Clin Oncol 2002; 21: (Abstr).