A Flawed Study Should Not Define a New Standard of Care

Abstract

Targeted intraoperative radiation therapy (TARGIT) puts forward the tantalizing prospect of marked simplification of breast radiation therapy to a single treatment delivered with an easily shielded low-energy (50-kV) x-ray device at the time of breast surgery. Based upon the results of the TARGIT-A trial, on July 25, 2014, the National Institute of Health and Care Excellence (NICE) in the United Kingdom gave preliminary recommendation for the use of TARGIT within the UK National Health Service (1Smyth C. Single-dose radiotherapy eases breast cancer stress. The Times of London, July 25, 2014http://www.thetimes.co.uk/tto/health/news/article4157186.ece?shareToken=55daafa616d42f8a6e51072c991c46f5Google Scholar). This endorsement and the growth in the use of this modality appear to be occurring without the full knowledge and recognition of the methodologic flaws of the TARGIT-A trial. These flaws are sufficiently profound as to undermine confidence in both the efficacy and the safety of TARGIT and should provide pause to any clinician considering its application. The prospective TARGIT-A trial used a noninferiority design to randomly assign patients to TARGIT or whole breast external beam radiation therapy (EBRT). For 67% of the study participants, the randomization and treatment occurred at the lumpectomy (classified as “pre-pathology”). The patients with high-risk features went on to receive EBRT after TARGIT. These features were defined as margin <1 mm, extensive ductal carcinoma in situ, invasive lobular carcinoma, or “individual centers could specify more than these core factors” (2Vaidya J.S. Wenz F. Bulsara M. et al.Risk-adapted targeted intraoperative radiotherapy versus whole-breast radiotherapy for breast cancer: 5-year results for local control and overall survival from the TARGIT-A randomized trial.Lancet. 2013; 383: 603-613Abstract Full Text Full Text PDF PubMed Scopus (612) Google Scholar) at their discretion. For 33% of the study participants, the definitive pathology was already available from a prior lumpectomy. This “post-pathology” stratum of patients was predetermined as being at low risk and, if randomized to TARGIT, was returned for reopening of the lumpectomy wound specifically for intraoperative irradiation. Vaidya et al (2Vaidya J.S. Wenz F. Bulsara M. et al.Risk-adapted targeted intraoperative radiotherapy versus whole-breast radiotherapy for breast cancer: 5-year results for local control and overall survival from the TARGIT-A randomized trial.Lancet. 2013; 383: 603-613Abstract Full Text Full Text PDF PubMed Scopus (612) Google Scholar) reported the TARGIT-A trial with a median follow-up time of 2.4 years. Of the entire cohort (including “pre- and post-pathology”), EBRT was given to 15% of patients in the TARGIT arm. The results would seem to indicate a clear failure of TARGIT, given that the 5-year actuarial local failure (LF) rate was higher than with EBRT (3.3% vs 1.3%, respectively; P=.042). However, the authors concluded that the predetermined 2.5% noninferiority threshold was met, and, as such, TARGIT successfully proved itself as equivalent to EBRT. Subgroup analysis revealed a 5-year LF for “pre-pathology” of 2.5% with TARGIT versus 1.7% with EBRT (P=.31). In “post-pathology,” the 5-year LF was 5.4% with TARGIT versus 1.7% with EBRT (P=.069). The investigators concluded that TARGIT is more effective “pre-pathology” than “post-pathology.” The articles that report the TARGIT-A trial often leave readers pondering some direct questions. Has a radically new paradigm been defined that upends our understanding of the pathologic anatomy of early breast cancer, the physical laws that define the dosimetry of a 50-kV x-ray source, and the well-established biological principles that underpin the therapeutic effect of radiation therapy? Or is something amiss in how the TARGIT-A trial was designed, executed, and analyzed? A critical review of the methods and data would suggest the latter. In a statement that pointedly encapsulates the shortcomings associated with the TARGIT-A trial, the internationally renowned statistician Professor Jack Cuzick wrote, “The TARGIT-A trial is a good example of trying to make data fit a preexisting hypothesis; there are several major deficiencies in the analysis” (3Cuzick J. Radiotherapy for breast cancer, The TARGIT-A trial.Lancet. 2014; 383: 1716Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar). The force of his words is underscored by the fact that he served as the initial chair of the Data Monitoring Committee for the TARGIT-A trial. A cogent critique of the TARGIT-A trial by Professor Cuzick appears alongside those of other distinguished experts in an extraordinary series of letters published recently in The Lancet (3Cuzick J. Radiotherapy for breast cancer, The TARGIT-A trial.Lancet. 2014; 383: 1716Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar, 4Haviland J. A'Hern R. Bentzen S. et al.Radiotherapy for breast cancer: The TARGIT A trial.Lancet. 2014; 383: 1716-1717Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar, 5Harness J. Silverstein M. Wazer D. et al.Radiotherapy for breast cancer: The TARGIT A trial.Lancet. 2014; 383: 1718-1719Abstract Full Text Full Text PDF PubMed Scopus (14) Google Scholar, 6Yarnold J. Offersen B. Olivotto I. et al.Radiotherapy for breast cancer, The TARGIT A trial.Lancet. 2014; 383: 1717-1718Abstract Full Text Full Text PDF PubMed Scopus (9) Google Scholar, 7Mackenzie P. Fyles A. Chung C. Radiotherapy for breast cancer: The TARGIT A trial.Lancet. 2014; 383: 1717Abstract Full Text Full Text PDF PubMed Scopus (17) Google Scholar). We will briefly summarize the points raised by these authors and discuss additional questions that surround the TARGIT technique. The fundamental design and statistical analysis of the TARGIT-A trial has been challenged: “Paramount among these is the misuse of noninferiority criterion” (3Cuzick J. Radiotherapy for breast cancer, The TARGIT-A trial.Lancet. 2014; 383: 1716Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar). This requires the upper 90% confidence interval be below the predetermined threshold of 2.5%. Contrary to the conclusions of Vaidya et al (2Vaidya J.S. Wenz F. Bulsara M. et al.Risk-adapted targeted intraoperative radiotherapy versus whole-breast radiotherapy for breast cancer: 5-year results for local control and overall survival from the TARGIT-A randomized trial.Lancet. 2013; 383: 603-613Abstract Full Text Full Text PDF PubMed Scopus (612) Google Scholar), this criterion was not met. When the appropriate 5-year LF rates are used, there was a significant 2% superiority of EBRT and a confidence interval that extended beyond 2.5% (3Cuzick J. Radiotherapy for breast cancer, The TARGIT-A trial.Lancet. 2014; 383: 1716Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar, 4Haviland J. A'Hern R. Bentzen S. et al.Radiotherapy for breast cancer: The TARGIT A trial.Lancet. 2014; 383: 1716-1717Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar). Haviland et al (4Haviland J. A'Hern R. Bentzen S. et al.Radiotherapy for breast cancer: The TARGIT A trial.Lancet. 2014; 383: 1716-1717Abstract Full Text Full Text PDF PubMed Scopus (28) Google Scholar) stated that the noninferiority test statistic is unreliable because its appropriate application requires that 5-year follow-up data be available for all patients. Such data were available for fewer than 20% of the TARGIT-A cohort. Professor Cuzick concludes that “the present attempt to argue for virtually no difference by misuse of the noninferiority criterion….does not give an objective assessment of this treatment modality.” The median follow-up time for the TARGIT-A cohort of just over 2 years is inadequate to enable conclusions to be drawn regarding risk of LF or normal tissue toxicity. For LF, the distorting effect of short follow-up is particularly applicable to the high proportion of low-risk patients with small, estrogen receptor (ER) positive tumors. For such a population, LF will mostly occur after 5 years (8Hughes K. Schnaper L. Bellon J. et al.Lumpectomy plus tamoxifen with or without irradiation in women age 70 years or older with early breast cancer: Long-term follow-up of CALGB 9343.J Clin Oncol. 2013; 31: 2382-2387Crossref PubMed Scopus (716) Google Scholar). TARGIT patients with high-risk features also received EBRT (22% of patients in the “pre-pathology” stratum), obscuring any difference in outcome between TARGIT alone and EBRT. For normal tissue toxicity associated with high-dose focal breast irradiation, the brachytherapy-based accelerated partial breast irradiation experience has shown that the evolution of late effects requires at least 5 years of follow-up to be accurately characterized (9Kaufman S. DiPetrillo T. Price L.L. et al.Long-term outcome and toxicity in a Phase I/II trial using high-dose-rate multicatheter interstitial brachytherapy for T1/T2 breast cancer.Brachytherapy. 2007; 6: 286-292Abstract Full Text Full Text PDF PubMed Scopus (69) Google Scholar). In subgroup analysis, the 5-year LF rate for “pre-pathology” TARGIT (2.5%) was higher than that of “post-pathology” (5.4%) and was not statistically different from EBRT (1.7%). On the basis of this observation, the trial authors concluded that the timing of TARGIT in relationship to lumpectomy was an important variable. Professor Cuzick has succinctly described that for any subgroup inquiry, to prevent spurious statistical results, correction for multiple comparisons and tests for heterogeneity between subgroups (eg final margin status) should be performed (3Cuzick J. Radiotherapy for breast cancer, The TARGIT-A trial.Lancet. 2014; 383: 1716Abstract Full Text Full Text PDF PubMed Scopus (27) Google Scholar). The results of such analyses were not provided in the TARGIT-A trial. The TARGIT-A investigators claim that non–breast cancer deaths were increased almost immediately after treatment in the EBRT arm secondary to a greater incidence of lethal cardiovascular events and rapidly fatal radiation-induced malignancies. Antiquated EBRT techniques can cause a very small rise in non–breast cancer (mostly cardiac) mortality (10Early Breast Cancer Trialists' Collaborative Group (EBCTCG)Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: An overview of randomized trials.Lancet. 2005; 366: 2087-2106Abstract Full Text Full Text PDF PubMed Scopus (3914) Google Scholar, 11Darby S.C. Cutter D.J. Boerma M. et al.Radiation-related heart disease: Current knowledge and future prospects.Int J Radiat Oncol Biol Phys. 2010; 76: 656-665Abstract Full Text Full Text PDF PubMed Scopus (440) Google Scholar). Death secondary to radiation-induced morbidity is usually not discernible until a minimum latency of 10 years (10Early Breast Cancer Trialists' Collaborative Group (EBCTCG)Effects of radiotherapy and of differences in the extent of surgery for early breast cancer on local recurrence and 15-year survival: An overview of randomized trials.Lancet. 2005; 366: 2087-2106Abstract Full Text Full Text PDF PubMed Scopus (3914) Google Scholar, 11Darby S.C. Cutter D.J. Boerma M. et al.Radiation-related heart disease: Current knowledge and future prospects.Int J Radiat Oncol Biol Phys. 2010; 76: 656-665Abstract Full Text Full Text PDF PubMed Scopus (440) Google Scholar). By contrast, the TARGIT-A trial observed virtually no latency in that a mortality difference was apparent after only 2 years of follow-up. This finding strains credibility because it asserts a heretofore unreported rapidity by which contemporary EBRT allegedly causes not only injury but nearly immediate lethal consequences. The TARGIT-A investigators further claim that some of the deaths due to EBRT included stroke and bowel ischemia. In the TARGIT-A trial, the toxic effect of EBRT appeared to be related to whether randomization occurred before or after clinicians had access to the definitive pathology results. On subgroup analysis, “pre-pathology” TARGIT resulted in a lower rate of non–breast cancer deaths, whereas for “post-pathology” TARGIT, non–breast cancer mortality was no different than that for EBRT. The most direct explanation, and one that cuts to the core of the structural design of the TARGIT-A trial and the integrity of its enrollment and randomization process, is that patients were not appropriately assessed and stratified for preexisting comorbidities. The predictable end product of such a grievous design flaw would be unbalanced allocation to the treatment arms and resultant regimen-associated mortality statistics that defy common sense. A common question when highly favorable results are presented for partial breast irradiation is whether any breast radiation was truly necessary. Most patients enrolled in TARGIT-A were postmenopausal and had tumors that were grade 1 to 2, ER positive, Her2 negative, size <2 cm, node negative, and treated with systemic therapy. Any patient randomized to TARGIT alone who had high-risk features went on to receive EBRT. As such, only the most favorable patients were treated exclusively with TARGIT. What is the expected risk of LF if they were to receive no radiation at all? Several clinical trials have evaluated the utility of EBRT in low-risk women with small, node negative, ER positive tumors treated with endocrine therapy. These trials (8Hughes K. Schnaper L. Bellon J. et al.Lumpectomy plus tamoxifen with or without irradiation in women age 70 years or older with early breast cancer: Long-term follow-up of CALGB 9343.J Clin Oncol. 2013; 31: 2382-2387Crossref PubMed Scopus (716) Google Scholar, 12Fisher B. Bryant J. Dignam J.J. et al.Tamoxifen, radiation therapy, or both for prevention of ipsilateral breast tumor recurrence after lumpectomy in women with invasive breast cancers of one centimeter or less.J Clin Oncol. 2002; 20: 4141-4149Crossref PubMed Scopus (527) Google Scholar, 13Fyles A.W. McCready D.R. Manchul L.A. et al.Tamoxifen with or without breast irradiation in women 50 years of age or older with early breast cancer.N Engl J Med. 2004; 351: 963-970Crossref PubMed Scopus (533) Google Scholar) reported 4% to 8% 5-year rates of ipsilateral breast plus regional nodal recurrence after lumpectomy and hormonal therapy without radiation therapy. These results compare favorably with the 5-year 3.3% in-breast only recurrence rate reported in the TARGIT-A trial. This suggests unnecessary irradiation of many patients who might have been best served with hormonal therapy alone. The seductive power of a large randomized trial to influence clinical practice cannot be dismissed. However, robust accrual, prominent publication, and aggressive promotion cannot substitute for critically objective evaluation of the methods and data that have been advanced in support of the TARGIT technique. The TARGIT-A trial has many methodologic and analytic flaws that deeply undermine the scientific validity of its claims. In the interest of all women with early breast cancer, clinicians and policy makers must carefully assess the actual state of our current knowledge associated with this modality and recognize that many more questions need to be addressed before we can declare that we have arrived at a new standard of care. In Regard to Vaidya et alInternational Journal of Radiation Oncology, Biology, PhysicsVol. 92Issue 5PreviewTo the Editor: In Vaidya's response to Hepel and Wazer (1), Fig 1 illustrates how the upper confidence limit for the difference between treatment outcomes allows assessment of noninferiority in a trial (2). This follows US Food and Drug Administration guidelines (3). It is imperative, however, that the underlying statistical analysis method and associated confidence limit are appropriate to the event of interest and the completeness of data. It is disappointing, therefore, that the TARGeted Intra-operative radioTherapy (TARGIT) trialists persist in using a noninferiority test statistic based on binomial proportions. Full-Text PDF In Regard to Vaidya et alInternational Journal of Radiation Oncology, Biology, PhysicsVol. 92Issue 5PreviewTo the Editor: The manner in which results of the TARGeted Intra-operative radioTherapy (TARGIT)-A trial (1) have been presented has precipitated a strong response from the international community of doctors and scientists working in breast cancer treatment (2-6). The patient community has had little input to date, but their interest has been provoked by Figure 4 from the TARGIT-A coinvestigators' recent manuscript (7). This figure is potentially misleading to women trying to make decisions regarding their breast cancer treatment. Full-Text PDF In Regard to Vaidya et alInternational Journal of Radiation Oncology, Biology, PhysicsVol. 92Issue 5PreviewTo the Editor: Intrabeam (Carl Zeiss Meditec AG, Germany) is 1 of the few technologies to have been the subject of dispute between world experts in the most prestigious of scientific journals (1, 2). Such public debate mandates that readers have more than average understanding of statistical methods, radiotherapeutic technique (kV-x rays, electrons, MV) and familiarity with current data regarding this unique subgroup of low-risk breast cancer (BC) patients (1-3). Low-risk BC has become a prevalent disease thanks to early detection, and from a public standpoint, we cannot afford to lessen the level of “skepticism and criticism” of any new treatment because mistakes or hasty decisions might constitute a heavy burden on public health in the long term. Full-Text PDF In Regard to Hepel and WazerInternational Journal of Radiation Oncology, Biology, PhysicsVol. 92Issue 5PreviewTo the Editor: The efficacy and safety of low photon energy intraoperative radiation therapy (IORT) during breast-conserving surgery (BCS) was evaluated in the TARGIT-A trial (1), which provides robust level 1 evidence for acceptance of IORT as an alternative to whole-breast radiation (WB-EBRT) for selected patients. Several recent articles (2-6) have criticized IORT on the basis of misinterpretation or misrepresentation of TARGIT-A. Full-Text PDF In Regard to Hepel and WazerInternational Journal of Radiation Oncology, Biology, PhysicsVol. 92Issue 5PreviewTo the Editor: I read the Red Journal's editorial by Hepel and Wazer (1) with interest. I agree with many issues raised by the authors, yet the profound thoughts need further scrutiny. Full-Text PDF In Regard to Hepel and WazerInternational Journal of Radiation Oncology, Biology, PhysicsVol. 92Issue 5PreviewTo the Editor: The initial editorial by Hepel and Wazer (1) contained several scientific, conceptual, and factual inaccuracies. In our own article, we had attempted to correct these and clarify the results of the trial. We find a striking absence of true scientific discourse in much of the content of the subsequent correspondence. Instead it seems we are now dealing with a clash of ideologies. In consequence, the objections to the trial seem to be theoretical or ad hominem attacks. Full-Text PDF In Regard to Hepel and WazerInternational Journal of Radiation Oncology, Biology, PhysicsVol. 92Issue 5PreviewTo the Editor: We read with great interest the recent editorial by Hepel and Wazer (1) on the TARGIT-A trial (2), a study that has generated a tremendous amount of scientific debate (3-6). Although we firmly believe that scientific debate is essential, the authors' main argument against Intrabeam therapy is based on a flawed premise. Full-Text PDF