Letters
tal benefit of testing on diagnostic thinking, therapy, and patient outcomes must be pursued to maximize the societal benefit of TTE. Susan A. Matulevicius, MD, MSCS Sandeep R. Das, MD, MPH Sharon C. Reimold, MD Author Affiliations: Department of Medicine, University of Texas Southwestern Medical Center, Dallas. Corresponding Author: Susan A. Matulevicius, MD, MSCS, Department of Medicine, University of Texas Southwestern, 5909 Harry Hines Blvd, Dallas, TX 75390-9047 (
[email protected]). Conflict of Interest Disclosures: None reported. 1. Mark DB, Anderson JL, Brinker JA, et al. ACC/AHA/ASE/ASNC/HRS/IAC/ Mended Hearts/NASCI/RSNA/SAIP/SCAI/SCCT/SCMR/SNMMI 2014 health policy statement on use of noninvasive cardiovascular imaging: a report of the American College of Cardiology Clinical Quality Committee. J Am Coll Cardiol. 2014;63(7):698-721. 2. Fryback DG, Thornbury JR. The efficacy of diagnostic imaging. Med Decis Making. 1991;11(2):88-94. 3. Matulevicius SA, Rohatgi A, Das SR, Price AL, DeLuna A, Reimold SC. Appropriate use and clinical impact of transthoracic echocardiography. JAMA Intern Med. 2013;173(17):1600-1607. 4. Tam JW, Nichol J, MacDiarmid AL, Lazarow N, Wolfe K. What is the real clinical utility of echocardiography? a prospective observational study. J Am Soc Echocardiogr. 1999;12(9):689-697. 5. Waggoner AD, Harris KM, Braverman AC, Barzilai B, Geltman EM. The role of transthoracic echocardiography in the management of patients seen in an outpatient cardiology clinic. J Am Soc Echocardiogr. 1996;9(6):761-768. 6. Fitch K, Bernstein SJ, Aguilar MS, et al. The RAND/UCLA Appropriateness Method User's Manual. Santa Monica, CA: RAND Corporation; 2001.
Estimating Overdiagnosis in Lung Cancer Screening To the Editor We read with interest the article by Patz et al1 investigating overdiagnosis in the National Lung Screening Trial (NLST). In their investigation, the authors found the upper bound for probability of overdiagnosis to be 11.0% to 18.5% for all lung cancers and even higher for bronchioloalveolar carcinoma (BAC) (67.6% to 78.9%). However, this risk assessment did not consider the lead- and the length-time biases.2 If data for non-BAC non–small cell lung cancer (NSCLC) (from Table 1 of the article by Patz et al1) had been plotted, we would have observed that the data actually distributed into 2 phases. First, in the active screening period, more non-BAC NSCLCs are diagnosed in the low-dose computed tomography (LDCT) arm than in the chest radiography (CXR) arm, emphasizing that some tumors are detected early by LDCT but not by CXR (overdiagnosis). During the second period beginning at the end of screening, the number of non-BAC NSCLCs became lower in the LDCT arm than in the CXR arm. This suggests the tumors not previously diagnosed by CXR were diagnosed later (lead-time bias) or grew slowly (length-time bias) but ultimately became symptomatic. Therefore, an accurate way to estimate actual overdiagnosis risk would be to use areas between the 2 curves in these 2 phases. We calculated a 5.3% larger area in the first phase than in the second one. We believe this proportion is a better estimator of overdiagnosis because it integrates time biases. Moreover, some previous lung cancer screening trials demonstrated jamainternalmedicine.com
that follow-up have to be extended up to 9 or 10 years to completely take into account both overdiagnosis and timebased biases.3,4 Conversely, numbers of diagnosed BACs in the LDCT arm are much higher than in the CXR arm during but overlap after the active screening period, up to study year 6. This illustrates that BAC are at high-risk of overdiagnosis during the screening period without effects of time-based biases. Indeed, BACs are mainly ground glass opacities with very mild growth. Using a volume-doubling time approach for nodule management (as done by the NELSON [Nederlands Leuvens Longkanker Screenings Onderzoek] trialists), would have resulted in ruling out these lesions as positive screening results.5 In summary, lung cancer screening highlights 2 different clusters. Non-BAC NSCLC could be overdiagnosed. Such a risk should be investigated and explained to patients, but it is overestimated by Patz et al.1 Otherwise, BACs are at high risk for overdiagnosis, but this risk may be easily managed by using a volume-monitoring approach and should not hide the true benefit for lung cancer screening on lung cancer mortality. Sébastien Couraud, MD, MSc Laurent Greillier, MD, PhD Bernard Milleron, MD; for the IFCT Lung Cancer Screening Group Author Affiliations: Service de Pneumologie Aiguë Spécialisée et Cancérologie Thoracique, CH Lyon Sud, Hospices Civils de Lyon, Pierre Bénite, France (Couraud); Faculté de Médecine Lyon-Sud, Université Lyon 1, Oullins, France (Couraud); Aix Marseille Univ–Assistance Publique–Hôpitaux de Marseille, Multidisciplinary Oncology & Therapeutic Innovations Department, Marseille, France (Greillier); Respiratory Disease Department, Tenon Hospital, Assistance Publique–Hôpitaux de Paris, Paris, France (Milleron); Intergroupe Francophone de Cancérologie Thoracique (IFCT), Paris, France (Milleron). Corresponding Author: Sébastien Couraud, MD, MSc, Service de Pneumologie Aiguë Spécialisée et Cancérologie Thoracique, CH Lyon Sud, Hospices Civils de Lyon, 165 Chemin du Grand Revoyet, 69495 Pierre Bénite CEDEX, France (
[email protected]). Conflict of Interest Disclosures: None reported. Group Information: Members of the IFCT Lung Cancer Screening Group are listed in Ann Oncol. 2013;24(3):586-597. 1. Patz EF Jr, Pinsky P, Gatsonis C, et al; NLST Overdiagnosis Manuscript Writing Team. Overdiagnosis in low-dose computed tomography screening for lung cancer. JAMA Intern Med. 2014;174(2):269-274. 2. Patz EF Jr, Goodman PC, Bepler G. Screening for lung cancer. N Engl J Med. 2000;343(22):1627-1633. 3. Marcus PM, Bergstralh EJ, Zweig MH, Harris A, Offord KP, Fontana RS. Extended lung cancer incidence follow-up in the Mayo Lung Project and overdiagnosis. J Natl Cancer Inst. 2006;98(11):748-756. 4. Oken MM, Hocking WG, Kvale PA, et al; PLCO Project Team. Screening by chest radiograph and lung cancer mortality: the Prostate, Lung, Colorectal, and Ovarian (PLCO) randomized trial. JAMA. 2011;306(17):1865-1873. 5. Horeweg N, van der Aalst CM, Vliegenthart R, et al. Volumetric computed tomography screening for lung cancer: three rounds of the NELSON trial. Eur Respir J. 2013;42(6):1659-1667.
To the Editor Patz et al1 attempt to report on the rate of overdiagnosis that occurred as part of the National Lung Cancer Screening Trial (NLST), however their conclusion that “the probability is 18.5%…that any lung cancer detected by LDCT JAMA Internal Medicine July 2014 Volume 174, Number 7
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Hawaii at Manoa User on 06/09/2015
1197
Letters
[low-dose computed tomography] was an overdiagnosis”1(p269) is very misleading. In the introduction, the authors correctly assert that the excess rate of cancers detected during LDCT screening compared with chest radiography is attributable to either overdiagnosis or lead time. Lead time refers to cancers that will eventually be detected in the nonscreened arm after the screening period has ended. Lead time can be beneficial (and the purpose of screening) if the cancer detected would ultimately prove harmful to the individual, or it can simply reflect a bias of screening (lead-time bias) if the cancer would never become clinically meaningful. On the basis of the available NSLT data, it is impossible to distinguish between beneficial lead time and lead-time bias. However, it is possible to estimate that computed tomographic screening provides a consistent 1- to 2-year lead-time effect over chest radiography in the NSLT trial.2 The authors do attempt to make a more accurate estimate of the overdiagnosis rate using the convolution-type model, which removes the lead-time effect. Using this model, the authors found that the number of excess lung cancers that would ultimately be discovered with lifetime follow-up was 9%, likely the true overdiagnosis rate for all lung cancers. When you further exclude diagnosis of bronchioalveolar carcinoma, the overdiagnosis rate falls to 1.2%. These findings are consistent with what is clinically recognized, ie, bronchioalveolar carcinomas are likely indolent tumors, whereas most non–small cell lung cancers are not. Unfortunately, the article by Patz et al1 repeatedly conflates the 2 terms excess rate and overdiagnosis: 18.5% represents the excess percentage of cancers detected during the screening period. The probability of overdiagnosis is much lower, closer to 1.2% after excluding bronchioalveolar carcinoma. Brian D. Gelbman, MD Daniel M. Libby, MD Author Affiliations: Division of Pulmonary and Critical Care Medicine, Weill Cornell Medical Center, New York, New York. Corresponding Author: Brian Gelbman, MD, Division of Pulmonary and Critical Care Medicine, Weill Cornell Medical Center, 635 Madison Ave, Ste 1101, New York, NY 10022 (
[email protected]). Conflict of Interest Disclosures: None reported. 1. Patz EF Jr, Pinsky P, Gatsonis C, et al; NLST Overdiagnosis Manuscript Writing Team. Overdiagnosis in low-dose computed tomography screening for lung cancer. JAMA Intern Med. 2014;174(2):269-274. 2. Grannis FW Jr. Minimizing over-diagnosis in lung cancer screening. J Surg Oncol. 2013;108(5):289-293.
To the Editor Stedman’s Medical Dictionary defines the word indolent as follows: “inactive; sluggish; painless or nearly so, said of a morbid process.”1 In the arena of cancer screening, indolent disease refers to disease that need not be detected (or treated) because it is not and never will be life-threatening. The identification of indolent disease is most definitely a harm of cancer screening, and indolent disease is one component of overdiagnosed disease. Overdiagnosis also occurs when screen1198
ing detects nonindolent disease that never would have been diagnosed in the absence of screening; that occurs because death due to other causes occurs prior to what would have been the date of symptomatic diagnosis in the absence of screening. We can refer to that as overdiagnosis due to competing causes of mortality. In the abstract of their important article quantifying overdiagnosis in lung cancer screening with low– radiation dose computed tomography, Patz et al2 incorrectly equate overdiagnosed cancers with indolent cancers. It is fair to assume that some portion of overdiagnosed lung cancers are indolent, but it must also be recognized that competing causes of mortality contribute to overdiagnosis. This is of particular importance in lung cancer screening because those most likely to be screened often have other life-threatening conditions, such as cardiovascular disease or chronic obstructive pulmonary disorder. Pamela M. Marcus, PhD, MS Author Affiliation: National Cancer Institute, Bethesda, Maryland. Corresponding Author: Pamela M. Marcus, PhD, MS, National Cancer Institute, 9609 Medical Center Dr, Room 4E-608, Bethesda, MD 20892-9763 (
[email protected]). Conflict of Interest Disclosures: None reported. 1. Stedman’s Medical Dictionary. 28th ed. Baltimore, MD: Lippincott, Williams & Wilkins; 2006. 2. Patz EF Jr, Pinsky P, Gatsonis C, et al; NLST Overdiagnosis Manuscript Writing Team. Overdiagnosis in low-dose computed tomography screening for lung cancer. JAMA Intern Med. 2014;174(2):269-274.
In Reply We appreciate the letters by Drs Marcus, Couraud et al, and Gelbman and Libby in response to our recent study on estimating overdiagnosis in low-dose computed tomographic screening for lung cancer.1 We agree with Dr Marcus that not all cases of overdiagnosis in a cancer screening program are necessarily caused by indolent tumors and should have made this clearer in our article. Given the natural history of most clinically apparent lung cancers, however, we estimate that the majority of overdiagnosis cases in the National Lung Cancer Screening Trial (NLST) were indolent tumors but acknowledge that this may not be exactly the case in other cancers or in population-based screening programs, with extended follow-up. Both letters by Couraud et al and Gelbman and Libby raised concerns about the magnitude of overdiagnosis we estimated in screening trials for lung cancer. In our article we calculated 2 standard measures of overdiagnosis using data from the NLST and then fit a standard convolution model to estimate overdiagnosis under various screening conditions. As the authors suggest, there are different methods by which one could approach this problem and probably arrive at slightly different values, but we do not believe the end result would change substantially or affect the overall interpretation of our results. Separating the analysis into bronchioloalveolar cell carcinoma (BAC) and non-BAC non–small cell lung cancer provides important scientific insights into the phenomenon of overdiagnosis. However, for the goal of informing patients and clinicians of overdiagnosis as a potential limitation of screening (ie, unnecessary treatment, morbidity and mortality in rare
JAMA Internal Medicine July 2014 Volume 174, Number 7
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Hawaii at Manoa User on 06/09/2015
jamainternalmedicine.com
Letters
cases, surveillance follow-up studies, costs, and anxiety), the rates for all non–small cell lung cancers are more relevant. While Courand et al suggest that patients with BAC may be managed by using volumetric monitoring, this approach has yet to be validated. We recognize that much work is being done in this area, but until better noninvasive tools are developed to clearly identify individuals with nonaggressive tumors that do not require treatment, all of these patients are likely to be treated, with similar potential for harms of screening. Gelbman and Libby suggest that the convolution model provides a “more accurate estimate of the overdiagnosis” than the calculations derived directly from the NLST data. Both methods have strengths and weaknesses as we have documented in our article; the validity of the assertion that the convolution model approach is more accurate is not clear. In fact, potential problems with relying on such statistical models to estimate overdiagnosis have been recently detailed.2 We believe that the calculations of overdiagnosis reported in our study accurately reflect excess rates of cancer in a screening trial, and although they may not be exactly the same as in a mass screening program, these data provide a general framework to inform patients and clinicians of this potential limitation when considering screening for lung cancer. Edward F. Patz Jr, MD Paul Pinsky, PhD Barnett S. Kramer, MD Author Affiliations: Department of Radiology, Duke University Medical Center, Durham, North Carolina (Patz); National Cancer Institute, Bethesda, Maryland (Pinsky, Kramer). Corresponding Author: Edward F. Patz Jr, MD, Department of Radiology, Duke University Medical Center, PO Box 3808, Durham, NC 27710 (patz0002@mc .duke.edu). Conflict of Interest Disclosures: None reported. 1. Patz EF Jr, Pinsky P, Gatsonis C, et al; NLST Overdiagnosis Manuscript Writing Team. Overdiagnosis in low-dose computed tomography screening for lung cancer. JAMA Intern Med. 2014;174(2):269-274. 2. Zahl PH, Jørgensen KJ, Gøtzsche PC. Lead-time models should not be used to estimate overdiagnosis in cancer screening [published online March 4, 2014]. J Gen Intern Med. doi:10.1007/s11606-014-2812-2.
Percutaneous Coronary Intervention vs Medical Treatment in Stable Angina: The Never-Ending Story To the Editor We read with interest the article by Stergiopoulos et al,1 who conclude that in patients with stable coronary artery disease and objectively documented ischemia, percutaneous coronary intervention (PCI) was not associated with a reduction in clinical end points compared with medical therapy (MT) alone. While interesting, this meta-analysis might be flawed in several ways. Study inclusion: Of the 5 randomized trials included in the current meta-analysis, 4 (Medicine, Angioplasty, or Surgery Study II [MASS-II], Clinical Outcomes Using Revascularization and Aggressive Drug Evaluation [COURAGE], Hambrecht et al, and Bypass Angioplasty Revascularization 2 Diabetes [BARI-2D]) do not reflect current clinical practice anymore jamainternalmedicine.com
with regard to ischemia-driven strategy of both primary PCI and PCI deferral with optimal MT, with a minimal use of contemporary revascularization strategies.2 Ischemia evaluation: Various methods for documenting ischemia were used across the studies with greater use rate of treadmill stress tests and nuclear imaging. Pooled together, as an ischemia-guided approach, treadmill exercise and invasive assessment, through fractional flow reserve, results in an operation conceptually wrong; stress and perfusion imaging have poor concordance with fractional flow reserve and tend to underestimate or overestimate the functional importance of coronary stenosis. Therefore, a serious issue of heterogeneity arises, with each ischemia test associated with different sensitivity and specificity profile, which translates into a nonhomogeneous classification of patients undergoing coronary angiography.3 This is particularly true in the absence of a flow-limiting lesion or when the lesion is not associated with significant ischemia burden (fractional flow reserve ≤0.80), and the patient is directed to MT.4 Ischemia threshold: Broad inclusion (any patients with documented signs of ischemia) allowed the patients with mild and mild to moderate ischemia to enter the analysis, with the COURAGE trial accounting for most of the studied population. It is well recognized that the incidence of the primary end point, regardless of definitions, correlates with the degree of ischemia, with extensive ischemia associated with the highest risk; therefore, inclusion of low-risk patients might drive the results toward no benefit of PCI over MT. Furthermore, the results of the current study are in contradiction to a recently published meta-analysis that demonstrated a 44% mortality reduction with PCI compared with MT in stable patients selected on the basis of ischemia.5 Patient crossover: Conclusions regarding recurrent and persistent angina are merely speculative and not justified because in a prespecified follow-up of 5 years, a considerable number of the MT patients (approximately 31%) had undergone revascularization, potentially leading once again to spurious findings. In conclusion, the heterogeneity that arises from pooling together studies using different testing for ischemia evaluation might have substantially biased the results toward the neutrality of effect of PCI vs MT. Eliano Pio Navarese, MD, PhD Mariusz Kowalewski, MD Harry Suryapranata, MD Author Affiliations: Department of Cardiology, Radboud University Nijmegen Medical Center, Nijmegen, the Netherlands (Navarese, Suryapranata); 10th Military Research Hospital and Polyclinic, Bydgoszcz, Poland (Kowalewski); Systematic Investigation and Research on Interventions and Outcomes (SIRIO) MEDICINE Research Network (Navarese, Kowalewski). Corresponding Author: Eliano Pio Navarese, MD, PhD, Department of Cardiology, Radboud University Nijmegen Medical Center, Geert Grooteplein 10, PO Box 9101, 6500 HB Nijmegen, the Netherlands (
[email protected]). Conflict of Interest Disclosures: None reported. 1. Stergiopoulos K, Boden WE, Hartigan P, et al. Percutaneous coronary intervention outcomes in patients with stable obstructive coronary artery disease and myocardial ischemia: a collaborative meta-analysis of contemporary randomized clinical trials. JAMA Intern Med. 2014;174(2):232-240.
JAMA Internal Medicine July 2014 Volume 174, Number 7
Copyright 2014 American Medical Association. All rights reserved.
Downloaded From: http://archinte.jamanetwork.com/ by a University of Hawaii at Manoa User on 06/09/2015
1199