Axel Bauer, M.D. Konstantinos D. Rizas, M.D. Ludwig-Maximilians Universität Munich, Germany axel.bauer@med.uni-muenchen.de 1. Materson BJ, Reda DJ, Cushman WC, et al. Single-drug therapy for hypertension in men a comparison of six antihypertensive agents with placebo. N Engl J Med 1993;328:914-21. 2. Zuern CS, Eick C, Rizas KD, et al. Impaired cardiac baroreflex sensitivity predicts response to renal sympathetic denervation in patients with resistant hypertension. J Am Coll Cardiol 213;62:2124-3. DOI: 1.156/NEJMc145677 The authors reply: We agree that the lack of efficacy in our study may relate to the catheter that was used or the way in which it was used. However, this was the same catheter that had been used in previous trials, and each procedure was supervised by an experienced proctor and performed per protocol instructions. If those instructions did not allow for aggressive enough denervation owing to an insufficient number of ablations, lack of four-quadrant ablations, or other technical features, that might explain the failure to lower blood pressure significantly. It is indeed possible that multielectrode devices could improve efficacy by ensuring appropriate vessel contact and a sufficient number and most effective pattern of ablations, though the safety of more aggressive ablation would need to be reconfirmed. The immediate assessment of the adequacy of renal denervation (e.g., from imaging or biomarkers) would provide reassurance that the procedure was efficacious in real time and potentially allow for dose modulation with adjustment for efficacy. Although it is possible that operator experience affected the results, all the operators were experienced proceduralists who were being carefully proctored. Within the trial, we found no evidence that either operator volume or operator experience affected the results. It is conceivable that the effects of medication changes occurring before screening had not been fully manifested. However, even after the exclusion of patients with medication changes before randomization or during the trial, the overall results remained unchanged. We are proud that we enrolled a substantial number of black patients. The lack of these patients in previous trials was a limitation. We hypothesized that there would be a greater, not lesser, benefit among black patients than among non-black patients, but our findings did not confirm this hypothesis. Detailed analysis of ambulatory blood-pressure monitoring did not find a significant benefit regardless of race. 1 We agree that various selection criteria and characteristics of our patient population such as the exclusion of patients with white-coat hypertension, the inclusion of obese patients, and a variety of other baseline characteristics or medications could account for the null results of this trial, as compared with the findings of previous trials. Identifying which patients may have a response to renal-artery denervation remains a major challenge. We did not find a significant benefit in any prespecified subgroup in the evaluation of ambulatory blood pressure. 1 Potentially, baroreflex sensitivity or other baseline physiological measurements will help to identify who might benefit from renal-artery denervation in future clinical trials. Deepak L. Bhatt, M.D., M.P.H. Brigham and Women s Hospital Heart and Vascular Center Boston, MA dlbhattmd@post.harvard.edu George L. Bakris, M.D. University of Chicago Medicine Chicago, IL Since publication of their article, the authors report no further potential conflict of interest. 1. Bakris GL, Townsend RR, Liu M, et al. Impact of renal denervation on 24-hour ambulatory blood pressure: results from SYMPLICITY HTN-3. J Am Coll Cardiol 214 May 17 (Epub ahead of print). DOI: 1.156/NEJMc145677 Multitarget Stool DNA Testing for Colorectal-Cancer Screening To the Editor: Imperiale et al. (April 3 issue) 1 report higher sensitivity and lower specificity with the use of a multitarget stool DNA test than with a commercial fecal immunochemical test (FIT; OC FIT-CHEK, Polymedco) for detecting colorectal neoplasms in participants who were un- 184 n engl j med 371;2 nejm.org july 1, 214
correspondence Table 1. Test Performance in Two Studies Involving Persons Who Underwent Screening Colonoscopy in the United States and Germany.* Test Multitarget stool DNA test FIT Sensitivity Results of Multitarget Stool DNA Test and FIT Reported in U.S. Study Cutoff Adapted to Yield Same Specificity as FIT in U.S. Study Results of FIT Derived from Additional Analyses of German Study percent (95 percent confidence interval) Cutoff Adapted to Yield Same Specificity as Multitarget Stool DNA Test in U.S. Study Colorectal cancer 92.3 (83. 97.5) 86.7 (59.5 98.3) Advanced adenoma 42.4 (38.9 46.) 41.6 (34.8 48.6) Nonadvanced adenoma 17.2 (15.9 18.6) 2.6 (16.7 24.9) Specificity: no advanced neoplasm Sensitivity 86.6 (85.9 87.2) 85.7 (83.9 87.4) Colorectal cancer 73.8 (61.5 84.) 73.3 (44.9 92.2) Advanced adenoma 23.8 (2.8 27.) 29.5 (23.4 36.2) Nonadvanced adenoma 7.6 (6.7 8.6) 8.3 (5.8 11.5) Specificity: no advanced neoplasm 94.9 (94.4 95.3) 94.9 (93.7 95.9) * CI denotes confidence interval, and FIT fecal immunochemical test. In the U.S. study, these data refer to a combined category of advanced adenomas and sessile serrated polyps measuring 1 cm or more. dergoing screening colonoscopy. The sensitivity and specificity of the FIT depend on the specific cutoff value used to define test positivity. To compare diagnostic performance, it is helpful to use cutoff levels yielding similar specificity. We have evaluated a similar FIT (OC-SENSOR, Eiken Chemical) in participants who underwent screening colonoscopy in Germany. 2 Adaptation of the FIT cutoff value in our study to yield the same specificity as that reported for the FIT used by Imperiale et al. (94.9%) indicated an essentially equivalent performance of the FITs in both studies (Table 1). Adaption of the cutoff value to yield specificity similar to that reported for the multitarget stool DNA test used by Imperiale et al. (86.6%) resulted, apart from some variation that is likely to be random, in very similar sensitivities for both tests (Table 1). Taken together, these results suggest that the high sensitivities reported for the multitarget stool DNA test can be achieved at similar specificity and much lower costs with the use of FITs. Hermann Brenner, M.D., M.P.H. Simone Werner, M.Sc. Hongda Chen, M.Sc. German Cancer Research Center Heidelberg, Germany The German Cancer Research Center reports receiving grant support from Eiken Chemical for testing the OC-SENSOR. No other potential conflict of interest relevant to this letter was reported. 1. Imperiale TF, Ransohoff DF, Itzkowitz SH, et al. Multitarget stool DNA testing for colorectal-cancer screening. N Engl J Med 214;37:1287-97. 2. Brenner H, Tao S. Superior diagnostic performance of faecal immunochemical tests for haemoglobin in a head-to-head comparison with guaiac based faecal occult blood test among 2235 participants of screening colonoscopy. Eur J Cancer 213;49: 349-54. To the Editor: The approach adopted by Imperiale et al. to report the findings of a large study comparing a widely adopted FIT with a multitarget stool DNA test in a population at average risk for colorectal cancer does not convey, n engl j med 371;2 nejm.org july 1, 214 185
Table 1. Extrapolation of Findings to a Population of 1, Persons Undergoing Screening with a Multitarget Stool DNA Test and FIT. Variable Sample size Multitarget Stool DNA Test number FIT Inadequate samples 625 31 Adequate samples 9375 9969 Positive tests 15 681 Cancers detected Colorectal 56 47 Advanced adenomas 31 175 Needed to screen colorectal cancer advanced adenoma Needed to undergo colonoscopy colorectal cancer advanced adenoma 179 213 33 57 27 14 5 4 in our opinion, the information that is relevant to assess the role of screening tests. The number of persons who would need to be screened is not really informative if the cost and screening interval are not known for both tests. The number of persons who would need to be referred for colonoscopy on the basis of positive test results to detect one advanced neoplasm (which is actually lower for FIT than for stool DNA) would allow a comparison of the two methods, based on the same unit cost, even if it may vary across jurisdictions. The extrapolation of the findings to a population of 1, persons undergoing screening does not account for the expected rate of inadequate samples, which was higher with the stool DNA test (6.2%) than with the FIT (.3%). Table 1, which is based on the reported prevalence, sensitivity estimates, and rates of inadequate samples and positivity, would offer a more appropriate comparative picture of the expected effect of screening. Carlo Senore, M.D. Nereo Segnan, M.D., M.P.H. Piedmont Reference Center for Epidemiology and Cancer Prevention Turin, Italy carlo.senore@cpo.it To the Editor: Imperiale and colleagues evaluate the use of a single-application multitarget stool DNA test for colorectal-cancer screening and compare its performance characteristics with those of the FIT. Two important points warrant further discussion. First, the investigators used a FIT cutoff value of 1 ng of hemoglobin per milliliter of buffer (equivalent to 2 μg of hemoglobin per gram of feces) 1 instead of a lower cutoff value; this may explain the decreased sensitivity of FIT (73.8%) for colorectal cancer in their study. Recently, our systematic review and meta-analysis showed that FITs with a cutoff value of less than 2 μg per gram had a sensitivity of 89% and a specificity of 91% for colorectal cancer in an asymptomatic, average-risk population. 2 Second, the overall positive rate for the stool DNA test was higher (16.1%) than that of the FIT (7.%); this was mainly due to the higher false positive rate for the stool DNA test. This finding is an important issue to consider given the unclear screening interval for stool DNA testing and the lack of colonoscopy resources across the world. Jeffrey K. Lee, M.D. Jonathan P. Terdiman, M.D. University of California, San Francisco San Francisco, CA jeff.lee@ucsf.edu Douglas A. Corley, M.D., Ph.D. Kaiser Permanente Oakland, CA 1. Fraser CG, Allison JE, Halloran SP, Young GP. A proposal to standardize reporting units for fecal immunochemical tests for hemoglobin. J Natl Cancer Inst 212;14:81-4. 2. Lee JK, Liles EG, Bent S, Levin TR, Corley DA. Accuracy of fecal immunochemical tests for colorectal cancer: systematic review and meta-analysis. Ann Intern Med 214;16:171-81. 186 n engl j med 371;2 nejm.org july 1, 214
correspondence To the Editor: In the article by Imperiale and colleagues, the calculation of the number of persons who would need to be screened is not correct. The authors used the total number of persons in a theoretical population (1,) and the number of detected cancers (65 with the use of colonoscopy, 6 with DNA testing, and 48 with FIT). The ratio 1, divided by the number of detected cancers is used as the number of persons who would need to be screened. With the use of that calculation, the DNA test provides about the same result as colonoscopy. This reasoning is not valid, since the DNA test also provided 732 false positive results, whereas the FIT produced 248 false positive results. Both the DNA test and the FIT have a very low positive predictive value of 6 lesions per 792 persons with positive results (7.6%) and 48 lesions per 296 persons with positive results (16.2%), respectively. A patient who receives a positive FIT result has a greater chance of having the disease than a patient who receives a positive DNA result. Because colorectal-cancer screening is performed to detect cancers, it is better to have a relatively small number of false positive results. The DNA test presented here seems to be better at ruling out the disease than diagnosing it. John H.M. Souverijn, Ph.D. Stichting Huisartsenlaboratorium Alphen aan den Rijn, the Netherlands jhmsouverijn@kpnmail.nl The Authors Reply: We appreciate the analyses performed by Brenner and colleagues. However, the receiver operating characteristic curve in Figure 3 of our article which was generated by direct comparison of both tests in the same participants clearly shows that FIT alone does not achieve the sensitivity of the combined markers across a broad range of specificity. At matched specificities of 86.6%, sensitivities for the detection of cancer were 92.3% with multitarget stool DNA and 76.9% with FIT (P =.15), with the use of a cutoff of 23 ng of hemoglobin per milliliter of buffer and a recognized technical uncertainty of FIT below 5 ng. 1 The rate of unsatisfactory samples noted by Senore and Segnan was due to stringent criteria for study quality control that required patients to undergo colonoscopy within 9 days after enrollment; this often precluded collection of a second specimen. This limitation would not occur in clinical practice. The few cases in which the collection device leaked were addressed by improving the seal. With regard to the concerns of Lee and colleagues: we used the FIT manufacturer s cutoff value of 1 ng of hemoglobin per milliliter of buffer. The observed sensitivity of FIT (73.8%) is consistent with the pooled estimate of 71% from their recent meta-analysis of studies in which colonoscopy was the reference standard. 2 They correctly point out the higher false positive rate due to the DNA component of the test. However, two thirds of the false positive results were associated with nonadvanced polyps, which most clinicians would not consider false positive findings because surveillance colonoscopy may be recommended. 3 Also, the lack of detection of adenomas by means of colonoscopy 4 may further inflate the false positive rate. In practice, a positive stool test result would be expected to enhance the vigilance of the endoscopist and perhaps reduce the false positive rate of the multitarget stool DNA test. Furthermore, in a screening program, the cumulative false positive rate is a function of both point specificity and testing frequency. For example, a point-in-time 12% false positive rate for multitarget stool DNA testing performed every 3 years would translate into a 4% average annual rate for FIT. Subsequent cost-effectiveness analyses may shed light on this important issue. We agree with Souverijn that the multitarget stool DNA test is better than FIT at ruling out colorectal cancer because of its higher sensitivity and resulting higher negative predictive value. Contrary to what Souverijn implies, however, the calculation of the number needed to screen does consider the false positive test results. Thomas F. Imperiale, M.D. Indiana University School of Medicine Indianapolis, IN David F. Ransohoff, M.D. University of North Carolina at Chapel Hill Chapel Hill, NC Steven H. Itzkowitz, M.D. Icahn School of Medicine at Mount Sinai New York, NY n engl j med 371;2 nejm.org july 1, 214 187
Since publication of their article, the authors report no further potential conflict of interest. 1. Evaluation report: immunochemical faecal occult blood tests (report no. CEP942). Alfreton, United Kingdom: NHS Purchasing and Supply Agency, November 29 (http://www.cancerscreening.nhs.uk/bowel/ifobt.pdf). 2. Lee JK, Liles EG, Bent S, Levin TR, Corley DA. Accuracy of fecal immunochemical tests for colorectal cancer: systematic review and meta-analysis. Ann Intern Med 214;16:171-81. 3. Lieberman DA, Rex DK, Winawer SJ, Giardiello FM, Johnson DA, Levin TR. Guidelines for colonoscopy surveillance after screening and polypectomy: a consensus update by the US Multi- Society Task Force on Colorectal Cancer. Gastroenterology 212;143:844-57. 4. Rex DK, Cutler CS, Lemmel GT, et al. Colonoscopic miss rates of adenomas determined by back-to-back colonoscopies. Gastroenterology 1997;112:24-8. Interferon Alfa Therapy in CALR-Mutated Essential Thrombocythemia To the Editor: Somatic mutations in the gene encoding calreticulin (CALR) were recently described in the majority of patients with myeloproliferative neoplasms without mutations in the Janus kinase 2 gene (JAK2). 1,2 We and other groups have previously shown that interferon alfa was able to reduce the JAK2-mutated clone and to induce molecular complete responses in patients with myeloproliferative neoplasms. 3 Because the effect of interferon alfa on mutant clones in myeloproliferative neoplasms with other molecular lesions is still questioned and because calreticulin was involved in resistance to interferon alfa in hepatitis B virus infection, 4 we wish to report on the positive effect of interferon alfa therapy in patients with CALR-mutated myeloproliferative neoplasms. Two unrelated patients with high-risk essential thrombocythemia and without mutations in JAK2 or the thrombopoietin receptor gene (MPL) who were being followed in our institution were treated off-label with peginterferon alfa-2a owing to their relatively young age and a history of thrombosis, in agreement with local and international guidelines. 5 Sequential samples were stored after the patients provided written informed consent for somatic mutation analyses. We used direct Sanger sequencing to detect CALR mutations and fragment analysis to calculate the mutant allele burden. 1 B Patient 2: CALR Mutation p.l367fs*46 A Patient 1: CALR Mutation p.k385fs*47 Peginterferon 2 alfa-2a 6 1 Hydroxyurea Peginterferon alfa-2a 6 Platelet Count (x1 9 /liter) 16 12 8 4 42 46 6 Platelet count Mutant allele burden.5 1 2 3 4 5 6 7 8 Years since Diagnosis 12 1 5 4 3 2 1 CALR Mutant Allele Burden (%) Platelet Count (x1 9 /liter) 8 6 4 2 46 48 Platelet count Mutant allele burden 3.2 2.8 1 2 3 4 5 6 7 8 9 1 11 12 13 14 Years since Diagnosis 5 4 3 2 1 CALR Mutant Allele Burden (%) Figure 1. Evolution of Platelet Count and CALR Mutation Burden during Treatment in Two Patients with Essential Thrombocythemia. The treatment periods with hydroxyurea or peginterferon alfa-2a are indicated by arrows. CALR mutations were identified with the use of direct Sanger sequencing, and the CALR mutant allele burden was calculated with the use of DNA fragment analysis and area-underthe-peak measurement: % of CALR mutant allele burden = (mutated CALR [nonmutated CALR + mutated CALR]) 1. Patient 1 harbored a 5-base insertion (p.k385fs*47) and received peginterferon alfa-2a at a dose of 9 μg per week for the first 6 months and then 9 μg every other week. Patient 2 had a 52-base deletion (p.l637fs*46) and was treated with 18 μg of peginterferon alfa-2a every 2 weeks for the first year, followed by 18 μg every 3 weeks for the second year. 188 n engl j med 371;2 nejm.org july 1, 214