1496
Letters to the Editor
interpret the estimates suggested by Suk et al. with caution. CONFLICT OF INTEREST The authors declare no conflict of interest. REFERENCES 1. Suk KT, Kim DJ, Kim CH et al. A prospective nationwide study of drug-induced liver injury in Korea. Am J Gastroenterol 2012;107:1380–7. 2. Park H, Jang I, Lee S. Hepatotoxic events associated with herbal medicinal products, folk remedies and food supplements in Korea. J Korean Oriental Med 2005;26:152–65. 1
Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA; 2Department of Internal Medicine, College of Korean Medicine, Woosuk University, Wanju, Jeollabukdo, South Korea; 3Department of Internal Medicine, College of Korean Medicine, Dongguk University, Goyang, Gyunggido, South Korea. Correspondence: Sunjae Bae, KMD, MPH, Johns Hopkins Bloomberg School of Public Health, 615 North Wolfe Street, Baltimore, Maryland 21205, USA. E-mail: [email protected]
Response to Bae et al.
of causative agents specific to each hospital for direct comparison among hospitals. We analyzed all of the patients (371 cases), and in the χ2-analysis there was no difference in the proportion of causative agents among the hospitals (P = 0.199) (Table 1). In this study, we included all kinds of university hospitals, from relatively small university hospitals to large ones, in order to evaluate the incidence rate of DILI more accurately (2). Therefore, heterogeneity in the number of DILI cases per 100 beds at each hospital might have occurred. The number of hepatologists in each hospital was another reason for the heterogeneity in the number of DILI cases per 100 beds at each hospital. In most hospitals, one hepatologist in each hospital participated in this study. Other hepatologists who did not participate in this study might have shown relatively little interest in enrolling or transferring patients to the investigator. Hospitals 4, 8, and 14 had three hepatologists, which is a larger number than that of the other hospitals (which had only one or two hepatologists). In hospitals 4, 8, and 14, the number of DILI cases per 100 beds
was lower compared with the other hospitals. In hospital 17, there was only one hepatologist, who took as much interest as a principal investigator in this study. Therefore, the number of DILI cases per 100 beds was the highest at the hospital with the deeply engaged hepatologist (41 (11.1%)). As Bae et al. mentioned, the causes of DILI displayed a large inconsistency from the interim data, which included 304 cases (81.9% of the total cases). However, in the χ2-analysis of the final number of patients (n = 371), there was no difference in the causes of DILI (P = 0.199). Moreover, if we were to include the 25 patients who had been excluded because the patients did not meet the time-to-onset criteria of the RUCAM - herbs (n = 21) and medication (n = 4) from the original cases - the inconsistency from the interim data would be reduced. With regard to the low level of agreement with the previous literature (3), we found that previous literature has referenced retrospectively analyzed studies with a small number of cases (68 cases/84 months and
Ki Tae Suk, MD1 doi:10.1038/ajg.2014.240
Table 1. Number of cases according to causative agents
To the Editor: We appreciate Bae et al. (1) for their interest in our article, and their comments regarding the accuracy and precision of the study’s estimates. Bae et al. make a very good point and provide constructive comments. They calculated the number of drug-induced liver injury (DILI) cases per 100 beds at each hospital from interim data, which included 304 cases (81.9% of the total cases), to enable a direct comparison among hospitals. We did not use the number of DILI cases per 100 beds at each hospital for the direct comparison among hospitals, because there is great variation in the bed-utilization rate (70–99%) based on certain characteristics of the hospital, including its size ( < 1,000 beds vs. > 1,000 beds or < 700 beds vs. > 700 beds), location (rural vs. urban), and ownership structure (national hospital vs. private hospital), as well as the season of the year. We had performed a statistical analysis by using the proportion The American Journal of GASTROENTEROLOGY
Medications (n=101)
Number of cases Others (n=270)
Total (n=371)
P value 0.199
Hospital 1
6 (40)
9 (60)
15
Hospital 2
6 (26)
17 (74)
23
Hospital 3
5 (18)
23 (82)
28
Hospital 4
0 (0)
3 (100)
3
Hospital 5
11 (44)
14 (56)
25
Hospital 6
4 (33)
8 (67)
12
Hospital 7
5 (42)
7 (58)
12
Hospital 8
5 (36)
9 (64)
14
Hospital 9
2 (13)
13 (87)
15
Hospital 10
8 (30)
19 (70)
27
Hospital 11
8 (21)
31 (79)
39
Hospital 12
2 (12)
15 (88)
17
Hospital 13
7 (30)
16 (70)
23
Hospital 14
4 (17)
19 (83)
23
Hospital 15
7 (19)
30 (81)
37
Hospital 16
4 (24)
13 (76)
17
Hospital 17
17 (42)
24 (58)
41
VOLUME 109 | SEPTEMBER 2014 www.amjgastro.com
Letters to the Editor
48 cases/72 months), and enrollment criteria for the studies differed from those of our prospective study (4,5). These references enrolled patients with all kinds of DILI. However, our study enrolled hospitalized adults with suspected DILI as defined by alanine aminotransferase > 3× upper normal limit or total bilirubin > 2× upper normal limit. We believe that these retrospective references with a small number of cases have not generated strong evidence or a reliable reference range. With regard to Table 4, we classified causative agents by forms. In the causative agents, we described all kinds of suspected causal agents. As the first prospective multicenter study in Asia on this topic, we believe that our study produced new knowledge about DILI. CONFLICT OF INTEREST The author declares no conflict of interest. REFERENCES 1. Bae S, Jang I, Ham C-h. Concerns on the precision of the estimation and the quality management of the data. Am J Gastroenterol 2014;109:1495–6 (this issue). 2. Suk KT, Kim DJ, Kim CH et al. A prospective nationwide study of drug-induced liver injury in Korea. Am J Gastroenterol 2012;107:1380–7. 3. Paik H, Jang I, Lee S. Hepatotoxic events associated with herbal medicinal products, folk remedies and food supplements in Korea. J Korean Oriental Med 2005;26:152–65. 4. Kim JB, Shon JH, Lee HL. et al. Clinical characteristics of acute toxic liver injury. Clin Mol Hepatol 2004;10:125–34. 5. Seo JC, Jeon WJ, Park SS. et al. Clinical experience 48 acute toxic hepatitis patients. Clin Mol Hepatol 2006;12:74–81. 1
Department of Internal Medicine, Hallym University College of Medicine, Chuncheon, South Korea. Correspondence: Ki Tae Suk, MD, Department of Internal Medicine, Hallym University College of Medicine, Gyo-dong, Chuncheon, South Korea. E-mail: [email protected]
Level of Evidence Supporting the Use of Thiopurine in Crohn’s Disease Tetsuji Fujita, MD1 doi:10.1038/ajg.2014.191
© 2014 by the American College of Gastroenterology
To the Editor: On the basis of a metaanalysis of 10 cohort studies, Chatu et al. (1) argue that their findings support the concept that the use of thiopurine analogs, such as azathioprine (AZA) and 6-mercaptopurine, may impact significantly on disease progression, reducing the need for first intestinal surgery by up to 40%. Although sensitivity analysis was performed to reduce the heterogeneity, there are concerns about the inconsistency of the studies, which is not likely to have been measured by a heterogeneity test. For instance, although stricture (hazard ratio (HR) 12.01, 95% confidence interval (CI) 5.97–24.17) and penetration (HR 10.77, 95% CI 4.87–23.80) of the affected intestine were important predictors of major abdominal surgery (2), 1 of the 10 studies included in the meta-analysis excluded patients with stricturing or penetrating Crohn’s disease (3). Recent cohort studies indicate much wider usage and early use of thiopurine analogs and fewer surgical resection rates, but these findings do not necessarily lead to the causal relationship between thiopurine uses and decreased surgical rates, because treatment guidelines were revised with advances in medical and surgical treatment during the study period. Level I evidence is based on the results of large-scale randomized controlled trials (RCTs) or a meta-analysis of RCTs with a narrow CI, and level II evidence is derived from small RCTs. Although aggregating several level II studies can provide level I evidence, the result of a meta-analysis of level III data from nonrandomized comparative cohort studies and level IV data from nonrandomized historical cohort studies would not shift these data to a higher level (4). In an updated Cochrane review of 13 RCTs including 1,211 patients with the primary outcome of the efficacy of AZA or 6-mercaptopurine for induction of remission in active Crohn’s disease, AZA and 6-mercaptopurine offered no advantage over placebo for induction of remission or clinical improvement of active Crohn’s disease (5). In a recent RCT with the primary outcome of the efficacy of use of AZA within 6 weeks of diagnosis of Crohn’s disease, AZA did not reduce the rate for intestinal surgery at 3 years after treatment (6). Taken together, there is no high-level evidence that overall and
early use of thiopurine in Crohn’s disease reduces the rate for intestinal surgery. CONFLICT OF INTEREST
The author declares no conflict of interest. REFERENCES 1. Chatu S, Subramanian V, Saxena S et al. The role of thiopurines in reducing the need for surgical resection in Crohn’s disease: a systematic review and meta-analysis. Am J Gastroenterol 2014;109:23–34. 2. Peyrin-Biroulet L, Oussalah A, Williet N et al. Impact of azathioprine and tumour necrosis factor antagonists on the need for surgery in newly diagnosed Crohn’s disease. Gut 2011;60:930–6. 3. Picco MF, Zubiaurre I, Adluni M et al. Immunomodulators are associated with a lower risk of first surgery among patients with nonpenetrating non-stricturing Crohn’s disease. Am J Gastroenterol 2009;104:2754–9. 4. Fujita T. Levels of evidence for laparoscopic surgery for colorectal cancer. J Am Coll Surg 2011;212:269–70. 5. Chande N, Tsoulis DJ, MacDonald JK. Azathioprine or 6-mercaptopurine for induction of remission in Crohn’s disease. Cochrane Database Syst Rev 2013;4:CD000545. 6. Cosnes J, Bourrier A, Laharie D et al. Early administration of azathioprine vs conventional management of Crohn’s disease: a randomized controlled trial. Gastroenterology 2013;145:758–65. 1
Department of Surgery, Jikei University of School of Medicine, Tokyo, Japan. Correspondence: Tetsuji Fujita, MD, Department of Surgery, Jikei University School of Medicine, 3-25-8 Nishishinbashi, Minato-ku, Tokyo 105-8461, Japan. E-mail: [email protected]
Response to Fujita Sukhdev Chatu, MBBS, MRCP, MD1, Venkataraman Subramanian, MBBS, MRCP, MD1, Sonia Saxena, MBBS, MSc, MD, FRCGP1 and Richard Pollok, MBBS, FRCP, BSc, PhD, DTM&H1 doi:10.1038/ajg.2014.192
To the Editor: We read with interest the concerns raised by Fujita (1) about our meta-analysis of the long-term impact of thiopurines on surgical resection in Crohn’s disease (2). We have never stated that this paper provides level 1 evidence and made the study limitations clear. However, as there are no randomized controlled trials in this area, the best available evidence is derived from pooling cohort studies. The American Journal of GASTROENTEROLOGY
1497
Letters to the Editor
interpret the estimates suggested by Suk et al. with caution. CONFLICT OF INTEREST The authors declare no conflict of interest. REFERENCES 1. Suk KT, Kim DJ, Kim CH et al. A prospective nationwide study of drug-induced liver injury in Korea. Am J Gastroenterol 2012;107:1380–7. 2. Park H, Jang I, Lee S. Hepatotoxic events associated with herbal medicinal products, folk remedies and food supplements in Korea. J Korean Oriental Med 2005;26:152–65. 1
Johns Hopkins Bloomberg School of Public Health, Baltimore, Maryland, USA; 2Department of Internal Medicine, College of Korean Medicine, Woosuk University, Wanju, Jeollabukdo, South Korea; 3Department of Internal Medicine, College of Korean Medicine, Dongguk University, Goyang, Gyunggido, South Korea. Correspondence: Sunjae Bae, KMD, MPH, Johns Hopkins Bloomberg School of Public Health, 615 North Wolfe Street, Baltimore, Maryland 21205, USA. E-mail: [email protected]
Response to Bae et al.
of causative agents specific to each hospital for direct comparison among hospitals. We analyzed all of the patients (371 cases), and in the χ2-analysis there was no difference in the proportion of causative agents among the hospitals (P = 0.199) (Table 1). In this study, we included all kinds of university hospitals, from relatively small university hospitals to large ones, in order to evaluate the incidence rate of DILI more accurately (2). Therefore, heterogeneity in the number of DILI cases per 100 beds at each hospital might have occurred. The number of hepatologists in each hospital was another reason for the heterogeneity in the number of DILI cases per 100 beds at each hospital. In most hospitals, one hepatologist in each hospital participated in this study. Other hepatologists who did not participate in this study might have shown relatively little interest in enrolling or transferring patients to the investigator. Hospitals 4, 8, and 14 had three hepatologists, which is a larger number than that of the other hospitals (which had only one or two hepatologists). In hospitals 4, 8, and 14, the number of DILI cases per 100 beds
was lower compared with the other hospitals. In hospital 17, there was only one hepatologist, who took as much interest as a principal investigator in this study. Therefore, the number of DILI cases per 100 beds was the highest at the hospital with the deeply engaged hepatologist (41 (11.1%)). As Bae et al. mentioned, the causes of DILI displayed a large inconsistency from the interim data, which included 304 cases (81.9% of the total cases). However, in the χ2-analysis of the final number of patients (n = 371), there was no difference in the causes of DILI (P = 0.199). Moreover, if we were to include the 25 patients who had been excluded because the patients did not meet the time-to-onset criteria of the RUCAM - herbs (n = 21) and medication (n = 4) from the original cases - the inconsistency from the interim data would be reduced. With regard to the low level of agreement with the previous literature (3), we found that previous literature has referenced retrospectively analyzed studies with a small number of cases (68 cases/84 months and
Ki Tae Suk, MD1 doi:10.1038/ajg.2014.240
Table 1. Number of cases according to causative agents
To the Editor: We appreciate Bae et al. (1) for their interest in our article, and their comments regarding the accuracy and precision of the study’s estimates. Bae et al. make a very good point and provide constructive comments. They calculated the number of drug-induced liver injury (DILI) cases per 100 beds at each hospital from interim data, which included 304 cases (81.9% of the total cases), to enable a direct comparison among hospitals. We did not use the number of DILI cases per 100 beds at each hospital for the direct comparison among hospitals, because there is great variation in the bed-utilization rate (70–99%) based on certain characteristics of the hospital, including its size ( < 1,000 beds vs. > 1,000 beds or < 700 beds vs. > 700 beds), location (rural vs. urban), and ownership structure (national hospital vs. private hospital), as well as the season of the year. We had performed a statistical analysis by using the proportion The American Journal of GASTROENTEROLOGY
Medications (n=101)
Number of cases Others (n=270)
Total (n=371)
P value 0.199
Hospital 1
6 (40)
9 (60)
15
Hospital 2
6 (26)
17 (74)
23
Hospital 3
5 (18)
23 (82)
28
Hospital 4
0 (0)
3 (100)
3
Hospital 5
11 (44)
14 (56)
25
Hospital 6
4 (33)
8 (67)
12
Hospital 7
5 (42)
7 (58)
12
Hospital 8
5 (36)
9 (64)
14
Hospital 9
2 (13)
13 (87)
15
Hospital 10
8 (30)
19 (70)
27
Hospital 11
8 (21)
31 (79)
39
Hospital 12
2 (12)
15 (88)
17
Hospital 13
7 (30)
16 (70)
23
Hospital 14
4 (17)
19 (83)
23
Hospital 15
7 (19)
30 (81)
37
Hospital 16
4 (24)
13 (76)
17
Hospital 17
17 (42)
24 (58)
41
VOLUME 109 | SEPTEMBER 2014 www.amjgastro.com
Letters to the Editor
48 cases/72 months), and enrollment criteria for the studies differed from those of our prospective study (4,5). These references enrolled patients with all kinds of DILI. However, our study enrolled hospitalized adults with suspected DILI as defined by alanine aminotransferase > 3× upper normal limit or total bilirubin > 2× upper normal limit. We believe that these retrospective references with a small number of cases have not generated strong evidence or a reliable reference range. With regard to Table 4, we classified causative agents by forms. In the causative agents, we described all kinds of suspected causal agents. As the first prospective multicenter study in Asia on this topic, we believe that our study produced new knowledge about DILI. CONFLICT OF INTEREST The author declares no conflict of interest. REFERENCES 1. Bae S, Jang I, Ham C-h. Concerns on the precision of the estimation and the quality management of the data. Am J Gastroenterol 2014;109:1495–6 (this issue). 2. Suk KT, Kim DJ, Kim CH et al. A prospective nationwide study of drug-induced liver injury in Korea. Am J Gastroenterol 2012;107:1380–7. 3. Paik H, Jang I, Lee S. Hepatotoxic events associated with herbal medicinal products, folk remedies and food supplements in Korea. J Korean Oriental Med 2005;26:152–65. 4. Kim JB, Shon JH, Lee HL. et al. Clinical characteristics of acute toxic liver injury. Clin Mol Hepatol 2004;10:125–34. 5. Seo JC, Jeon WJ, Park SS. et al. Clinical experience 48 acute toxic hepatitis patients. Clin Mol Hepatol 2006;12:74–81. 1
Department of Internal Medicine, Hallym University College of Medicine, Chuncheon, South Korea. Correspondence: Ki Tae Suk, MD, Department of Internal Medicine, Hallym University College of Medicine, Gyo-dong, Chuncheon, South Korea. E-mail: [email protected]
Level of Evidence Supporting the Use of Thiopurine in Crohn’s Disease Tetsuji Fujita, MD1 doi:10.1038/ajg.2014.191
© 2014 by the American College of Gastroenterology
To the Editor: On the basis of a metaanalysis of 10 cohort studies, Chatu et al. (1) argue that their findings support the concept that the use of thiopurine analogs, such as azathioprine (AZA) and 6-mercaptopurine, may impact significantly on disease progression, reducing the need for first intestinal surgery by up to 40%. Although sensitivity analysis was performed to reduce the heterogeneity, there are concerns about the inconsistency of the studies, which is not likely to have been measured by a heterogeneity test. For instance, although stricture (hazard ratio (HR) 12.01, 95% confidence interval (CI) 5.97–24.17) and penetration (HR 10.77, 95% CI 4.87–23.80) of the affected intestine were important predictors of major abdominal surgery (2), 1 of the 10 studies included in the meta-analysis excluded patients with stricturing or penetrating Crohn’s disease (3). Recent cohort studies indicate much wider usage and early use of thiopurine analogs and fewer surgical resection rates, but these findings do not necessarily lead to the causal relationship between thiopurine uses and decreased surgical rates, because treatment guidelines were revised with advances in medical and surgical treatment during the study period. Level I evidence is based on the results of large-scale randomized controlled trials (RCTs) or a meta-analysis of RCTs with a narrow CI, and level II evidence is derived from small RCTs. Although aggregating several level II studies can provide level I evidence, the result of a meta-analysis of level III data from nonrandomized comparative cohort studies and level IV data from nonrandomized historical cohort studies would not shift these data to a higher level (4). In an updated Cochrane review of 13 RCTs including 1,211 patients with the primary outcome of the efficacy of AZA or 6-mercaptopurine for induction of remission in active Crohn’s disease, AZA and 6-mercaptopurine offered no advantage over placebo for induction of remission or clinical improvement of active Crohn’s disease (5). In a recent RCT with the primary outcome of the efficacy of use of AZA within 6 weeks of diagnosis of Crohn’s disease, AZA did not reduce the rate for intestinal surgery at 3 years after treatment (6). Taken together, there is no high-level evidence that overall and
early use of thiopurine in Crohn’s disease reduces the rate for intestinal surgery. CONFLICT OF INTEREST
The author declares no conflict of interest. REFERENCES 1. Chatu S, Subramanian V, Saxena S et al. The role of thiopurines in reducing the need for surgical resection in Crohn’s disease: a systematic review and meta-analysis. Am J Gastroenterol 2014;109:23–34. 2. Peyrin-Biroulet L, Oussalah A, Williet N et al. Impact of azathioprine and tumour necrosis factor antagonists on the need for surgery in newly diagnosed Crohn’s disease. Gut 2011;60:930–6. 3. Picco MF, Zubiaurre I, Adluni M et al. Immunomodulators are associated with a lower risk of first surgery among patients with nonpenetrating non-stricturing Crohn’s disease. Am J Gastroenterol 2009;104:2754–9. 4. Fujita T. Levels of evidence for laparoscopic surgery for colorectal cancer. J Am Coll Surg 2011;212:269–70. 5. Chande N, Tsoulis DJ, MacDonald JK. Azathioprine or 6-mercaptopurine for induction of remission in Crohn’s disease. Cochrane Database Syst Rev 2013;4:CD000545. 6. Cosnes J, Bourrier A, Laharie D et al. Early administration of azathioprine vs conventional management of Crohn’s disease: a randomized controlled trial. Gastroenterology 2013;145:758–65. 1
Department of Surgery, Jikei University of School of Medicine, Tokyo, Japan. Correspondence: Tetsuji Fujita, MD, Department of Surgery, Jikei University School of Medicine, 3-25-8 Nishishinbashi, Minato-ku, Tokyo 105-8461, Japan. E-mail: [email protected]
Response to Fujita Sukhdev Chatu, MBBS, MRCP, MD1, Venkataraman Subramanian, MBBS, MRCP, MD1, Sonia Saxena, MBBS, MSc, MD, FRCGP1 and Richard Pollok, MBBS, FRCP, BSc, PhD, DTM&H1 doi:10.1038/ajg.2014.192
To the Editor: We read with interest the concerns raised by Fujita (1) about our meta-analysis of the long-term impact of thiopurines on surgical resection in Crohn’s disease (2). We have never stated that this paper provides level 1 evidence and made the study limitations clear. However, as there are no randomized controlled trials in this area, the best available evidence is derived from pooling cohort studies. The American Journal of GASTROENTEROLOGY
1497
