Redundant Data in the Meta-analysis on Helicobacter pylori Eradication

TO THE EDITOR: We read with interest the recent meta-analysis by Fuccio and colleagues (1). There is an ongoing Cochrane review on this subject; its protocol is published (2), and its results have been reported in abstract form (3). Fuccio and colleagues report that eradication therapy for Helicobacter pylori reduces the subsequent incidence of gastric cancer. However, we have serious doubts about the accuracy of their analysis. In the Methods section of their article, Fuccio and colleagues state that when they found multiple articles for a single study, they used the latest publication from each eligible study. However, the meta-analysis erroneously incorporated data from the same randomized, controlled trial (RCT) twice (4, 5). Tables 1 and 2 (1) showed 2 studies that were conducted in the Shandong province of China, commenced in 1996, had very similar numbers of participants, and used identical eradication regimens of the same duration. We contacted the original investigators of these studies (2, 3) directly when collecting data for our review. Those investigators confirmed that the 2 publications were in fact 5-year (4) and 10-year (5) follow-up studies from the same RCT. The erroneous inclusion of repeated data led Fuccio and colleagues to report a statistically significant effect of H. pylori eradication therapy in reducing incidence of gastric cancer, when this is not the case. If only the 10-year follow-up data are included (5), the pooled relative risk for subsequent gastric cancer is 0.65 (95% CI, 0.42 to 1.01). Also, the pertinent trial publications reported fewer cases of gastric cancer after 10-year follow-up (5) than at 5-year follow-up (4), raising concerns about the accuracy of data collection or of reporting for the trial. Taken together, this information convinces us that the effect of eradication therapy on gastric cancer is not as clear-cut as Fuccio and colleagues suggest. We delayed publishing our full Cochrane review because of these concerns about the duplicate Chinese studies. Our knowledge and experience regarding errors in meta-analyses and the fact that Annals published the meta-analysis with erroneous data highlight several important issues for authors of meta-analyses, for journal editors, and for peer reviewers. First, transparent reporting of follow-up studies conducted at various time points from the same RCT is needed to ensure correct identification of trials and to avoid miscounting of data by authors. Second, when studies are reported only in abstract form, or if the accuracy of the data is in doubt, authors should directly contact original investigators for clarifications. Finally, peer reviewers (and journal editors) may have difficulty confirming the results of systematic reviews and meta-analyses. In some cases, independent verification of the meta-analysis may be needed to ensure the results are truly accurate.

IN RESPONSE: Drs. Ford and Moayyedi write that we counted the same trial twice in our recently published meta-analysis. Reports of the "2" pertinent trials (1,2) were not written by the same authors. The 2 reports showed that the work had been carried out in the Shandong province of China in the 1990s and used the most common worldwide eradication treatment at that time (1 week of omeprazole, clarithromycin, and amoxicillin). The 2 reports, as well as a third related report, showed confusing discrepancies regarding recruitment methods for participants and numbers of patients randomly assigned to eradication treatment and control groups (1)(2)(3). These reports also gave discrepant information on eradication rate of H. pylori in the treated group, baseline histologic characteristics of included patients, histologic classification system used, age inclusion range and mean age of included patients, and number of gastric cancer cases in each group. Therefore, we thought the 2 studies were different, although we did try to contact authors for clarification and confirmation. Our attempt to contact Leung and colleagues (1) by e-mail was unsuccessful. Zhou responded to our request for additional information by sending us 3 publications (2-4), but made no reference to the study by Leung and colleagues.
We thank Drs. Ford and Moayyedi for pointing out that Zhou's report (2) may represent 10-year follow-up data from the same trial for which Leung and colleagues reported 5-year follow-up data (1). However, if this is so, how can there be fewer cases of gastric cancer in the treatment group at 10 years (n ϭ 2) than at 5 years (n ϭ 4)? It calls the accuracy of these data into serious doubt, and one must ask which of the 2 groups of data, if either, should be considered. We strongly agree with Drs. Ford and Moayyedi about the need for transparent reporting for follow-up studies conducted at different stages of the same RCT.
Finally, we avoided emphasizing the statistical significance of the results of our meta-analysis because the few available studies did not allow firm conclusions. A relative risk of 0.65 (CI, 0.43 to 0.98; P ϭ 0.04) is very close to that obtained after excluding the 5-year follow-up data reported by Leung and colleagues (1): relative risk, 0.65 (CI, 0.42 to 1.01; P ϭ 0.05). Clinical and statistical significance do not always correspond. Therefore, even after excluding the study by Leung and colleagues, our conclusions remain the same: Helicobacter pylori eradication treatment seems to modestly reduce the risk for gastric cancer. However, from a clinical point of view, even a small reduction in risk and incidence achieved with H. pylori

Annals of Internal Medicine
Letters eradication treatment will probably give a huge advantage in terms of social health, especially in high-risk areas.

Granulysin as a Marker for Early Diagnosis of the Stevens-Johnson Syndrome
Background: The Stevens-Johnson syndrome (SJS) and toxic epidermal necrolysis (TEN) are life-threatening adverse drug reactions characterized by massive epidermal necrosis. In the early stage, clinical presentations of SJS/TEN are very similar to those of ordinary drug-induced skin reactions (ODSRs); therefore, SJS/TEN is difficult to diagnose and the start of treatment is often delayed, resulting in high mortality rates. Other investigators (1) reported that granulysin is highly expressed in blisters of SJS/TEN and causes disseminated keratinocyte death. Because SJS/TEN progresses and spreads rapidly, the granulysin level should be increased in the serum of patients with active SJS/TEN if it is a key mediator of these diseases.
Objective: To determine whether serum granulysin levels are higher in patients with SJS/TEN than in healthy control participants or those with ODSRs.
Methods: We measured granulysin in the sera of 31 healthy control participants, 24 patients with ODSR, 13 patients with SJS, and 7 patients with TEN by using enzyme-linked immunosorbent assay (2). Disease onset in patients with SJS/TEN was defined as the day (day 1) on which the mucocutaneous or ocular lesion first eroded or ulcerated (3), and we collected sera from these patients from 4 days before to 10 days after ulceration. We used the Tukey-Kramer test to conduct multiple comparisons between groups.
Results: None of the 31 healthy control participants had a granulysin level greater than the upper limit of normal, which was 10 ng/mL (0% elevated; mean, 1.6 ng/mL [SD, 0.6]), and among 24 patients with ODSRs, only 1 patient had an elevated granulysin level (4.2% elevated; mean, 3.5 ng/mL [SD, 3.4]) (Figure). We obtained samples from 5 patients with SJS/TEN on day Ϫ4 to day Ϫ2, and we detected the highest granulysin concentrations (elevated in 80% of patients); mean, 24.8 ng/mL [SD, 21.2]). Granulysin levels were lower in the 14 samples collected on day Ϫ1 to day 2 (28.6% elevated; mean, 13.7 ng/mL [SD, 16.0]), and were even lower in the 10 samples collected from day 3 to day 5 (10.0% elevated; mean, 4.2 ng/mL [SD, 3.0]) and in the 13 samples collected from day 6 to day 10 (7.7% elevated; mean, 4.5 ng/mL [SD, 4.5]). When we compared granulysin levels from day Ϫ4 to day Ϫ2 among patients with SJS/TEN, patients with ODSRs, and healthy control participants, the differences were statistically significant (P Ͻ 0.010).
Discussion: Granulysin is cytotoxic for tumor cells, transplant cells, bacteria, fungi, and parasites, in which it damages negatively charged cell membranes because of its positive charge (4). It plays an important role in the host defense against pathogens, and it induces apoptosis of target cells by using a mechanism involving caspases and other pathways (4). Its potency makes it a credible mediator of skin damage in patients with SJS/TEN. Adding to this credibility is a report (1) that granulysin is the most highly expressed cytotoxic molecule in the blisters of patients with SJS/TEN. We show that serum granulysin levels in 4 of 5 patients with SJS/TEN were elevated before skin detachment or mucosal lesions develop. Soluble Fas ligand (sFasL) shares some properties with granulysin: It contributes to keratinocyte death in SJS/TEN (3,5), and levels are elevated in the sera of patients with SJS/TEN (3). Serum granulysin levels, however, are approximately 100 times higher than those of sFasL on day Ϫ4 to day Ϫ2 (23.

Localized Amyloidosis at the Site of Enfuvirtide Injection
Background: Enfuvirtide is the first of a new class of antiretroviral agents that block fusion of the viral particle with the host target cell. Its safety and antiviral activity have been demonstrated (1,2). In clinical trials, injection site reactions occurred in 80% to 100% of patients (3). The most common signs and symptoms reported were induration in 94%, erythema in 91%, and subcutaneous nodules and cysts in 70% (4).
Objective: To describe a case of amyloidosis at the injection site of enfuvirtide.
Case Report: The patient was a man aged 47 years who had a history of sexual intercourse with men and extensive treatment for HIV with a triple-class viral resistance profile. He also had longstanding leg pain thought to be secondary to HIV neuropathy and no history of intravenous drug use. There was no history of opportunistic or chronic infections.
Because of a persistently elevated viral load, enfuvirtide by subcutaneous injection was added to his highly active antiretroviral treatment regimen for 41 months; enfuvirtide therapy was then stopped in February 2007 because of intolerable injection site reactions. While he was receiving enfuvirtide, his viral loads were completely suppressed. Eighteen months after enfuvirtide therapy was stopped, large, tender, indurated reactions with fragile epithelial sur-faces persisted at all injection sites (Figure, top). These reactions bled extensively into the subcutaneous tissue with minor trauma (Figure,  bottom). A lesion on the triceps was excised surgically, and the wound healed without complications. Pathologic examination showed extensive deposits of proteinaceous material with intense Congo red staining that was consistent with amyloid. A lesion on the opposite arm was resected and showed similar findings. The patient had a normal leukocyte count and normal hemoglobin, blood urea nitrogen, and creatinine levels and had no evidence of plasma cell dyscrasia and no history of organ dysfunction to suggest systemic amyloidosis.
Discussion: In 7 patients receiving enfuvirtide, biopsy of injection site reactions revealed an inflammatory response consistent with a localized hypersensitivity reaction (5), and other studies (3) have reported similar findings. Other reports (6) have described 3 histologic patterns: an acute urticaria-or vasculitis-like pattern with inflammation of the fat tissue, a subacute pattern with an initial dermal sclerosis, and a long-term scleroderma-like pattern.
In our patient, surgical excision of enfuvirtide injection site reactions revealed subcutaneous nodular amyloidosis. Localized  subcutaneous nodular amyloidosis is rare (7,8). The lesions can present as waxy nodules with or without overlying atrophic epidermis, and they may ulcerate with minimal trauma, causing cutaneous hemorrhage (7)(8)(9). Some authors (7) have reported that nodular cutaneous amyloidosis may occur in relation to cutaneous plasmocytoma. Amyloid formation at the site of drug injection has been described in 5 previous patients. Of these, 4 were receiving either porcine or human insulin for glucose control (10 -13), and 1 was given injections of an unknown medication during the Korean War (14). To our knowledge, this is the first case of localized amyloidosis associated with the use of enfuvirtide. We believe that localized amyloidosis should be considered in patients with severe, persistent injection site reactions and suggest that subcutaneous hemorrhage may help make the diagnosis.
A recent article (1) that pooled data from 6 trials reported that, compared with no treatment, Helicobacter pylori eradication treatment reduced the relative risk (RR) for gastric cancer (0.65 [95% CI, 0.43 to 0.98]). The article described and counted data from "2" trials that the editors believe were actually 5-and 10-year follow-up data from the same trial (2,3). Furthermore, the reported data for the eradication treatment group for the trial showed fewer cases of gastric cancer at 10-year follow-up (n ϭ 2) than at 5-year follow-up (n ϭ 4). Reported numbers for the 10-year follow-up were extracted from an abstract presentation (3). If the reported data from the 5-or 10-year follow-up are excluded, the pooled RR is 0.65 (CI, 0.42 to 1.01) or 0.70 (CI, 0.46 to 1.08), respectively. If reported data from both the 5-year and 10-year follow-up are excluded, the pooled RR is 0.71 (CI, 0.45 to 1.23).