Gastroenterology
Volume 131, Issue 3 , Pages 980-981, September 2006

Correction

Indianapolis, Indiana USA

Article Outline

 

Kahi CJ, Jensen DM, Sung JJY, Bleau BL, Jung HK, Eckert G, Imperiale TF. Endoscopic Therapy Versus Medical Therapy for Bleeding Peptic Ulcer With Adherent Clot: A Meta-analysis. Gastroenterology 2005;129:855–862.

In this erratum we correct our original analysis1 and further address Dr. Laine’s correspondence.2, 3

In our article, an analysis of randomized trials of endoscopic therapy vs medical therapy in patients with bleeding from peptic ulcer with adherent clot,1 we assessed data both at the level of the study (ie, a meta-analysis, where the study is the unit of analysis) and at the individual patient level (ie, the patient is the unit of analysis, using data we obtained from the authors of the 4 published trials).4, 5, 6, 7 The reason for obtaining patient-level data was to facilitate an analysis of heterogeneous factors among the trials, such as differences in ulcer location and differences in endoscopic therapy. Furthermore, a patient-level analysis is considered to be the gold standard for meta-analysis of study-level data and is more sensitive for detecting treatment effects.8, 9 As there are different ways to aggregate study-level data, and because we appreciated the potential for clinical heterogeneity, we chose a priori a random-effects model for the study-level analysis, which adjusts for heterogeneity among studies and provides a more conservative estimate of a treatment effect through use of wider confidence intervals. Our analysis confirmed the presence of statistical heterogeneity (greater-than-expected variability among studies), supporting the need to use a random-effects model. However, as pointed out to us by Dr Laine, we mistakenly reported results for a fixed effects model, which is best used in the absence of heterogeneity, rather than the intended random effects model.

In this erratum, we provide random-effects relative risks (RR) and 95% confidence intervals (CIs) for the meta-analysis and clarify Table 4 from our article,1 which was supposed to have contained individual patient-level RRs, as well as random-effects results.

Table 4. Results of Patient-level Analysisa and Meta-analysisb
Trial subgroup (subject n)RebleedingSurgeryMortality
Relative risk (95% CI)
4 fully published trials (n = 146)
Patient-level analysis0.30(0.10–0.77)0.72(0.17–2.06)1.40(0.46–3.87)
Meta-analysis0.45(0.09–2.19)0.73(0.20–2.74)1.28(0.41–3.96)
All 6 trials (n = 240)c
Meta-analysis0.48(0.18–1.30)0.42(0.17–1.04)0.80(0.31–2.03)

a Patient-level relative risks and 95% confidence intervals (CIs) were generated using the raw data from the 4 fully published trials.

b Logistic regression analysis was performed using SAS, version 8.0, with the study as a random effects covariate. The resulting odds ratios and 95% CIs were converted to relative risks and CIs using the method of Zhang and Yu.10 Random effects relative risks and 95% CIs for meta-analysis of the 4 published trials and for all 6 studies were generated using RevMan 4.2 software for meta-analysis.

c Includes 2 trials published only in abstract form.11, 12

Among the 4 published trials, Table 4 shows that there is agreement between the patient-level analysis and meta-analysis for the outcomes of surgery and mortality. However, the results for rebleeding may be considered discrepant. Although the RRs for rebleeding are consistent, the 95% CI for the individual patient-level analysis excludes 1 (suggesting a protective effect of endoscopic therapy), while that for the meta-analysis includes 1 (suggesting either no effect of endoscopic therapy as compared with medical therapy or a lack of power to detect a clinically important effect).

Since the patient-level analysis has greater statistical power than the meta-analysis and is considered the gold standard, it represents the more accurate and robust numerical results. Discrepant (though not inconsistent) findings between meta-analysis and individual patient-level analysis are not uncommon or unexpected.13, 14 While these new analyses appropriately correct our numerical errors, they have no effect on our interpretation and discussion of the results and their limitations, as described in the original manuscript, both with respect to validity and generalizability.

Back to Article Outline

References 

  1. Kahi CJ, Jensen DM, Sung JJY, Bleau BL, Jung HK, Eckert G, et al. Endoscopic therapy versus medical therapy for bleeding peptic ulcer with adherent clot: a meta-analysis . Gastroenterology . 2005;129:855–862
  2. Laine L . Systemic review of endoscopic therapy for ulcers with clots: can a meta-analysis be misleading? . Gastroenterology . 2005;129:2127
  3. Kahi CJ , Jensen DM , Imperiale TF . Systemic review of endoscopic therapy for ulcers with clots: can a meta-analysis be misleading? . (response) Gastroenterology . 2005;129:2127–2128
  4. Bleau BL , Gostout CJ , Sherman KE , et al.   Recurrent bleeding from peptic ulcer associated with adherent clot: a randomized study comparing endoscopic treatment with medical therapy . Gastrointest Endosc . 2002;56:1–6
  5. Jensen DM , Kovacs TO , Jutabha R , et al.   Randomized trial of medical or endoscopic therapy to prevent recurrent ulcer hemorrhage in patients with adherent clots . Gastroenterology . 2002;123:407–413
  6. Jung HK , Son HY , Jung SA , et al.   Comparison of oral omeprazole and endoscopic ethanol injection therapy for prevention of recurrent bleeding from peptic ulcers with nonbleeding visible vessels or fresh adherent clots . Am J Gastroenterol . 2002;97:1736–1740
  7. Sung JJ , Chan FK , Lau JY , et al.   The effect of endoscopic therapy in patients receiving omeprazole for bleeding ulcers with nonbleeding visible vessels or adherent clots: a randomized comparison . Ann Intern Med . 2003;139:237–243
  8. Lambert PC , Sutton AJ , Abrams KR , Jones DR . A comparison of summary patient-level covariates in meta-regression with individual patient data meta-analysis . J Clin Epidemiol . 2002;53:86–94
  9. Teramuki S , Matsuyama Y , Mizuno S , Sakamoto J . Individual patient-level and study-level meta-analysis for investigating modifiers of treatment effect . Jpn J Clin Oncol . 2004;34:717–721
  10. Zhang J , Yu KF . What’s the relative risk? A method of correcting the odds ratio in cohort studies of common outcomes . JAMA . 1998;280:1690–1691
  11. Jensen DM, Kovacs TO, Jutabha R, Randall GM, Cheng S, Jensen ME, et al. Final results and cost assessment of endoscopic vs. medical therapies for prevention of recurrent ulcer hemorrhage from adherent clots in a randomized controlled trial . (abstr) Gastrointest Endosc . 1995;41:279
  12. Gonzalez-Huix F , Figa M , Hombrados M , Gavalda L , Acero D . Randomized controlled trial of endosocpic sclerotherapy compared with medical treatment for the prevention of recurrent ulcer hemorrhage in patients with non-bleeding adherent clot . (abstr) Endoscopy . 2000;32(Suppl 1):E29
  13. Stewart LA , Parmar MK . Meta-analysis of the literature or of individual patient data: is there a difference? . Lancet . 1993;341:418–422
  14. Steinberg KK, Smith SJ, Stroup DF, Olkin I, Lee NC, Williamson GD, et al. Comparison of effect estimates from a meta-analysis of summary data from published studies and from a meta-analysis using individual patient data for ovarian cancer studies . Am J Epidemiol . 1997;145:17–25

PII: S0016-5085(06)01669-6

doi:10.1053/j.gastro.2006.07.026

Refers to article:

  • Endoscopic Therapy Versus Medical Therapy for Bleeding Peptic Ulcer With Adherent Clot: A Meta-analysis

    Charles J. Kahi, Dennis M. Jensen, Joseph J.Y. Sung, Brian L. Bleau, Hye Kyung Jung, George Eckert, Thomas F. Imperiale
    Gastroenterology September 2005 (Vol. 129, Issue 3, Pages 855-862)

Gastroenterology
Volume 131, Issue 3 , Pages 980-981, September 2006

Article Tools