Upper GI bleed: Diagnostic endoscopy had no effect on rebleeding or mortality

Clinical bottom line (level 1b-)

  1. Patients who underwent routine diagnostic endoscopy were not clearly less likely to rebleed or die than those who do not.
Peterson et al: The New England Journal of Medicine 1981; 304: 925-929
Expires October 2002

The study

Unblinded ?concealed randomised trial without intention-to-treat
Setting: Veteran Affairs hospital, USA

206 patients (aged 55, ?% male) with untreated upper gastrointestinal bleeding (a recent history of haematemesis or melaena documented by a bloody nasogastric aspirate or by a stool sample positive for gross or occult blood)

Excluded if
  • condition not stabilised within 6 hours of treatment (presence of fresh blood on gastric lavage, or clinical evidence of hypovolaemia or haematocrit <25%)
  • contraindication to endoscopy: uncooperative, presence of myocardial ischaemia, pulmonary or cardiac failure, or evidence of perforated viscus.
  • abdominal vascular grafts


Control Group: (n = 100, 100 analysed): Routine endoscopy within 4 hours of stabilisation
Experimental Group: (n = 106, 106 analysed): No routine endoscopy.
All patients were treated with antacids hourly for the first 24 hours, then qds for 6 weeks. All patients had an upper GI series within 2 to 4 days of admission.
92% followed for 12 months
Outcome notes:
  • recurrent bleeding : during hospitalisation
  • complications : pneumonia, myocardial infarction

The evidence

Outcome Time to outcome CEREERRRR
(95% CI)		ARR
(95% CI)		NNT
(95% CI)
recurrent bleeding 4 days 33
(33%)		 32
(30.19%)		 9%
(-37% to 39%)		 2.81%
(-9.89% to 15.51%)		 36
(NNT = 10 to infinity;
NNH = 6 to infinity)
death 12 months 11
(11%)		 8
(7.55%)		 31%
(-64% to 71%)		 3.45%
(-4.48% to 11.38%)		 29
(NNT = 9 to infinity;
NNH = 22 to infinity)
complications 4 days 6
(6%)		 4
(3.77%)		 37%
(-116% to 82%)		 2.23%
(-3.67% to 38.13%)		 45
(NNT = 12 to infinity;
NNH = 27 to infinity)
GI bleed after discharge from hospital 12 months 6
(6%)		 7
(6.6%)		 -10%
(-216% to 62%)		 -0.60%
(-7.24% to 6.03%)		 -166
(NNT = 17 to infinity;
NNH = 14 to infinity)

Comments

  1. Patients were randomised in blocks of ten.
  2. Note that the study excluded the sickest patients - i.e. the ones who might have got most benefit
  3. The study is too small to exclude any benefit from routine endoscopy. It was also performed before interventional endoscopy was common-place.

    Citation

    1. Peterson WL, Barnett CC, Smith HJ, et al: Routine endoscopy in upper-gastrointestinal-tract bleeding. The New England Journal of Medicine 1981; 304: 925-929
    Contributor: Chris Ball and Musab Hayatli, October 1999
    Reviewer:

    Clinical Question.
    Patient upper GI bleed
    Intervention or Exposure endoscopy
    Outcome death, rebleeding