Harm Scenario

You are referred a 68 year old retired company director with cancer of the proximal stomach. He has angina on moderate effort, is moderately obese (despite recent weight loss) and has a history of excessive alcohol consumption, ceasing about seven years previously. He asks you what type of surgery is available for his condition. Staging investigations including CT and laparoscopy have not shown any liver or peritoneal metastasis, nor any invasion of adjacent organs. You therefore recommend a total gastrectomy with curative intent. He has heard about the controversy over the risks and benefits of extended lymphadenectomy in the Japanese style, and has asks whether you would recommend the Japanese operation for him. You explain that a survival benefit has not been proven for this surgery, although many surgeons remain enthusiastic. He asks, very reasonably, whether the increased risk attributed to the Japanese operation is also unproven. Together, you formulate the question: Is radical lymphadenectomy more dangerous than conventional resection in patients with gastric cancer?

A Medline search (using the terms ‘gastric cancer’ and ‘lymph node excision’ yields two recent randomised controlled trials looking at this question. You choose the first of these Lancet 1995; 345: 745-8).

Read the article and decide:

  • Are the results of this harm study valid?
  • Are the results of the harm study important?
  • Should these valid, important results change the treatment of your patient?

Completed Harm Worksheet for Evidence-Based General Surgery


Bonenkamp JJ et al. Randomised comparison after D1 and D2 dissection for gastric cancer in 1996 Dutch patients.Lancet 1995; 345: 745-8).

Are the results of this harm study valid?

Were there clearly defined groups of patients, similar in all other important ways other than exposure to the treatment (D2 resection).
Yes. All subjects taking part in a randomised trial, for entry to which they had to be fit to receive either D1 or D2 operation, and potentially curable as judged preoperatively. They could, however, be excluded for inoperability at laparotomy, after randomisation. There is a slight risk that surgeons might exclude more patients on these grounds if they were assigned to an operation THEY believed to be more dangerous. There were slightly more exclusions for inoperability in the D2 group (163) than in the D1 group (142).
Were treatment exposures and clinical outcomes measures in the same way in both groups? Was assessment objective, or blinded as to treatment exposure.
Partly. Postoperative death was an objective measure. Complications and reoperations were recorded in the knowledge of the treatment given. This knowledge may have affected the decision to class an incident as a complication, or to perform a re-exploration.
Was the follow-up of study patients complete and long enough?
Do the results satisfy some “diagnostic tests for causation”?
Is it clear that exposure preceded outcome?
Yes: operation preceded death in all cases!
Is there a dose-response gradient?
No. Patients either got D1 or D2 surgery.
Is there positive evidence from a “dechallenge-rechallenge” study?
Not possible in surgery.
Is the association consistent from study to study?
Yes: similar results in Cuschieri et al, Lancet 1996; 347: 995-9
Does the association make biological sense?
Yes, in that bigger surgery is usually more dangerous surgery.

BUT: The important question of the learning curve for a surgical operation was inadequately addressed by both this and the other large trial. D1 surgery was widely practised in the Netherlands before the study, but the “supervising surgeons” got only 1-3 opportunities to practise the D2 technique with expert supervision before supervising others doing it. There is evidence that minimising the complication rate for D2 resection may take 20-25 operations (Parikh et al, 1996).

AND: There was a very clear association between additional resection of the spleen or distal pancreas and mortality on multivariate analysis. After allowing for this, the association of mortality with extended lymph node dissection disappeared.

Are the valid results of this harm study important?

Adverse Outcome (postoperative death)
Present Absent
Exposed to the treatment?
(D2 resection)
Yes 13% 87%
No 6.5% 92.5%

begin{align} text{Relative Risk} &= 13/6.5 \
&= 2.0 text{ ($p$=0.04)} end{align}

Should these valid, potentially important results of a critical appraisal about a harmful treatment change the treatment of your patient?

Can the study results be extrapolated to my patient?
If you accept them at face value, yes. You may also interpret them as suggesting that extended nodal dissection by a suitably experienced surgeon, avoiding additional organ resection, may carry no increased risk.
What are your patient’s risks of postoperative death?
If we assume our patient is like the average study patient (and we are like the average study surgeon!) then his Absolute risk is
13%-6.5% = 6.5%: 1/6.5% = 15
What are your patient’s preferences, concerns and expectations from this treatment?
Need to be determined.
What alternative treatments are available?
Since the evidence against conventional D2 looks strong enough to be worrying, you could offer him D2 with preservation of the spleen and pancreas (if technically possible) or D1 resection.