Avoidance of ingestion of anti-inflammatory drugs in dyspepsia is a confounding variable in apparent
Editor - Langman et al describe a method using coded data from the
general practice research database to support their hypothesis that anti-
inflammatory drugs may protect against oesophageal and gastric cancer. 1
The study reported the association between a reduction of coded cases of
cancer of the oesophagus, stomach, colon and rectum in a sub-population
who had received at least seven prescriptions in the 13-36 months before
diagnosis. This conclusion seems to be ambitious as aetiologically it
seems unreasonable to anticipate that the use of a drug in the 36 months
prior to diagnosis will halt a neoplastic process that may have began many
months or years before. Gastrointestinal cancer (particularly of the
oesophagus and stomach) is frequently associated with abdominal pain and
dyspepsia. Patients (and their attending clinicians) will avoid the use of
anti-inflammatory drugs in the presence of such dyspepsia. Thus surely the
reported association between the use of anti-inflammatory drugs near the
time of diagnosis is more elegantly explained by the confounding avoidance
of these drugs in dyspepsia associated with malignancy. The authors also
describe a possible dose effect and once again this is equally well
explained by the greater avoidance of these drugs in patients with
increased dyspepsia rather than invoking a hypothetical mechanism of gut
epithelial protection.
One of the great advantages of collecting clinical information in a coded
format is that new associations may be discovered using a variety of
techniques including knowledge discovery in databases (KDD). 2 However if
such exploitation of repositories of data is to gain recognition and
acceptance in medicine vigilance is needed in interpreting these
associations and a full consideration of possible confounding factors is
essential before the proposal of new theories.
Philip JB Brown general practitioner
Humbleyard Practice, Norfolk
1. Langman MJS, Cheng KK, Gilman EA, Lancahire RJ. Effect of anti-
inflammatory drugs on overall risk of common cancer: case-controlled study
in general practice research database BMJ 2000;320:1642-6. (17 June)
2. Brown PJB, Rayward-Smith V. Theoretical potential of semantic
based knowledge discovery from terminology populated clinical databases.
(Prepared for submission)
Rapid Response:
Avoidance of ingestion of anti-inflammatory drugs in dyspepsia is a confounding variable in apparent
Editor - Langman et al describe a method using coded data from the
general practice research database to support their hypothesis that anti-
inflammatory drugs may protect against oesophageal and gastric cancer. 1
The study reported the association between a reduction of coded cases of
cancer of the oesophagus, stomach, colon and rectum in a sub-population
who had received at least seven prescriptions in the 13-36 months before
diagnosis. This conclusion seems to be ambitious as aetiologically it
seems unreasonable to anticipate that the use of a drug in the 36 months
prior to diagnosis will halt a neoplastic process that may have began many
months or years before. Gastrointestinal cancer (particularly of the
oesophagus and stomach) is frequently associated with abdominal pain and
dyspepsia. Patients (and their attending clinicians) will avoid the use of
anti-inflammatory drugs in the presence of such dyspepsia. Thus surely the
reported association between the use of anti-inflammatory drugs near the
time of diagnosis is more elegantly explained by the confounding avoidance
of these drugs in dyspepsia associated with malignancy. The authors also
describe a possible dose effect and once again this is equally well
explained by the greater avoidance of these drugs in patients with
increased dyspepsia rather than invoking a hypothetical mechanism of gut
epithelial protection.
One of the great advantages of collecting clinical information in a coded
format is that new associations may be discovered using a variety of
techniques including knowledge discovery in databases (KDD). 2 However if
such exploitation of repositories of data is to gain recognition and
acceptance in medicine vigilance is needed in interpreting these
associations and a full consideration of possible confounding factors is
essential before the proposal of new theories.
Philip JB Brown general practitioner
Humbleyard Practice, Norfolk
1. Langman MJS, Cheng KK, Gilman EA, Lancahire RJ. Effect of anti-
inflammatory drugs on overall risk of common cancer: case-controlled study
in general practice research database BMJ 2000;320:1642-6. (17 June)
2. Brown PJB, Rayward-Smith V. Theoretical potential of semantic
based knowledge discovery from terminology populated clinical databases.
(Prepared for submission)
Competing interests: No competing interests