Risk of adverse gastrointestinal outcomes in patients taking cyclo-oxygenase-2 inhibitors or conventional non-steroidal anti-inflammatory drugs: population based nested case-control analysis
BMJ 2005; 331 doi: https://doi.org/10.1136/bmj.331.7528.1310 (Published 01 December 2005) Cite this as: BMJ 2005;331:1310All rapid responses
Rapid responses are electronic comments to the editor. They enable our users to debate issues raised in articles published on bmj.com. A rapid response is first posted online. If you need the URL (web address) of an individual response, simply click on the response headline and copy the URL from the browser window. A proportion of responses will, after editing, be published online and in the print journal as letters, which are indexed in PubMed. Rapid responses are not indexed in PubMed and they are not journal articles. The BMJ reserves the right to remove responses which are being wilfully misrepresented as published articles or when it is brought to our attention that a response spreads misinformation.
From March 2022, the word limit for rapid responses will be 600 words not including references and author details. We will no longer post responses that exceed this limit.
The word limit for letters selected from posted responses remains 300 words.
The paper by Hippisley-Cox and colleagues has, like other apparently
negative papers regarding COX-2-selective drugs, been widely reported.
However, it seems to us that as remarked upon in other rapid responses
there are confounding factors. Firstly, as mentioned by Drs. Lewis and
Kay, there is the key issue of differences in the patients receiving the
various drugs. The NICE guidelines in place during much of this period
recommended COX-2-selective drugs only to be used for patients at
particular gastrointestinal risk, such as older people or those with a
history of GI disturbances, and this has been reported in the UK as
consistent with channelling of higher risk patients to COX-2-selective
drugs (ref 1). So although the study matched patients to controls we are
not told if, for instance, the celecoxib or rofecoxib patient groups were
comparable to the ibuprofen or naproxen groups (NB we can only see that
the pooled drug taking group is comparable to the pooled non-drug taking
group). Without this information regarding the various drug taking
patient groups one might put forward the alternate conclusion that this
study demonstrates that celecoxib and rofecoxib are as safe in at risk
populations as other NSAIDs are in general populations. The question we
should ask, of course, is how would the NSAIDs have fared in the higher
risk groups?
A second point as referred to by Prof. Sanchez-Delgado is the dose of
each drug taken. For example, in a report published in this particular
journal 10 years or so ago, the relative gastrointestinal toxicities of a
range of NSAIDs showed that in general terms ibuprofen was associated with
lower gastrointestinal risk than other NSAIDs. However, at higher doses
ibuprofen was as risky as other NSAIDs (ref 2). So ibuprofen's apparent
safety was associated with its particular use at low dosage. Current
prescription guidelines seem also to encourage use of low dose ibuprofen
as an early measure with alternative drugs used if ibuprofen proves
ineffective. This would suggest that much of the ibuprofen data is from
lower dose. This consideration also touches upon the comments of Dr.
Underhill. While it is true that in the CLASS study celecoxib did not
show GI benefit except in secondary analysis one should also be aware that
this particular study (as did VIGOR) employed the COX-2-selective drug at
double the usual maximum dose; i.e. in CLASS celecoxib was given at 400 mg
twice daily and in VIGOR rofecoxib was given at 50 mg once daily. But the
comparator NSAIDs were given at standard doses. Seen in that light the
COX-2-selective drugs didn't do as badly as one is often encouraged to
believe.
So, in simple terms, the COX-2-selective drugs might not actually
look so bad in this analyis if one thinks about the patients that probably
received them and the doses of NSAIDs that were probably prescribed.
Finally, why is there continual pressure for COX-2-selective drugs to
outperform traditional NSAIDs in unbalanced competitions? How would the
standard NSAIDs have fared at double dose in CLASS or VIGOR? Naproxen at
2000mg a day for one year? Ibuprofen at 4800mg? How would these standard
NSAID look in the "real world" if channelled to high risk patients?
Yours sincerely
Timothy D. Warner and Jane A. Mitchell
1 - MacDonald TM, Pettitt D, Lee FH, Schwartz JS. Channelling of
patients taking NSAIDs or cyclooxygenase-2-specific inhibitors and its
effect on interpretation of outcomes. Rheumatology (Oxford). 2003 Nov;42
Suppl 3:iii3-10.
2 - Henry D, Lim LL, Garcia Rodriguez LA, Perez Gutthann S, Carson
JL, Griffin M, Savage R, Logan R, Moride Y, Hawkey C, Hill S, Fries JT.
Variability in risk of gastrointestinal complications with individual non-
steroidal anti-inflammatory drugs: results of a collaborative meta-
analysis. BMJ. 1996 Jun 22;312(7046):1563-6.
Competing interests:
Have received research, lecturing and/or consulting fees from AAiPharma, Boehringer Ingelheim, Merck, Novopharm, Pfizer, Schering-Plough and Shire Pharmaceuticals,
Competing interests: No competing interests
EDITOR- We read with great interest the study by Hippisley-Cox et al.
1 about the gastro-intestinal risks of cyclo-oxygenase-2 (COX 2)
inhibitors and conventional non steroidal anti-inflammatory drugs. Our
goal is not to discuss the data but only to underline that the conclusion
of a lack of enhanced safety of COX 2 inhibitors was already widely
published in the literature. For example, the Laporte’s study 2, a
multicentre case-control study, found an odds ratio of 7.2 for rofecoxib.
In a case/non-case study performed in the French PharmacoVigilance
database between May 2000 and December 2002 3, we also found an increased
risk of exposure to coxibs available in France (rofecoxib; celecoxib) in
oeso-gastro-intestinal adverse drug reactions (ADRs). In fact,
epidemiological results coming from PharmacoVigilance databases are often
criticized due to their limitations: selection biases, underreporting,
“Weber” effect… This is only partly true. For example, for coxibs, it is
not possible to translate into the “real life”, the odds ratio values
found in a case/non case study, since a notoriety bias probably led to
over reporting of such oeso-gastro-intestinal ADRs.
However, comparison of
the case/non case study and the present one clearly shows that utilization
of automated ADR database (as, for example, a PharmacoVigilance database)
appears to be very useful in order to generate signals in
pharmacovigilance. Such a method (which is relatively rapid and
inexpensive) should be included in regular surveillance of drugs’ risk,
especially during the first years of drugs’ life, i.e. before than their
effects could be studied using more conventional pharmacoepidemiological
methods (like cohort or case-control studies). Thus, from a methodological
point of view, we believe that automatic analyses of PharmacoVigilance
databases should be added to the methods used for detection of ADRs. True
quantification of the risk will be performed later. This conclusion seems
important at the time where, following rofecoxib withdrawal, methods of
drugs’ surveillance are widely discussed and criticized.
The Hippisley-Cox’s study 1 and the two others cited above 2,3 also
demonstrate that the oeso-gastro-intestinal risk of coxibs is clearly
higher than that previously suggested by the sole clinical trials.
Finally, they also recall that, after randomised controlled double-blind
studies, pharmacoepidemiological studies are necessary to definitively
quantify drug’s benefit risk ratio in the real word and to verify (or not)
the conclusions of clinical trials.
Jean-Louis Montastruc, MD, PhD, pharmacoepidemiologist, Maryse
Lapeyre-Mestre, MD, PhD, pharmacoepidemiologist, Stephanie Lugardon, MD,
pharmacoépidémiologise, Agnes Sommet, MD, pharmacoépidémiologist.
Laboratoire de Pharmacologie Médicale et Clinique, Unité de
Pharmacoépidémiologie, EA 3696, IFR 126, Université Paul Sabatier, Faculté
de Médecine, 37 allées Jules-Guesde, 31000 Toulouse, France.
Address all correspondence to J.L. Montastruc montastruc@cict.fr
No conflicts of interest
References
1. Hippisley-Cox J, Coupland C, Logan R. Risk of adverse gastrointestinal
outcomes in patients taking cyclo-oxygenase-2 inhibitors or conventional
non-steroidal anti-inflammatory drugs: population based nested case-
control analysis. BMJ, 2005; 331: 1310-6
2. Laporte JR, Ibanez L, Vidal X, Vendrell L, Leone R. Upper
gastrointestinal bleeding associated with the use of NSAIDs: newer versus
older agents. Drug Saf 2004; 27: 411-20
3. Lugardon S, Lapeyre-Mestre M, Montastruc J.L. Upper gastrointestinal
adverse drug reactions and cyclo-oxygenase-2 inhibitors (celecoxib and
rofecoxib): a case/non-case study from the French Pharmacovigilance
database. Eur J Clin Pharmacol 2004; 60: 676-7
Competing interests:
None declared
Competing interests: No competing interests
I was fascinated by Dr Joe Feczko's comments about the apparently
misleading concusions drawn by the authors of the Hippisley-Cox paper.
It is important to note that this was an observational study, which
we all know, occupies a lower place on the hierarchy of evidence than
large, prospective randomised controlled trials measuring outcomes that
are important to patients (Patient-oriented evidence that matters or
POEMs)
Perhaps Dr Feczko would care to comment on the only POEM evidence
that exists for celecoxib, namely from the CLASS study?(1)
CLASS showed no statisically significant difference between celecoxib
and the comparators (diclofenac and ibuprofen) in terms of the primary
outcome of the study - GI ulcer complications.(2) Only when a post-hoc sub
-group analysis of these data was performed in those not taking aspirin
was a statistically significant benefit seen (with a P value of 0.04).
This was one of over 34 post-hoc analyses performed on this study (so, by
chance, we would expect at least one of these to show a difference with a
P value slightly less than 0.05.(3) Others have recently highlighted the
folly of post-hoc subgroup analyses where the primary endpoint does not
show a significant difference.(4)
This trial has also been the subject of an in-depth critique by Juni,
which readers are encouraged to read.(2)
The regulators in the USA and in Europe have concluded that the data
from the CLASS study does not show a meaningful benefit for celecoxib.(5)
It is disappointing that the prescribing of this drug continues to
increase despite the lack of good quality evidence for its usefulness in
providing a benefit to our patients.
Perhaps we should stick to the facts rather than argue about the
wording of the conclusions of hypothesis generating data (like the
Hippisley-Cox paper). Although such data are interesting, they do not
inform our practice in the same way as a NEGATIVE prospective randomised
controlled trial does.
Yours
Jonathan Underhill
1 Silverstein FE, et al. JAMA 2000; 284: 1247-1255
2 Jüni P, et al. BMJ 2002; 324: 1287-1288
http://bmj.bmjjournals.com/cgi/content/full/324/7349/1287
3 Lu HL
(www.fda.gov/ohrms/dockets/ac/01/briefing/3677b1_04_stats.doc)
4 Freemantle N. How well does the evidence on pioglitazone back up
researcher’s claims for a reduction in macrovascular events? BMJ 2005;
331: 836-838 http://bmj.bmjjournals.com/cgi/content/full/331/7520/836
5 FDA Talk Paper, June 7,
2002.http://www.fda.gov/bbs/topics/ANSWERS/2002/ANS01151.html
Competing interests:
None declared
Competing interests: No competing interests
Dear Sir
We believe that the study by Hippisley-Cox into the safety of cyclo-
oxygenase-2 (COX-2) inhibitors and non-steroidal anti-inflammatory drugs
(NSAIDS) is subject to more indication bias than they suggest1. When the
study began five years ago, COX-2 inhibitors were widely regarded as safer
to prescribe in patients at high risk of developing gastro-intestinal (GI)
side effects than NSAIDS2 , although there were doubts over their
cardiovascular safety 3. Working in general practice at the time, we saw
COX-2 inhibitors used in primary care patients who had mild upper GI
symptoms or history, as a ‘safe option’. They were prescribed
particularly in the elderly, in whom opiate analgesics often cause
significant adverse effects. This practice would have led to a higher
risk cohort of patients being prescribed COX-2 inhibitors rather than
NSAIDS, thus reducing the apparent safety of COX-2 s in this study.
Additionally, during the study period, both ulcer healing drugs and
NSAIDS were readily available ‘over the counter’ (OTC). The study relied
on using computerised prescribing records from general practices as the
main source of prescribing data. We feel that this neglects the
confounding influence of OTC drugs which have an important influence on
the prevalence of GI pathology 4 5.
Yours sincerely
Michael Lewis BSc MRCGP MRCOphth
Dr Diana Kay MB.BS, MRCGP, DCH
1. Hippisley-Cox J, Coupland C, Logan R. Risk of adverse
gastrointestinal outcomes in patients taking cyclo-oxygenase-2 inhibitors or
conventional non-steroidal anti-inflammatory drugs: population based nested case-
control analysis. BMJ, Dec 2005; 331: 1310 - 1316; doi:10.1136/bmj.331.7528.1310
2. Efficacy, tolerability, and upper gastrointestinal safety of
celecoxib for treatment of osteoarthritis and rheumatoid arthritis: systematic review of
randomized controlled trials. Deeks JD, Smith LA, Bradley MD. BMJ 2002: 325;619
3. Efficacy and safety of COX-2 inhibitors. Jones R. BMJ 2002:325;
607-608
4. Non-steroidal anti-inflammatory drugs and elderly patients.
Bateman DN, Kennedy JG. BMJ 1995;310:817-818
5. Are H2 receptor antagonists safe over the counter drugs? Andersen
M, Schou JS BMJ 1994;309:493-494
Competing interests:
None declared
Competing interests: No competing interests
Since the beginning of the drama with Vioxx (Rofecoxib) in September
2004, many studies have revised the odd ratios for cardiovascular risks of
Cox-2 inhibitors and of non selective NSAIDs. This study by Julia
Hippisley-Cox et al, BMJ 3 December 2005, revises the gastrointestinal
risks.
In the meantime, a clear picture is evolving, which shows that, as
predicted by Paracelsus five hundred years ago, the risks and benefits are
determined by their doses and potencies, more than by their selectivity to
the Cox enzymes, with the particular exception of low dose aspirin, which
permanently inhibits the platelets function without affecting the
endothelial prostacycline, producing the useful antithrombotic effect. But
in higher doses, aspirin is also toxic.
Besides their analgesic and anti-inflamatory effects, the Cox
inhibitors, both Cox-2 and non-selective, increase the cardiovascular,
renal and gastrointestinal risks, among other mechanisms, by losing the
gastroprotective and the antithrombotic and vasodilator effects,
increasing the water and sodium retention, the blood pressure (even with
Acetaminophen), the activity of pro inflammatory lipooxygenases and of
free oxygen radicals. Despite being less gastro-toxic, the Cox-2 still can
produce these adverse reactions, depending on their potency and dose.
Over 20 years ago, a report from England demonstrated that all NSAIDs
could produce some deaths per million of prescriptions. The least potent,
like Ibuprofen, produced the lowest risk of death: 1.5; with increasing
risk depending of increasing potency: Naproxen 4.6, Diclofenac 5.3,
Piroxicam 7.0, and Indometacin 7.1. The most potent, Meloxicam, did not
exist then. It would be useful to have data on its toxicity.
The experience with the cardiovascular risk of the new Cox-2
inhibitors follows a similar pattern. The most potent, like Rofecoxib and
Valdecoxib, were already retired.
We have no data from the following in potency, Etoricoxib, although it is
known to worsen hypertension. The least potent, like Celecoxib (still in
use), and Lumiracoxib,
help us to understand that the class effect is not of Cox-2, but of Cox
inhibitors in general.
The odds ratios for adverse effects oscillate, depending on the
endpoint, the reference parameters and the dose, from the most to the
least potent and toxic, as follows: Rofecoxib (1.32 to 3.58, in 12
studies), Valdecoxib (3 fold, 1 study), Diclofenac (1.55, 1 study), Other
NSAIDs (1.16 to 2.06, 3 studies), Other Cox-2 (1.45, 1 study), Naproxen
(1.14 to 1.5, 3 studies), Ibuprofen (1.09 to 1.24, 2 studies), Celecoxib
(0.43 to 1.26 in 6 studies, in 3 of them approximates 1.25), Lumiracoxib
(1.14, ns, in one study).
As end results, the patients show benefits in their suffering and
quality of life, but pay with an increased risk in other important aspects
that could eventually shorten their lifespan.
On the other hand, the continuous suffering of pain or inflammation
could also shorten their lifespan through the associated stress and
cardiovascular or gastrointestinal pathophysiological adverse reactions to
the disease mechanisms. Remember: “There is no free lunch”. We doctors
must explain to our patients and they decide.
We should recommend the less potent, less toxic, like Acetaminophen,
Celecoxib or Ibuprofen, (these two have similar odds ratios of toxicity),
in the lowest doss, for the shortest time, and also care for better
protection of the cardiovascular risk factors, like hypertension or
hyperlipidemia, or the gastrointestinal adverse effects.
Competing interests:
None declared
Competing interests: No competing interests
Sir,
Is is well known that one of the strongest risk factor for NSAiDs-
induced G-I toxicity is just previous G-I events.
The Authors in the patients and methods section have stated that patients
were excluded who had already had a diagnosis of an adverse upper
gastrointestinal event before the study period. By exluding those patients
with previous G-I events the high risk cases have been exluded from the
analysis. By narrowing the analysis to cases without previous G-I events
the Authors have induced a fundamental bias. It Is quite reasonable that
the large majority of the prescriptions of selective Cox 2 inhibitors had
been delivered to G-I high risk patients who have been exluded in the
analysis.
A second issue is the Authors do not provide the crude number of patients
with concurrent prescriptions of ulcer healing drugs by each
antinflamatory drug. It is quite reasonable that the vast majority of
ulcer healing drugs had been prescribed to patients on non selective
NSAIDs.
Third issue is the timing effect of concurrent antinflamatory medications.
The Authors stated that the odds ratios were adjusted for each other non-
steroidal anti-inflammatory drug group, smoking status, comorbidity,
deprivation, use of selective serotonin reuptake inhibitors, tricyclic
antidepressants, statins, aspirin, and ulcer healing drugs. But it is
important not only to adjust for previous NSAIDs use but also considering
how many different drugs had been used until index, their timing and their
nature (aspirin or diclofenac, i.e.), as well. It is well known that the
NSAiDs - induced G-I effect may persist for long time, thus interacting
with the following drug.
The Authors use a nested case-control design, a model constitutionally
prone to a number of confounding factors.
The data are prescriptions; what about real drug consumption?
The conclusions do not reflect the results, at least in part. If
patients on Celecoxib were too few what is the sense in comparing a class
of non-selective NSAIDs with one drug (rofecoxib)?
Competing interests:
None declared
Competing interests: No competing interests
While I note the somewhat predictable response from a Pfizer Inc
employee, I would agree that the authors do seem to dismiss celecoxib a
bit too easily.
What was of more interest to note, was the lack of protection when
using PPI's with diclofenac, in comparison to other NSAID's, especially in
view of how commonly it is prescribed.
It would be interesting to note the breakdown of risks for different
doses of ibuprofen, since this data seems to suggest it is safer than COX
II's
Competing interests:
None declared
Competing interests: No competing interests
Dear Editor,
The major conclusion drawn from the nested case-control study by
Hippisley-Cox et al [1]was that no consistent evidence was found of
enhanced safety against gastrointestinal events with any of the new cyclo-
oxygenase-2 inhibitors compared with non-selective non-steroidal anti-
inflammatory drugs. I feel this conclusion deserves some comments.
First, there was no significant increased risk for gastrointestinal
events in current users of celecoxib compared with controls [adjusted odds
ratio: 1.11 (0.87 to 1.41), P=0.39]. Furthermore, the adjusted odds ratio
for current use of rofecoxib in patients currently taking aspirin was 2.98
(2.24 to 3.99) whereas it was 1.22 (0.97 to 1.54) in those not currently
taking aspirin. While denoting a strong interaction between rofecoxib and
aspirin, these data do not support the view of a significant increased
hazard of gastrointestinal outcome in patients taking rofecoxib alone.
Finally, the authors acknowledged that any observational study may be
subject to residual confounding that cannot be fully corrected for. In
this respect, the present study was based on drug prescriptions and not
actual drug consumption. It should be stressed that a low compliance rate
with a given drug might result in an improved gastrointestinal safety
profile, and vice versa. Unfortunately, whether the adherence to drug
treatment was similar across all non-steroidal anti-inflammatory drugs
studied has not been assessed.
1. Hippisley-Cox J, Coupland C, Logan R. Risk of adverse
gastrointestinal outcomes in patients taking cyclo-oxygenase-2 inhibitors
or conventional non-steroidal anti-inflammatory drugs: population based
nested case-control study. BMJ 2005;331:1310-2.
Competing interests:
Invited speaker at Merck sponsored symposia and participation in clinical studies sponsored by Merck
Competing interests: No competing interests
Sir, I wish to address the Hippisley-Cox paper published in the 3
December issue of BMJ since the conclusions drawn do not accurately
reflect the data as presented. I am concerned that the disconnect between
the study findings and the conclusions in this article may cause confusion
amongst doctors and patients as they are making important health care
decisions.
This study found important differences between celecoxib and the
studied NSAIDs in terms of the risk of GI adverse effects. Specifically,
celecoxib was the only treatment that did not significantly increase the
risk of GI adverse events (adjusted RR 1.11, 95% CI 0.87 to 1.41) compared
to control patients. The authors comment that the number of celecoxib
taking patients was low, yet the upper limit of the 95% CI for celecoxib
is less than the lower limits for naproxen (1.73), diclofenac (1.78),
other NSAIDs (1.43) and aspirin (1.49) - supporting the relative GI safety
of celecoxib.
The findings are consistent with the results of other studies. NSAIDs
typically increase GI bleed risks two to four fold. For example, Mamdani
(BMJ 2002; 325:624-627) studied 1.3 million elderly patients in the
Canadian population and found that celecoxib was not associated with
increased risk of hospitalization for GI hemorrhage - as opposed to the
significantly increased risk seen with other NSAIDs.
We do not believe that the conclusion of the Hippisley-Cox article -
as well as the accompanying BMJ press release -- fully acknowledge safety
differences that the data showed among various arthritis treatment
options. In this study, celecoxib was the treatment with the lowest risk
of GI complications - important information for doctors and patients who
are making health decisions. The study conclusions and the BMJ press
release may create an impression that is not fully supported by the data
as presented.
Competing interests:
Employee, Pfizer Inc.
Competing interests: No competing interests
Poor headlines
The paper reviewed a limited number of agents showing great
differences between non specific NSAIDs. Yet the headline read that "COX
II were no different"; dismissing all in the class, implying that there
was no reason to prescribe any particular drug rather than just writing
"Rx NSAID 2tabs QDS"
This is not what the paper demonstrated at all! Apart from selection bias
as previously noted by other responders there was an imballance between
the various agents too few celecoxib and no etoricoxib or others let alone
etidolac, keoralac &c.
As a sufferer from reflux oesophagitis I have personal experience of the
different unpleasant effects of different drugs.
I have used this personal data in discussions with pain patients and found
it of more use than such as your headline grabbing article. Indeed the
dubious use of dubious "evidence" made me feel that standards have slipped
at the BMJ.
At the end of the day the message should be instilled that "pain killers"
of any type, but especialy the NSAIDs should only be used in short bursts
as an aid to restoring mobility NOT, in the chronic pain patient, to "kill
the pain"
and that the drug should be carefully tailored to the patient not to the
theoretical opinion.
Competing interests:
None declared
Competing interests: No competing interests