Lalashan.mcmaster.ca

Safety of Celecoxib vs Other Nonsteroidal
Dohme for rofecoxib and of the advisory board of Pfizer/Searle for celecoxib. Dr Anti-inflammatory Drugs
Ko¨hler and Dr Kuipers have received travel grants from MSD Sharp & Dohme andfrom Pfizer/Searle.
1. Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity with celecoxib
To the Editor: The results of the Celecoxib Long-term Arthri-
vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis: tis Safety Study (CLASS)1 support the hypothesis that cele- the CLASS Study: a randomized controlled trial. JAMA. 2000;284:1247-1255.
2. Lichtenstein DR, Wolfe MM. COX-2–selective NSAIDs: new and improved?
coxib alone does not induce anatomical lesions in the gastro- 3. Mizuno H, Sakamoto C, Matsuda K, et al. Induction of cyclooxygenase-2 in
However, in the subgroup of patients who were receiving gastric mucosal lesions and its inhibition by specific antagonist delays healing inmice. Gastroenterology. 1997;112:387-397.
aspirin for cardiovascular disease there was no difference in non- 4. Wolfe MM, Lichtenstein DR, Singh G. Medical progress: gastrointestinal tox-
steroidal anti-inflammatory drug (NSAID)–induced GI com- icity of nonsteroidal anti-inflammatory drugs. N Engl J Med. 1999;340:1888-1899.
plications between celecoxib and aspirin vs conventional ibu-profen or diclofenac and aspirin. Because cyclooxygenase To the Editor: Dr Silverstein and colleagues1 suggest that ce-
(COX)-1 is responsible for GI integrity under physiological con- lecoxib appears to be associated with significantly less renal ditions and because COX-2–specific inhibitors alone do not toxicity than therapy with NSAIDs, as do Drs Lichtenstein and cause anatomical lesions, we would have expected a lower rate Wolfe2 in their Editorial. It is possible, however, that particu- of ulcer complications in the group receiving celecoxib and as- lar NSAIDs have worse outcomes than nonselective NSAIDs pirin. As outlined in the accompanying Editorial by Drs Licht- and celecoxib. If so, it would be more appropriate to examine enstein and Wolfe,2 a possible explanation for this surprising reasons for such a finding (eg, concurrent drug regimens that finding might be a type II error due to the small sample size.
could potentiate renal toxicity of NSAIDs), rather than con- However, the small numerical difference observed between the clude that celecoxib had a lower rate of renal adverse events 2 groups suggests that even a statistically significant result in than did all nonselective NSAIDs. It is similarly possible that a study with a much larger sample size would translate into the rates of GI ulcers and ulcer complications could have been questionable clinical benefit. Alternatively, healing of aspirin- influenced by a particular subgroup.
induced anatomical lesions, which normally involves COX-2 Bettina Taylor, MD
induction, might be delayed by celecoxib.3 In the CLASS trial, Medscheme, Integrated Care
there was no washout period, and many patients were receiv- Bernard van de Wal, MD
ing NSAIDs at baseline. Therefore, many of these patients were Tygerberg Hospital
at risk for asymptomatic NSAID-induced gastropathy at the time Cape Town, South Africa
they entered the study. Thus, although COX-2–specific inhibi-tors are not specifically ulcerogenic, they may interfere with 1. Silverstein FE, Faich G, Goldstein JL, et al. Gastrointestinal toxicity with celecoxib
vs nonsteroidal anti-inflammatory drugs for osteoarthritis and rheumatoid arthritis:
repair of NSAID-induced gastropathy.
the CLASS study: a randomized controlled trial. JAMA. 2000;284:1247-1255.
Osteoarthritis and cardiovascular disease frequently coex- 2. Lichtenstein DR, Wolfe MM. COX-2–selective NSAIDs: new and improved?
JAMA. 2000;284:1297-1299.
ist in elderly patients. With regard to the data of the CLASStrial, the use of celecoxib for osteoarthritis in patients receiv- In Reply: While we agree with Dr Ko¨hler and colleagues that
ing aspirin for cardiovascular disease may not be as safe as ex- the CLASS trial supports the hypothesis that COX-2–specific pected from the hypothesis that celecoxib alone does not in- inhibitors will produce fewer ulcers and ulcer complications duce lesions in the GI tract. At present we do not think that than conventional NSAIDs, we disagree that the data suggest celecoxib offers a proven advantage for patients who receive the possibility that celecoxib interferes with healing of aspirin- long-term aspirin therapy. For such patients who require ad-ditional analgesia, we think that a better choice would be con-ventional NSAIDs with misoprostol or omeprazole.4 GUIDELINES FOR LETTERS. Letters discussing a recent JAMA article should
be received within 4 weeks of the article’s publication and should not exceed 400
Lars Ko¨hler, MD
words of text and 5 references. Letters reporting original research should not ex- Henning Zeidler, MD
ceed 500 words and 6 references. All letters should include a word count. Lettersmust not duplicate other material published or submitted for publication. Letters Sonja Merkesdal, MD
will be published at the discretion of the editors as space permits and are subject Jens G. Kuipers, MD
to editing and abridgment. A signed statement for authorship criteria and respon- Division of Rheumatology
sibility, financial disclosure, copyright transfer, and acknowledgment is required Medical School Hanover
for publication. Letters not meeting these specifications are generally not con-sidered. Letters will not be returned unless specifically requested. Also see Instruc- Hanover, Germany
tions for Authors (July 5, 2000). Letters may be submitted by surface mail: LettersEditor, JAMA, 515 N State St, Chicago, IL 60610; e-mail: JAMA-letters Financial Disclosure: Dr Ko¨hler and Dr Zeidler receive funding from Merck/USA
@ama-assn.org; or fax (please also send a hard copy via surface mail): (312) 464-5824.
for a research project (Chlamydia pneumoniae in arteriosclerosis—role of cyclooxy-genase-2) and have received educational grants from MSD Sharp & Dohme and Letters Section Editors: Stephen J. Lurie, MD, PhD, Senior Editor; Phil B.
from Pfizer/Searle. Dr Zeidler is a member of the advisory board of MSD Sharp & Fontanarosa, MD, Executive Deputy Editor.
2000 American Medical Association. All rights reserved.
(Reprinted) JAMA, December 27, 2000—Vol 284, No. 24 3123
induced GI mucosal lesions. A prior study1 has found that the 1. Weil J, Colin JD, Langman M, Lawson D, Logan R, Murphy M et al. Prophy-
lactic aspirin and risk of peptic ulcer bleeding. BMJ. 1995;310:827-830.
relative risk of an ulcer complication in aspirin users vs non- 2. Emery P, Zeidler H, Kvien TK, et al. Celecoxib versus diclofenac in long-term
aspirin users is approximately the same as the relative risk of management of rheumatoid arthritis: randomised double-blind comparison. Lan- an ulcer complication in aspirin users receiving celecoxib vs cet. 1999;354:2106-2111.
3. Simon LS, Weaver AL, Graham DY, et al. Anti-inflammatory and upper gas-
nonaspirin users receiving celecoxib that we reported.
trointestinal effects of celecoxib in rheumatoid arthritis: a randomized controlled Two additional lines of evidence also argue against the hypoth- trial. JAMA. 1999;282:1921-1928.
4. Goldstein JL, Agrawal NM, Silverstein FE, et al. Effect of celecoxib versus placebo
esis that COX-2–specific inhibitors interfere with healing of upper and NSAIDs in healing of gastroduodenal erosions in arthritis patients: analysis of con- GI mucosal lesions. First, in a study in patients with rheuma- trolled endoscopic studies [abstract]. Ann Rheum Dis. 2000;59(suppl 1):159.
5. Rodriguez LA, Varas C, Patrono C. Differential effects of aspirin and non-
toid arthritis without an NSAID washout period, the incidence aspirin nonsteroidal antiinflammatory drugs in the primary prevention of myocar- of endoscopic ulcers in patients receiving celecoxib was equiva- dial infarction in postmenopausal women. Epidemiology. 2000;11:382-387.
lent to that observed in studies of patients with rheumatoid arthri- These letters were shown to Dr Wolfe, who declined to reply.—ED.
tis in which there was an NSAID washout period.2,3 Thus, patientsentering the first study with NSAID-induced mucosal injury pre-sumably healed while receiving celecoxib. Second, we have found Glycoprotein Inhibitors and Fibrinolysis
that conventional NSAIDs, but not celecoxib, interfere with the in Myocardial Infarction
healing of erosive gastritis as assessed by endoscopy.4 Finally, while the benefit of celecoxib may be attenuated by To the Editor: Drs Bhatt and Topol discussed the role of glyco-
the use of concomitant low-dose aspirin, the CLASS trial estab- protein (Gp) IIb/IIIa inhibitors in acute coronary syndromes,1 but lished that celecoxib provides other safety benefits (eg, de- did not comment on the use of these agents in patients who have creases in bleeding, renal, and hepatic adverse effects relative to received fibrinolytic therapy. It is not uncommon for patients who conventional NSAIDs) that are not provided by the addition of have received such treatment to require further acute interven- a gastroprotective agent to conventional NSAID therapy. More- tion. Are there any data about the risk-benefit ratio of Gp IIb/IIIa over, NSAIDs alone do not provide a cardioprotective effect as agents in patients who may subsequently require urgent angio- does aspirin, and the combination of NSAIDs and low-dose as- plasty following failed fibrinolytic therapy? pirin may be additively toxic to the upper GI mucosa.1,5 Bruce L. Ring, MD
In answer to Drs Taylor and van de Wal’s question about com- Caritas Good Samaritan Medical Practice
parisons between celecoxib and individual NSAIDs, our study was Stoughton, Mass
designed to compare celecoxib to NSAIDs as a class and was pow- 1. Bhatt DL, Topol EJ. Current role of platelet glycoprotein llb/IIIa inhibitors in acute
ered accordingly. In general, conventional NSAIDs have similar coronary syndromes. JAMA. 2000; 284:1549-1558.
adverse events. The use of 2 distinct NSAIDs as the comparatortreatment for celecoxib is, in fact, a particular strength of our study In Reply: Fibrinolysis has a proven role in the treatment of myo-
and renders the conclusions more generalizable and less suscep- cardial infarction (MI) with ST-segment elevation. Glycopro- tible to the potential interpretative biases that may result from tein IIb/IIIa inhibition is established as an adjunct to primary per- using a single NSAID comparator. Regardless, with respect to the cutaneous intervention (PCI) for acute MI. The Global Use of various study outcomes, celecoxib was consistently safer and bet- Strategies To Open Occluded Coronary Arteries (GUSTO) III trial ter tolerated than the individual NSAID comparators, although examined the use of abciximab in patients who had ongoing is- not every individual comparison was statistically significant. The chemia following fibrinolysis with full-dose reteplase or al- comparisons between celecoxib and individual NSAIDs must also teplase and subsequently underwent rescue PCI with abcix- be viewed in light of the fact that celecoxib was administered at imab.1 There was a trend towards a lower 30-day mortality in the a supratherapeutic dose while the conventional NSAIDs were ad- 83 patients who received abciximab, although these patients did ministered at the commonly used therapeutic doses.
have more episodes of severe bleeding. An analysis of the pa- We hope to share additional analyses of this database to fur- tients in Strategies for Patency Enhancement in the Emergency ther elucidate the therapeutic properties of COX-2–specific Department (SPEED) who underwent PCI within 60 to 90 min- utes of receiving abciximab and reduced-dose reteplase revealed James B. Lefkowith, MD
a low rate of mortality, MI, or urgent revascularization at 30 days.2 G. Steven Geis, MD, PhD
Similarly, an analysis from the Integrilin and Reduced Dose of Pharmacia Clinical Research
Thrombolytics in Acute Myocardial Infarction (INTRO-AMI) trial Skokie, Ill
showed that early PCI after reduced-dose alteplase and eptifi- Fred Silverstein, MD
batide also resulted in a relatively low rate of death or reinfarc- Department of Medicine
tion.3 This concept of prompt chemical reperfusion followed by University of Washington
early PCI is now commonly referred to as “facilitated PCI.” SPEED served as the pilot for the GUSTO IV acute MI trial, which is currently examining standard-dose abciximab plus re- Financial Disclosure: Dr Lefkowith is an employee of Pharmacia, manufacturer
duced-dose double-bolus reteplase in a cohort of 16 600 pa- of celecoxib. Dr Silverstein is a paid consultant for Pharmacia. The CLASS trial wasfunded by Pharmacia.
tients. This large sample size will allow a much more accurate 3124 JAMA, December 27, 2000—Vol 284, No. 24 (Reprinted)
2000 American Medical Association. All rights reserved.
assessment of the safety of facilitated PCI. Thus, the historical possible that some patients in this study did not have GER, but dichotomy of either mechanical or pharmacological reperfu- in fact had EE. Interestingly, a preliminary report suggests an sion may be coming to an end. Unlike older reports of early association of chromosome 13q with atopy in infancy.6 PCI after fibrinolysis, PCI after reduced-dose fibrinolytics plus Glenn T. Furuta, MD
Gp IIb/IIIa inhibitors appears safe and efficacious. In the fu- Samuel Nurko, MD
ture, treatment for acute ST-segment elevation MI will likely Athos Bousvaros, MD
consist of the combination of reduced-dose fibrinolytics and Don Antonioli, MD
Gp IIb/IIIa inhibitors, followed by PCI.
Kamran Badizadegan, MD
Department of Pathology
Deepak L. Bhatt, MD
and the Combined Program in Pediatric GI Nutrition
Eric J. Topol, MD
Children’s Hospital
Department of Cardiology
Boston, Mass
Cleveland Clinic Foundation
Cleveland, Ohio
1. Hu FZ, Preston RA, Post JC, et al. Mapping of a gene for severe gastroesopha-
geal reflux to chromosome 13q14. JAMA. 2000;284:325-334.
Financial Disclosure: Dr Bhatt has had meeting expenses and honoraria pro-
2. Orenstein SR, Shalaby TM, Di Lorenzo C, Putnam PE, Sigurdsson L, Kocoshis
vided by Centocor Inc and COR Therapeutics Inc and has received grant support SA. The spectrum of pediatric eosinophilic esophagitis beyond infancy: a clinical series of 30 children. Am J Gastroenterol. 2000;95:1422-1430.
1. Miller JM, Smalling R, Ohman EM, et al. Effectiveness of early coronary angio-
3. Walsh SV, Antonioli DA, Goldman H, et al. Allergic esophagitis in children: a
plasty and abciximab for failed thrombolysis (reteplase or alteplase) during acute myo- clinicopathological entity. Am J Surg Pathol. 1999:23:390-396.
cardial infarction (results from the GUSTO-III trial). Am J Cardiol. 1999;84:779-784.
4. Kelly KJ, Lazenby AJ, Rowe PC, Yardley JH, Perman JA, Sampson HA. Eosino-
2. Herrmann HC, Moliterno DJ, Bode C, Betriu A, Lincoff AM, Ohman EM. Com-
philic esophagitis attributed to gastroesophageal reflux: improvement with an amino bination abciximab and reduced-dose reteplase facilitates early PCI in acute MI: acid-based formula. Gastroenterology. 1995:109;1503-1512.
results from the SPEED trial. Circulation. 1999;100:I-188.
5. Liacouras CA, Wenner WJ, Brown K, Ruchelli E. Primary eosinophilic esopha-
3. L’Allier PL, Lopez JF, Vrobel TR, Strony JS, Brener SJ. Favorable results with im-
gitis in children: successful treatment with oral corticosteroids. J Pediatr Gastro- mediate coronary intervention after low dose t-PA and eptifibatide for acute MI: enterol Nutr. 1998:26;380-385.
results from the INTRO-AMI trial. Circulation. 1999;100:I-359.
6. Nickel R, Beyer K, Huang SK, Barnes KC, Wahn U. Genetic markers of atopy in
infancy: results from the German Multicenter Allergy Study. Clin Exp Allergy. 1999:
299(suppl 4);23-25.
The Spectrum of Pediatric
In Reply: Dr Furuta and colleagues raise an interesting and
Gastroesophageal Reflux
important point in asking for a detailed description of the spec-trum of phenotypes of the affected subjects in our genetic link- To the Editor: Ms Hu and colleagues1 sought to identify the
age study of severe pediatric GER. Clearly, the symptoms of genetic map for pediatric gastroesophageal reflux (GER). How- patients with primary nonsyndromic pediatric GER, EE, and ever, the authors do not report diagnostic criteria for GER, and hiatal hernia can appear to be similar. It is thus desirable when therefore we are concerned about the uniformity of their sample.
performing genetic linkage studies to ensure that all of the af- Children with chronic vomiting, spitting up, dysphagia, and fected individuals have the same phenotype.
food refusal are commonly diagnosed with GER and are treated Accordingly, we attempted to restrict the phenotypic (and there- with acid suppression. Alternatively, these symptoms may have fore genotypic) heterogeneity within our sample by specifically other causes such as anatomical malformations, food aller- excluding children with diagnoses of EE or hiatal hernia. Be- gies, functional bowel diseases, or metabolic disorders. Clini- cause the diagnosis of pediatric GER was made retrospectively cal criteria used to make an initial diagnosis of GER can be for some of the adult members of the study families, it was not subjective, but when symptoms are refractory to medical man- possible to formally rule out these confounding conditions for agement, patients will often undergo evaluation including these individuals. However, it is unlikely that the affected par- esophagogastroduodenoscopy with biopsy. One emerging cause ents of affected offspring with primary nonsyndromic GER would of such refractory symptoms is eosinophilic esophagitis (EE), have a materially different disease. The fact that the same ge- which has been reported in at least 81 children in 4 different netic region on chromosome 13q14 was found to be inherited academic institutions.2-5 The affected squamous epithelium con- with the GER phenotype in all 5 families argues strongly that we tains a large number of intraepithelial eosinophils, despite at were successful in our attempts to choose a phenotypically ho- least 2 months of acid suppression.2 Treatment with fundopli- mogeneous set of study families. We do not wish to imply that cation is not therapeutic, but corticosteroids or an elemental all forms of pediatric GER result from mutations in the same gene; diet provide effective symptom control and improvement of in fact, we suspect that there will be genetic heterogeneity among esophageal histology.3,4 To date, the incidence, etiology, and persons with similar and differing GER phenotypes.
long-term consequences of EE are not well defined.
Because the differential diagnosis for children with upper in- Garth D. Ehrlich, PhD
testinal symptoms is not limited to GER, it would be interesting Fen Ze Hu, MS
Robert Preston, PhD
to know the diagnostic criteria that Hu et al used. History and J. Christopher Post, MD, PhD
barium studies alone, which were used in 17 of 36 patients, are Center for Genomic Sciences and Department of Human Genetics
not sufficient to differentiate between GER and EE. Since clini- Allegheny General Hospital and MCP Hahnemann University
cal practice suggests that the incidence of EE is increasing, it is Pittsburgh, Pa
2000 American Medical Association. All rights reserved.
(Reprinted) JAMA, December 27, 2000—Vol 284, No. 24 3125
Lead in Calcium Supplements
Lead-related complications are compounded for pregnantwomen, older persons attempting to prevent or minimize os- To the Editor: We are gratified that, in his Editorial1 accompa-
teoporosis, and children with suspected milk allergy. A US Food nying our article,2 Dr Heaney agreed that the labeling of calcium and Drug Administration official stated, “However, individu- supplements with their lead levels would be a sound plan. We als who consume more than two to three times the recom- also share his hope that calcium intake may attenuate lead ab- mended dose of bonemeal would be at greater risk if the lead sorption; however, this optimism has not yet been supported by content of the particular bonemeal product is high.”4 I have the literature. In the study he cites,3 as well as in the more recent previously underscored the wide variations in individual comprehensive findings from the same group,4 the protective ef- fect of calcium was limited to 14-day-old chickens that were cal- The documentation of significant amounts of additional toxic cium deficient. Those authors suggested that the lead-calcium axis metals in vitamin, mineral, and calcium supplements that were was highly complex and varied with the duration of intake. Fur- analyzed by different laboratories2,5,6 also should be of con- thermore, the data suggested that some lead absorption may be cern. They include arsenic, mercury, aluminum, iron, anti- uncoupled from calcium transport. Unfortunately, the data that mony, nickel, tin, fluoride, cadmium, and others.
we cited5 failed to confirm the alleged protective effect of dietary H. J. Roberts, MD
Palm Beach Institute for Medical Research
Heaney also correctly pointed out that there are a variety of West Palm Beach, Fla
dietary foods that may have a relatively high lead content. Nev-ertheless we view those “single source” items as being the ex- 1. Heaney RP. Lead in calcium supplements: cause for alarm or celebration? JAMA.
ception rather than the rule, and US residents correctly should 2000;284:1432-1433.
2. Roberts HJ. Potential toxicity due to dolomite and bone meal. South Med J.
demand a long-term average exposure of no more than 6 µg/ day. The weight of the evidence indicates that this is a realistic 3. Ross EA, Szabo NJ, Tebbett IR. Lead content of calcium supplements. JAMA.
2000;284:1425-1429.
and attainable goal and that failure to achieve these low levels 4. Millar SA. Lead and calcium supplements. JAMA. 1987;257:1810.
would be a public disservice. We would instead challenge manu- 5. Roberts HJ. Dolomite as a source of toxic metals. N Engl J Med. 1981;304:
423.
facturers to improve their quality control and to label their prod- 6. Roberts HJ. More on dolomite. N Engl J Med. 1981;304:1367.
ucts accordingly so that consumers can make informed choices.
It is our hope that some of the brands without detectable lev- In Reply: In response to Dr Ross and colleagues, my state-
els have already accomplished this goal and thus the manu- ment about interference of calcium in lead absorption, while facturers have not had the motivation to include this informa- supported and predicted by animal studies, was based on stud- ies in humans. At a population level, blood lead concentration Edward A. Ross, MD
is inversely correlated with reported calcium intake.1,2 In con- Division of Nephrology, Hypertension, and Transplantation
trolled experiments, absorption of tracer-labeled lead sources College of Medicine
is reduced by factors of 6- to 50-fold by coingestion of cal- Nancy J. Szabo, PhD
Ian R. Tebbett, PhD
The mechanism of lead absorption is not relevant. Transfer Department of Physiological Sciences
across the intestinal membrane, whether active or passive, is College of Veterinary Medicine
limited, and when calcium is present in the digestate, it di- University of Florida
lutes any lead also present, thereby reducing the amount that Gainesville
crosses the mucosal barrier, as the above-referenced human stud- 1. Heaney RP. Lead in calcium supplements: cause for alarm or celebration? JAMA.
Dr Roberts provides no evidence to support his claim of harm 2. Ross EA, Szabo NJ, Tebbett IR. Lead content of calcium supplements. JAMA.
from lead-containing supplements. He cites only old litera- 2000;284:1425-1429.
3. Fullmer CS, Rosen JF. Effect of dietary calcium and lead status on intestinal cal-
ture describing the lead content of dolomite and bonemeal. Nei- cium absorption. Environ Res. 1990;51:91-99.
ther supplement was discussed by Ross et al,5 and I did not re- 4. Fullmer CS. Intestinal calcium and lead absorption: effects of dietary lead and
calcium. Environ Res. 1991;54:159-169.
fer to them in my Editorial. Dolomite is not a particularly good 5. Sargent JD, Dalton MA, O’Connor GT, Olmstead EM, Klein RZ. Randomized
calcium supplement, and there is little to recommend it. Bone- trial of calcium glycerophosphate-supplemented infant formula to prevent leadabsorption. Am J Clin Nutr. 1999;69:1224-1230.
meal, however, is another issue. Its lead content is a functionof environmental exposure of the animal contributing the bones.
To the Editor: In his Editorial, Dr Heaney1 states that there is
For horses and cattle, that would be grasses contaminated by little cause for alarm about lead contamination of calcium exhaust emissions, which have been reduced dramatically in supplements. However, the ongoing use of such contami- nated products, especially dolomite and bonemeal, can have If Roberts has data on lead content of current bonemeal serious neurologic, gastrointestinal, cutaneous, and hemato- sources, that would be relevant. However, it is inappropriate logic effects. In a previous report,2 I countered his criticism of to estimate the lead content of current bonemeal on the basis Ross et al3 who “do not demonstrate the dangers they fear.” of old data because the source of contamination has changed.
3126 JAMA, December 27, 2000—Vol 284, No. 24 (Reprinted)
2000 American Medical Association. All rights reserved.
Because bonemeal contains phosphorus, which may confer ben- fit into a specific category.4 More precise and simple systems, efits in some instances when calcium supplementation is called such as descriptive grading systems,5 should be explored and for, it should not be dismissed out of hand.
The point of my Editorial was not to say that calcium sources Nirav R. Shah, MD, MPH
low in lead would not be better than high-lead sources—just Department of Internal Medicine
that current lead levels are already so low and the calcium that Yale-New Haven Hospital
carries the lead provides such a high protective factor that ex- New Haven, Conn
penditures by the industry to reduce lead levels even further 1. Guyatt GH, Haynes RB, Jaeschke RZ, et al, for the Evidence-Based Medicine
would take us well past the point of diminishing returns.
Working Group. Users’ Guides to the medical literature: XXV: evidence-based medi- Calcium deficiency was 1 of 2 national nutrient inadequa- cine: principles for applying the User’s Guides to patient care. JAMA. 2000;284:1290-1296.
cies identified by the federal interagency task force “Healthy 2. Benson K, Hartz AJ. A comparison of observational studies and randomized,
People 2000.”6 There is today a recognized calcium deficiency controlled trials. N Engl J Med. 2000;342:1878-1886.
3. Concato J, Shah N, Horwitz RI. Randomized, controlled trials, observational stud-
in this country and any scare about the safety of the calcium ies, and the hierarchy of research. N Engl J Med. 2000;342:1887-1892.
supply would, in my judgment, do far more harm than good.
4. Woloshin S. Arguing about grades. Eff Clin Pract. 2000;3:94-95.
5. Roman SH, Silberzweig SB, Siu AL. Grading the evidence for diabetes perfor-
Robert P. Heaney, MD
mance measures. Eff Clin Pract. 2000;3:85-91.
Department of Medicine
Creighton University
In Reply: Dr Shah suggests that systematic reviews of obser-
Omaha, Neb
vational studies provide better evidence than single random-ized trials. We disagree strongly. Shah’s proposal would, for Financial Disclosure: Dr Heaney serves as a consultant for and his laboratory does
work for many of the major calcium suppliers in the United States, including food
instance, rank a meta-analyses of observational studies of hor- fortifiers, wholesale suppliers, the dairy industry, and supplement manufacturers.
mone replacement therapy1 (which suggested a 50% reduc- However, Dr Heaney is a full-time employee of Creighton University and derives tion in relative risk of coronary events) higher than the Heart no personal financial gain from any of these relationships, as all consulting rev-enues are paid directly to Creighton University.
and Estrogen/progestin Replacement Study (HERS) RCT,2 which 1. Mahaffey KR, Gartside PS, Glueck CJ. Blood lead levels and dietary calcium in-
found no effect of hormone replacement therapy for second- take in 1 to 11 year-old children: the Second National Health and Nutrition Ex- ary prevention of coronary events. We would rank the HERS amination Survey, 1976 to 1980. Pediatrics. 1986;78:257-262.
RCT higher and, pending further RCT evidence, would not rec- 2. Muldoon SB, Cauley JA, Kuller LH, Scott J, Rohay J. Lifestyle and sociodemo-
graphic factors as determinants of blood lead levels in elderly women. Am J Epi-
ommend hormone replacement therapy to women to de- crease their risk of cardiovascular events. When consistent bi- 3. Blake KC, Barbezat GO, Mann M. Effect of dietary constituents on the gastro-
intestinal absorption of 203Pb in man. Environ Res. 1983;30:182-187.
ases exist in observational studies (such as the tendency for 4. Heard MJ, Chamberlain AC. Effect of minerals and food on uptake of lead from
women with lower cardiovascular risk to receive hormone re- the gastrointestinal tract in humans. Human Toxicol. 1982;1:411-415.
5. Ross EA, Szabo N, Tebbett IR. Lead contamination of calcium supplements. JAMA.
placement therapy), a meta-analysis offers no protection.
However, we agree that poorly conducted RCTs can pro- 6. Healthy People 2000 [conference edition]. Washington, DC: US Dept of Health
and Human Services; 1990.
duce misleading results and that is why we have offered Users’Guides to distinguish methodologically rigorous RCTs fromthose that are flawed. While we also agree that observational What Is the Best Evidence
studies can produce results similar to those of RCTs and that for Making Clinical Decisions?
meta-analyses of observational studies can yield results simi-lar to meta-analyses of RCTs, they also can produce very dif- To the Editor: Dr Guyatt and colleagues1 suggest a hierarchy
ferent results. Prominent examples of discordant results in- of study design to evaluate the strength of evidence for mak- clude the effects of physiologic vs ventricular pacing in patients ing treatment decisions. In promoting the superiority of single with conduction abnormalities,3 the impact of diuretic therapy randomized trials over systematic reviews of observational stud- on cardiovascular disease in diabetic patients,4 and the pre- ies, the authors risk making the same mistake that evidence- vention of cardiovascular events with beta carotene5 and vita- based medicine professes to avoid—relying on personal expe- min E.6 The problem is that until one conducts the RCTs, cli- rience and disregarding the best available evidence.
nicians do not know if they are dealing with a concordant or Randomized trials are not always the best way to evaluate treatment in a given clinical scenario, and poorly conducted Our hierarchy is intended to address issues of therapeutic randomized trials may lead to erroneous conclusions and pro- benefit; for diagnostic accuracy and prognosis, other hierar- duce contradictory results. Evidence suggests that observa- chies are necessary. For questions of harm, RCTs will usually tional studies and randomized controlled trials (RCTs) can pro- be unavailable, and clinicians must, in most instances, rely on duce similar estimates of the effects of treatment,2 and that meta- observational studies. Furthermore, we are not suggesting that analyses of observational studies produce results that are similar all RCTs are superior to all observational studies, but that RCTs as a class provide stronger evidence than observational stud- Furthermore, current rigid hierarchy promoted by Guyatt ies as a class. Our hierarchy does not constitute the final word, et al will necessarily place less value on evidence that does not and evidence-based medicine and its tools will continue to evolve 2000 American Medical Association. All rights reserved.
(Reprinted) JAMA, December 27, 2000—Vol 284, No. 24 3127
and improve. Nevertheless, our hierarchy is simple, logical, and Table 1. Treatment Response by Baseline CD4 Cell Count Level and
evidence-based. We continue to recommend it to clinicians striv- ing to effectively apply findings from the medical literature to Variable
No. of Patients
Response
Response
Gordon H. Guyatt, MD, MSc
David Naylor MD, PhD
W. Scott Richardson, MD
Lee Green, MD, MPH
R. Brian Haynes, MD, PhD
Mark C. Wilson, MD, MPH
Deborah J. Cook, MD, MSc
Roman Z. Jaeschke MD, MSc
*Data are presented as number (percent) of patients in the row with the indicated out- Author affiliations are listed in Guyatt GH, Haynes RB, Jaeschke RZ, et al, for the
Evidence-Based Medicine Working Group. Users’ Guides to the medical litera- †P values for initial response and durable response were Ͻ.001 for both CD4 level and ture: XXV: evidence-based medicine: principles for applying the User’s Guides to patient care. JAMA. 2000;284:1290-1296.
1. Stampfer MJ, Colditz GA. Estrogen replacement therapy and coronary heart
disease: a quantitative assessment of the epidemiologic evidence. Prev Med. 1991;
bination antiretroviral therapy were included. We examined 2 20:47-63.
2. Hulley S, Grady D, Bush T, Furberg C, Herrington D, Riggs B, Vittinghoff E, for
outcomes: reduction of HIV RNA to less than 400 copies/mL the Heart and Estrogen/progestin Replacement Study (HERS) Research Group. Ran- on at least 1 occasion within 6 months of starting treatment domized trial of estrogen plus progestin for secondary prevention of coronary heartdisease in postmenopausal women. JAMA. 1998;280:605-613.
(initial response) and response with no subsequent elevation 3. Connolly SJ, Kerr CR, Gent M, et al, for the Canadian Trial of Physiologic Pac-
of HIV RNA to more than 1000 copies/mL (durable response).
ing Investigators. Effects of physiologic pacing versus ventricular pacing on the Baseline characteristics were compared with outcomes using risk of stroke and death due to cardiovascular causes. N Engl J Med. 2000;342:1385-1391.
4. Curb JD, Pressel SL, Cutler JA, et al, for the Systolic Hypertension in the Elderly
Results. A total of 553 patients starting highly active anti-
Program Cooperative Research Group. Effect of diuretic-based antihypertensivetreatment on cardiovascular disease risk in older diabetic patients with isolated retroviral therapy received treatment for at least 6 months and systolic hypertension. JAMA. 1996;276:1886-1892.
were eligible for analysis, with a mean follow up of 824 days.
5. Hennekens CH, Buring JE, Manson JE, et al. Lack of effect of long-term sup-
plementation with beta carotene on the incidence of malignant neoplasms and
The patients were 74% male, 70% black, 44% injecting drug cardiovascular disease. N Engl J Med. 1996;334:1145-1149.
users, 35% homosexual men, 18% heterosexually infected, and 6. Yusuf S, Dagenais G, Pogue J, Bosch J, Sleight P, for the Heart Outcomes
Prevention Evaluation Study Investigators. Vitamin E supplementation and car-
had a mean age of 37 years. The median baseline CD4 cell count diovascular events in high-risk patients. N Engl J Med. 2000;342:154-160.
was 202/mm3 and median baseline HIV RNA was 35810 copies/mL. Treatment included nucleoside analogues in all patientsand 1 protease inhibitor in 76%, 2 protease inhibitors in 21%, RESEARCH LETTER
An HIV RNA of less than 400 copies/mL was achieved by Association of Initial CD4 Cell Count and
351 patients (63.5%). Only 147 patients had durable re- Viral Load With Response to Highly Active
sponses (42% of initial responders, 27% of total). TABLE 1 shows
Antiretroviral Therapy
the proportion of patients with an initial response and a du-rable response by baseline CD4 and viral load level. Patients To the Editor: Considerable uncertainty and debate exist re-
with initial CD4 cell counts greater than 350/mm3 were sig- garding the appropriate time for initiating antiretroviral therapy nificantly more likely to achieve both an initial and a durable for the treatment of HIV (human immunodeficiency virus).1 response than patients with lower baseline CD4 cell count lev- British guidelines suggest that treatment should be initiated els. Patients with initial CD4 cell counts greater than 500/ based on CD4 cell count alone,2 while US guidelines recom- mm3 had a slightly higher rate of initial and durable response mend treatment based on a CD4 cell count below 500/mm3 or than those with initial CD4 cell counts of 350 to 500 per mm3 a viral load above 10 000 to 20 000 per mL.3,4 No prospective (82% vs 73% and 42% vs 35%, respectively). Similarly, pa- study of starting vs delaying the initiation of highly active an- tients with baseline HIV RNA levels of 25000 copies/mL or less tiretroviral therapy in patients with higher viral loads has been were significantly more likely to have virologic responses than Methods. We retrospectively compared the outcomes of start-
TABLE 2 shows conditional logistic regression analyses us-
ing highly active antiretroviral therapy at varying CD4 cell counts ing initial response and durable response as the dependent vari- and viral loads in a clinical cohort of HIV-infected patients in ables. Patients with initial CD4 cell count levels greater than an ongoing study.5 Patients who were starting their first com- 350/mm3 were almost twice as likely to achieve an initial and bination antiretroviral regimen with a protease inhibitor or non- a durable response than patients with lower CD4 cell count, nucleoside reverse transcriptase inhibitor after July 1, 1996, were after adjustment for other factors, including viral load. Pa- analyzed. Only patients who received at least 6 months of com- tients with initial RNA levels of 25 000 copies/mL or less were 3128 JAMA, December 27, 2000—Vol 284, No. 24 (Reprinted)
2000 American Medical Association. All rights reserved.
vational data are not derived from a randomized trial, we think Table 2. Conditional Logistic Regression Analyses for Achieving an
they indicate that highly active antiretroviral therapy is more likely to succeed in suppressing viremia when given before immuno- Odds Ratio (95% CI)
Odds Ratio (95% CI)
for Achieving
for Achieving
deficiency has progress to a moderate or severe level.
Variable†
Initial Response
Durable Response
Richard E. Chaisson, MD
Jeanne C. Keruly, RN
Richard D. Moore, MD, MHSc
Department of Medicine
Johns Hopkins University School of Medicine
Baltimore, Md
Funding/Support: This study was supported by National Institutes for Health grants
R01-DA-11602, AI-01637, and FD-U-001640.
1. Henry K. The case for more cautious, patient-focused antiretroviral therapy.
Ann Intern Med. 2000;132:306-311.
†Age, sex, and HIV transmission categories were not significant. The reference catego- 2. Gazzard B, Moyle G, for the BHIVA Guidelines Writing Committee. 1998 re-
ries used in the logistic analyses were CD4 cell count Ͻ200 and RNA Ͼ100 000 vision to the British HIV Association guidelines for antiretroviral treatment of HIV seropositive individuals. Lancet. 1998;352:314-316.
3. Panel on Clinical Practices for Treatment of HIV Infection. Guidelines for the
also significantly more likely to achieve initial and durable re- use of antiretroviral agents in HIV-infected adults and adolescents: the living docu-ment, January 28, 2000. Available at: http://hivatis.org/trtgdlns.html. Accessibil- sponses than those with higher viral loads, after adjustment for CD4 cell count and other factors. The number of nucleoside 4. Carpenter CC, Cooper DA, Fischl MA, et al. Antiretroviral therapy in adults:
updated recommendations of the International AIDS Society–USA Panel [re-
analogue drugs received prior to combination therapy was in- view]. JAMA. 2000;283:381-390.
versely associated with achieving an undetectable viral load.
5. Moore RD, Chaisson RE. Natural history of HIV infection in the era of combi-
Age, sex, and injection drug use history were not associated nation antiretroviral therapy. AIDS. 1999;13:1933-1942.
with outcomes, while black race was associated with a lowerodds of achieving a durable response.
Comment. These data suggest that the initial timing of an-
tiretroviral therapy should consider both CD4 cell count and CORRECTIONS
viral load. Patients with CD4 cell count levels greater than 350/mm3 or with viral loads of 25 000 copies/mL or less had more Omitted Acknowledgment: In the Original Contribution entitled “Postlicensure
Safety Surveillance for Varicella Vaccine” published in the September 13, 2000,
favorable initial and durable responses than those with lower issue of THE JOURNAL (2000;284:1271-1279), several contributors to the data from CD4 cell counts or higher viral loads. It is possible that pa- the Vaccine Adverse Event Reporting System (VAERS) were omitted from the Ac-knowledgment. During the primary analysis period (March 1995 through July 25, tients with lower CD4 cell counts and higher viral loads have 1998) the Columbia University laboratory of Drs Phillip LaRussa, Sharon Stein- more a virulent phenotype of HIV, although there is no evi- berg, and Anne Gershon conducted the polymerase chain reaction and restrictionfragment length polymorphism analyses to identify varicella-zoster virus and to dence of reduced in vitro susceptibility to antiretroviral drugs distinguish the vaccine strain from wild-type strains. They worked under contract of viruses from such patients. Because we focused on the spe- to the Varicella Zoster Virus Identification Program at Merck & Co Inc. VAERS re- cific outcome of virologic response rather than survival, we do ceived the results of these analyses from Merck as individual case reports. Al-though the article mentioned that the National Varicella Reference Laboratory at not think these data are subject to lead time bias.
the Centers for Disease Control and Prevention performs these analyses, during Although guidelines have been developed based on the natu- the time covered by the study only the Columbia University laboratory had thiscapability. In cooperation with the Centers for Disease Control and Prevention, ral history of untreated HIV infection, rather than the likeli- Merck also operates the varicella vaccine pregnancy registry, where Kristine Shields hood of treatment success, our results support both the British provided detailed information about multiple cases.
and US guidelines, which recommend starting treatment for pa- Incorrect Value: In the Clinical Crossroads entitled “An 80-Year-Old Man With
tients before the CD4 cell count falls to less than 350/mm3 and Memory Loss” published in the February 23, 2000, issue of THE JOURNAL (2000; for a CD4 cell count less than 500/mm3 or viral load greater than 283:1046-1053), there was an incorrect value in the Table on page 1050 and inthe text on page 1051. “Ginkgo biloba, 40 mg/d” should have read “Ginkgo bi- 10000 to 20000 copies/mL, respectively. Although these obser- 2000 American Medical Association. All rights reserved.
(Reprinted) JAMA, December 27, 2000—Vol 284, No. 24 3129

Source: http://lalashan.mcmaster.ca/theobio/MMEDclinic/images/9/9a/Shah.pdf