Manningclark.org.au

Smoking was an exclusion criterion for controls, whereas Figure 1. Methodological Quality of Animal Trials (n=76) 4 of the 21 cases were regular smokers of 2 to 10 cigarettesper day. Mean urinary excretion rates of 8-iso-PGF2␣ were similar in the 4 smokers (404 pg/mg of creatinine) and in the 21 cases considered as a whole (482 pg/mg of creati- nine). Urine albumin excretion rates were not tested. There was only a small glucose variability between each day (day 1 mean amplitude of glycemic excursions [MAGE], 74 mg/ dL; day 2 MAGE, 76 mg/dL), and MAGE values on day 1 and day 2 were highly correlated (r=0.87; PϽ.001).
Finally, conflicting observations in the study by O’Byrne et al4 could have resulted from the use of different methods in different groups of patients at different ages: enzyme im- munoassay in our study (21 patients with type 2 diabetes; mean age of 64 years) vs stable isotope dilution mass spec- Percentage of Trials Satisfying Each Quality Criterion trometry assay in O’Byrne et al (13 patients with type 1 dia-
betes; mean age of 36 years).
Louis Monnier, MD
how often highly cited animal studies translate into suc- [email protected]
Department of Metabolic Diseases
Methods. The 7 leading scientific journals by citation im-
Lapeyronie Hospital
pact factor (Journal Citation Reports, Thomson Scientific, Emilie Mas, PhD
Philadelphia, Pa, 2004) that regularly publish original ani- University Institute of Clinical Research
mal studies were searched: Science, Nature, Cell, Nature Medi- University of Montpellier
cine, Nature Genetics, Nature Immunology, and Nature Bio- Christine Ginet, MD
technology. Articles with more than 500 citations were Department of Metabolic Diseases
retrieved under the assumption that such prominent find- Lapeyronie Hospital
ings would more likely be tested in subsequent human trials.4 Franc¸oise Michel, MD
University Institute of Clinical Research
A total of 2000 articles published between 1980 and 2000 University of Montpellier
were screened, reflecting advances in molecular biology and Laetitia Villon, MD
recombinant genetics. Articles were included if they inves- Department of Metabolic Diseases
tigated a preventive or therapeutic intervention in an in vivo Lapeyronie Hospital
animal model. When there were multiple animal studies of Jean-Paul Cristol, MD
the same intervention, the most cited study was retained.
Claude Colette, PhD
Power calculations (␣=0.05, ␤=0.05) estimated that 49 ar- University Institute of Clinical Research
ticles were needed to exclude a translation rate below 5%.
University of Montpellier
For each included study, a literature search identified hu- Montpellier, France
man studies that translated the animal evidence. Success- Financial Disclosures: None reported.
ful translation was defined as replication in a randomized 1. Buckingham B, Block J, Wilson DM. Continuous glucose monitoring. Curr Opin
trial yielding results that were statistically positive accord- Endocrinol Diabetes. 2005;12:273-279.
ing to primary outcome. Interventions and diseases analo- 2. Roberts LJ, Morrows JD. Measurement of F(2)-isoprostanes as an index of oxi-
dative stress in vivo. Free Radic Biol Med. 2000;28:505-513.
gous to those studied in the animal study were allowed.
3. Brownlee M, Hirsch IB. Glycemic variability: a hemoglobin A1c-independent risk
MEDLINE, EMBASE, the Cochrane Central Register of factor for diabetic complications. JAMA. 2006;295:1707-1708.
4. O’Byrne S, Forte P, Roberts J, et al. Nitric oxide synthesis and isoprostane pro-
Controlled Trials, the Cochrane Database of Systematic Re- duction in subjects with type 1 diabetes and normal urinary albumin excretion.
views, the National Institutes of Health Clinical Trials Da- tabase, BIOSIS Previews, and the International Pharmaceu-tical Abstracts Database were searched from their inception RESEARCH LETTER
through May 2006. Bibliographies of topic-specific reviewarticles were manually searched for additional studies and experts were contacted if the search was negative.
The quality of the studies was assessed based on adapted standards for the conduct of animal research (FIGURE 1).5
To the Editor: Most medical therapies in use today were ini-
Good quality was defined as a global methodology score of tially developed and tested in animals,1 yet animal experi- 50% or higher. Multivariable logistic regression was used ments often fail to replicate when tested in rigorous hu- to assess predictors of translation. The Pearson correlation man trials.2,3 We conducted a systematic review to determine test was used to determine if methodological quality of ani- 2006 American Medical Association. All rights reserved.
(Reprinted) JAMA, October 11, 2006—Vol 296, No. 14 1731 Comment. Only about a third of highly cited animal re-
Figure 2. Search Flow and Article Retrieval search translated at the level of human randomized trials. Thisrate of translation is lower than the recently estimated 44% 2000 Potentially Relevant Citations Identified replication rate for highly cited human studies.4 Limitationsof this review include a focus on highly cited animal studies published in leading journals, which by their positive and highly of Title, Keywords, and Abstract918 No Disease visible nature may have been more likely to translate than less frequently cited research. In addition, this study had limited power to discern individual predictors of translation.
Nevertheless, we believe these findings have important implications. First, patients and physicians should remain cautious about extrapolating the findings of prominent ani- mal research to the care of human disease. Second, majoropportunities for improving study design and methodologi- 12 Studies Excluded (Identical Intervention cal quality are available for preclinical research. Finally, poorreplication of even high-quality animal studies should be expected by those who conduct clinical research.
Daniel G. Hackam, MD
[email protected]
mal studies improved over time. Significance level was set Donald A. Redelmeier, MD, MSHSR
at 2-sided PϽ.05. Analyses were conducted using SAS ver- Department of Medicine
sion 9.0 (SAS Institute Inc, Cary, NC).
University of Toronto
Results. Seventy-six animal studies fulfilling inclusion
Toronto, Ontario
criteria were identified (F
Author Contributions: Dr Hackam had full access to all the data in the study and
IGURE 2; details of studies avail-
takes responsibility for the integrity of the data and the accuracy of the data analysis.
able in online eTable, available at http://www.jama.com).
Study concept and design: Hackam, Redelmeier.
No animal study was negative. The median citation count Analysis and interpretation of data: Hackam, Redelmeier.
was 889 (range, 639-2233). The median publication year Drafting of the manuscript: Hackam, Redelmeier.
was 1992, yielding a median of 14 years for potential trans- Critical revision of the manuscript for important intellectual content: Hackam, lation. Of the animal studies, 37 (49%) were rated as hav- Statistical analysis: Hackam, Redelmeier.
ing good methodological quality. Most studies included Obtained funding: Redelmeier.
dose-response gradients, clinically relevant outcomes, and Administrative, technical, or material support: Redelmeier.
Study supervision: Redelmeier.
long-term end points (Figure 1). Few studies included ran- Literature retrieval: Hackam.
dom allocation of animals, adjustment for multiple hypoth- Financial Disclosures: None reported.
Funding/Support: Dr Hackam was supported by a Canadian Institutes of Health
esis testing, or blinded assessment of outcomes. Method- Research Fellowship Award, the Chisholm Memorial Fellowship, and the Clinician- ological quality did not improve during the study interval Scientist Training Program of the University of Toronto. Dr Redelmeier was sup- ported by the Canada Research Chair in Medical Decision Sciences, the Error Man- agement Unit of Sunnybrook Health Sciences Centre, the National Institutes of Of the animal studies, 28 (37%; 95% confidence interval Health Resuscitation Outcomes Consortium, and the Canadian Institutes of Health [CI], 26%-48%) were replicated in human randomized trials, Role of the Sponsors: The funding sources had no role in the design and conduct
14 (18%) were contradicted by randomized trials, and 34 of the study; in the col ection, management, analysis, and interpretation of the (45%) remain untested. Median time to replication was 7 data; or in the preparation, review, or approval of the manuscript.
Additional Information: The eTable is available at http://www.jama.com.
years (range, 1-15 years). Global methodology score did not Acknowledgment: We are grateful to Mark Crowther, MD, MSc, Gideon Koren,
predict translation in unadjusted analyses (odds ratio [OR], MD, Philippe Poussier, MD, Joel Ray, MD, MSc, William Sibbald, MD, MBA, and Matthew Stanbrook, MD, PhD, for comments on a previous draft of the manu- 1.28 per 10% higher score; 95% CI, 0.97-1.69) or in analy- script. No individual received compensation for their assistance.
ses adjusted for citation rate and length of time available for 1. Guidance for Industry: Nonclinical Safety Evaluation of Drugs or Biologic
human replication (OR, 1.27; 95% CI, 0.96-1.69). Animal Combinations. Rockville, Md: US Dept of Health and Human Services, Food and studies incorporating dose-response gradients were more Drug Administration, Center for Drug Evaluation and Research; 2006.
2. Kaste M. Use of animal models has not contributed to development of acute
likely to translate to humans (OR, 3.3; 95% CI, 1.1-10.1).
stroke therapies: pro. Stroke. 2005;36:2323-2324.
Other quality criteria, type of therapy, type of disease, spe- 3. Pound P, Ebrahim S, Sandercock P, Bracken MB, Roberts I. Where is the evi-
cies, journal, citation rate, length of follow-up, and year of dence that animal research benefits humans? BMJ. 2004;328:514-517.
4. Ioannidis JPA. Contradicted and initial y stronger effects in highly cited clinical
publication did not predict subsequent translation. Eight rep- research. JAMA. 2005;294:218-228.
licated interventions were subsequently approved for use 5. Stroke Therapy Academic Industry Roundtable. Recommendations for stan-
dards regarding preclinical neuroprotective and restorative drug development. Stroke.
1732 JAMA, October 11, 2006—Vol 296, No. 14 (Reprinted) 2006 American Medical Association. All rights reserved.
Neuronal nitric oxide synthase disruption E1 JAMA, October 11, 2006—Vol 296, No. 14 (Reprinted) 2006 American Medical Association. All rights reserved.
Glial cell line–derived neurotrophic factor Abbreviations: Ab, antibody; CFTR, cystic fibrosis transmembrane conductance regulator; EAE, experimental autoimmune encephalomyelitis; hcr, human-complement receptor; ICAM-1, intercellular adhesion molecule-1; IgG, immunoglobulin G; IL, interleukin; iNOS, inducible nitric oxide synthase; LFA-1, leukocyte function-associated antigen-1; NBQX, 2,3-Dihydroxy-6-nitro-7-sulfamoyl-benzo(F)quinoxaline; NMDA, N-Methyl-D-Aspartate; SCID, severe combined immunodeficiency; TGF, transforming growth factor; TNF, tumor necrosis factor; TRAIL, tumor necrosis factor related apoptosis-inducing ligand; VEGF, vascular endothelial growth factor.
2006 American Medical Association. All rights reserved.
(Reprinted) JAMA, October 11, 2006—Vol 296, No. 14 E2

Source: http://manningclark.org.au/sites/default/files/animal_human_research_evidence.pdf