Sir

Animal clinical research paradigms are part of the mechanism for assessing the risks of new drugs to humans. To extrapolate the results to humans requires that animal experiments are designed adequately and conducted properly, with the methods fully reported. But these goals are often not met, either because collection of equivalent clinical parameters is held to be too difficult or time-consuming, or because of an effort to oversimplify complex biological systems in animals that are not identical or useful to the human paradigm1,2,3. This problem needs rectifying.

The sort of information that should be recorded and reported includes means and standard deviations for weight, electrolytes and glucose and the exact time of day of key treatments (including weekends). This would reveal whether such variables were kept constant throughout the experiment, and would show whether failures were due to failure of treatment or of support care.

Another area requiring more detailed reporting is that of pain detection and alleviation, which is not evident in publications of studies supported by the US Public Health Service (PHS). The PHS and the US Department of Agriculture require alleviation of pain in most animals, but this information isn't appearing in the methods sections of publications.

In order to provide information about the dynamic nature of the success or failure of treatment, clinical pathological data should be used, from tests such as complete blood count, clinical chemistries and organ profiles, carried out on living animals at specified times during a disease. Pathological data alone (from tests carried out after death, such as liver biopsies) are not adequate.

Omissions in reporting are prevalent. Take for example recent problems in human gene therapy 4. Primate experiments clearly suggested risk. But in the animal work, the nature of the morbidity in one paper5 was described only in terms of pathological data. The reader is informed that appetite in the survivors was normal, but weight changes were not reported. It is not recorded whether the animals were housed in groups (usually required for primate social needs) or singly, leaving the knowledgeable reader wondering whether a cage-mate ate the food. Additionally, mortality was described only by pathological criteria based on liver biopsies, rather than including information from, for example, liver function tests, complete blood count information, haemodynamic criteria and subjective scores in both survivors and dead animals. This left the events occurring between life and death open to speculation.

In a slightly improved reporting scheme in canine research6, when using pericardial adenoviral-mediated vascular endothelial growth factor (VEGF), investigators reported a severe dose-dependent physiological reaction after administration. Substantial information about morbidity and mortality was included, but key aspects about related events were omitted. For example, no information was given about intense electrolyte changes, nausea and occasional thrombocytopaenia — seen and treated by the veterinary staff — which could have contributed to the later drug and technical failures of VEGF and adenovirus in human patients7.

Scientists and the public need access to correct information about clinical protocols in animals. We need more of the right kind of clinical study design, and more collaborative reporting from inside animal facilities.

Such partnerships between veterinarians and scientists are stuck — they require a change in thinking. Scientists need to stop viewing veterinary care as a burden and see it as a quality assurance for drugs that are destined for human use. As well as allowing a more accurate risk–benefit analysis for human drug trials, introduction of such reporting standards in non-human experiments could also reduce the amount of costly regulation required at the later stages of commercialization of drugs.