ANAESTHETIC mortality risk factors associated with anaesthetic-related death have been studied in canine populations in North American referral practices (Lumb and Jones 1973, 1984, Hosgood and Scholl 1998, Gaynor and others 1999). Mortality risks for canine referral populations undergoing anaesthetic have been documented at 0·5 to 1 per cent. Typically, these risks are higher than those documented in first-opinion practice (Clarke and Hall 1990, Dyson and others 1998). Factors associated with anaesthetic-related death in referred animals include poor patient health status, old age and belonging to a brachycephalic breed (Lumb and Jones 1973, Hosgood and Scholl 1998, Gaynor and others 1999). Only the study by Hosgood and Scholl (1998) adjusted the risks associated with specific factors for confounding variables, and found that increased risk was only associated with increasing age and poor patient health status. Assessment of risk factors in a canine referral practice population has not been published recently in the UK. The aims of this study were to identify major risk factors associated with anaesthetic-related mortality and to estimate the mortality risk in a single UK canine referral practice population. A case-control study was conducted retrospectively from dogs anaesthetised at the Queen Mother Hospital for Animals (QMHA), Royal Veterinary College (RVC), from the beginning of February 1999 until the end of April 2002. The study was nested in a cohort of all dogs that were anaesthetised using an inhalation agent at the RVC during the study period. The cohort was generated from the QMHA’s computer system, based on the inhalation agent administered. Cases were defined as anaesthetic-related deaths occurring within 48 hours of anaesthesia, in which anaesthesia could not be excluded from contributing to death. Cases were identified from case records and postmortem reports. Controls were randomly selected from the cohort of dogs anaesthetised at the RVC during the three-year period using a case:control ratio of 1:2. Potential risk factors assessed included patient health status (ASA [American Society of Anesthesiologists] grade 1 to 2 healthy or mild disease versus ASA grade 3 to 5 severe disease to moribund), age, premedication with acepromazine (ACP; Vericore), premedication with opioids (pethidine [Pethidine injection BP; Martindale Pharmaceuticals], morphine [Morphine sulphate injection BP; Martindale Pharmaceuticals], butorphanol [Torbugesic injection; Fort Dodge] or papaveretum [Papaveretum injection BP; Martindale Pharmaceuticals]), overall premedication (benzodiazepines [Diazemals; Dumex or Hypnovel; Roche] and/or opioids only, acepromazine and/or opioids), induction with thiopentone (Intraval Sodium; Merial Animal Health) or propofol (Rapiovet; Schering-Plough Animal Health), maintenance of anaesthesia with halothane (Halothane Vet; Merial Animal Health) or isoflurane (IsoFlo; Schering-Plough Animal Health) and duration of anaesthesia. Crude odds ratios (ORs) and 95 per cent confidence intervals (CIs) for the odds of anaesthetic-related death were calculated for the exposure variables. Either the chi-squared test or Fisher’s exact test was applied for categorical data and Student’s t test or Mann-Whitney U test was applied for continuous data (Kirkwood 1988). Risk factors significant at the 20 per cent level were taken forward for further evaluation in the multivariable analysis. Following exploratory stratification, logistic regression (STATA 7.0; Statcorp) was used, and the contribution of the explanatory variables to the model was assessed. Statistically significant variables (P<0·05) were retained in a logistic regression model using a forward stepwise approach. Fit of the model was evaluated using the Hosmer-Lemeshow statistic (Hosmer and Lemeshow 1989). The anaesthetic-related mortality risk was estimated from the number of cases divided by the number of anaesthetics undertaken during the year from April 2001 to April 2002. During the three-year period, 6026 dogs were anaesthetised and 20 anaesthetic-related deaths were recorded. The anaesthetic-related death risk in the referral population for the year April 2001 to April 2002 was 0·58 per cent (one in 173 dogs; 95 per cent CI 0·24 to 0·92). The median ages (interquartile range) of cases and controls were 6·8 (3·8 to 10) and 5·9 (1·5 to 9·8) years respectively (P=0·51). The median durations of anaesthesia (interquartile range) for cases and controls were 72·5 (45 to 235) minutes and 80 (50 to 140) minutes, respectively (P=0·14). At the univariable level, patients with a high-risk status (ASA grade 3 to 5) had a significantly increased odds of death compared with patients at low risk (ASA 1 to 2) (Table 1). Patients receiving acepromazine as part of the premedication were at reduced odds of death relating to anaesthesia, compared with not receiving acepromazine and compared with other premedicants. In the latter comparison, two cases that had received no premedication were Variable Category β se β OR 95% CI P