Recruitment to the study began in July 2006 and finished in December 2010. Follow-up was continued for a further 10 months after recruitment had been completed. An interim analysis was undertaken in February 2010 at which time 60 dogs had been recruited to the study, of which 27 had reached the primary endpoint. The criterion for unblinding was not met.
Approximately, 1,000 dogs were screened to determine whether they met the entry criteria of the study. Seventy-six dogs were recruited to the study. The outcome after recruitment for these 76 dogs is summarized in . The median time in the study for all recruited dogs was 427 days. Forty-four dogs reached the primary endpoint, resulting in an overall event rate for the study of 58%. The median time in study for dogs reaching the primary endpoint was 373.5 days. Ten dogs died or were withdrawn from the study for other reasons. The median time in study for these dogs was 445 days. Twenty-two dogs were still alive and had not reached the primary endpoint at the time of closure of the study. The median time in study for these dogs was 529 days. No dogs were lost to follow-up.
A flow chart indicating the outcome for the 76 dogs enrolled in the study.
Thirty-nine dogs were randomized to receive pimobendan and 37 were randomized to receive placebo. In the group administered pimobendan, the median dose received was 0.453 mg/kg/day (interquartile range (IQR), 0.310–0.503 mg/kg/day)c.
Analyses for homogeneity of groups showed no significant difference between groups for any of the characteristics measured at baseline ().
Selected baseline characteristics in the 2 treatment groups expressed as frequencies and (proportions%) or medians and [interquartile range]
In the placebo group, 25 dogs reached the primary endpoint, 4 dogs died or were withdrawn for other reasons, and 8 dogs were alive at the end of the study. In the pimobendan group, 19 dogs reached the primary endpoint, 6 dogs died or were withdrawn for other reasons, and 14 dogs were alive at the end of the study (). The proportions of dogs reaching the primary endpoint were not significantly different between groups (P = .1).
Reasons for Censoring of 32 Dogs by Treatment
The estimated median time to the primary endpoint was significantly longer for dogs receiving pimobendan (718 days, IQR 441–1152 days) than for dogs receiving placebo (441 days, IQR 151–641 days) (log-rank P = .0088) (). The median time to reach the primary endpoint for dogs in the pimobendan group was 63% (9 months) longer than for those in the placebo group. When a subanalysis was performed including only euthyroid dogs, the results were similar. The estimated median time to the primary endpoint was significantly longer for dogs receiving pimobendan (693 days, IQR 441–1,152 days) than for dogs receiving placebo (441 days, IQR 146–641 days) (log-rank P = .018). Too few hypothyroid dogs (n = 5) were included in the study to allow their analysis as a separate subgroup.
Kaplan Meier survival curves plotting the estimated percentage of dogs in each group that have not yet met the primary endpoint, against time. IQR, interquartile range
Of the 19 dogs reaching the primary endpoint in the pimobendan group, 12 (63%) developed CHF and 7 (37%) died suddenly. Of the 25 dogs in the placebo group that reached the primary endpoint, 17 (68%) developed CHF and 8 (32%) died suddenly. The proportion of dogs reaching the different components of the primary endpoint in the different groups was not significantly different (P = .76). Comparing only those dogs that reached each subendpoint, the median time to each subendpoint was not significantly different between groups ().
Table 4 The number of dogs in each treatment reaching the component parts of the primary endpoint and the median [IQR] time taken to do so. CHF, congestive heart failure. The comparison between groups compares the time to subendpoint only for those dogs that (more ...)
Kaplan Meier plot illustrating the estimated proportion of dogs in each group remaining alive (all-cause mortality) is illustrated in . The estimated median survival time was significantly longer for dogs receiving pimobendan (623 days, IQR 491–1,531 days) compared with dogs receiving placebo (466 days, IQR 236–710 days) (log-rank P = .034).
Kaplan Meier survival curves for the all-cause mortality analysis, plotting the estimated percentage of surviving dogs in each group, against time. MST, median survival time; IQR, interquartile range.
Univariable analysis showed that there was an increase in the hazard of reaching the primary endpoint in dogs with a greater LVIDSN (hazard ratio (HR) = 1.836 (95% confidence intervals (CI) = 1.457, 2.315) for a 0.1 unit increase in LVIDSN) and a greater LVIDDN (HR = 1.482 (95% CI = 1.213, 1.810) for a 0.1 unit increase in LVIDDN) (). There was also a significant increase in the hazard of reaching the primary endpoint in dogs with higher heart rates on physical examination (HR = 1.17 per 10 bpm increase in heart rate; 95% CI = 1.01, 1.35) and evidence of arrhythmias (the presence of any arrhythmia [if an arrhythmia present; HR = 2.88; 95% CI = 1.53, 5.41] on the resting ECG; greater number of VPCs on the resting ECG [if ≥4 VPCs; HR = 2.36; 95% CI = 1.12, 4.97]; and greater numbers of VPCs on the Holter recording [HR = 1.146 for each 1000 VPCs on the Holter ECG; 95% CI = 1.041, 1.261] were all associated with a worse outcome) ().
Fig. 4 A forest plot showing the hazard ratio and 95% confidence intervals associated with variables considered in the univariable analyses with time to the primary endpoint (congestive heart failure or sudden death) as the dependent variable. Circles represent (more ...)
The explanatory multivariable model created after the backward stepwise process, but before consideration of interaction terms, contained 4 variables. These were treatment group, LVIDSN, baseline heart rate, and whether or not a dog had ≥4VPCs. The effect of treatment remained significant when all other arrhythmia variables were substituted for ≥4 VPCs in the final model and when normalized diastolic diameter was used in the model instead of normalized systolic diameter.
Six interaction terms were created examining the interactions between each of the possible pairs of variables in the model. There was no significant interaction between treatment and any of the other variables. The only significant interaction was between heart rate and whether or not a dog had ≥4VPCs. Dogs were therefore categorized into 4 groups according to whether or not they had a heart rate above or below the median (110 bpm) at baseline, and whether or not they had ≥4 VPCs. Group 1 consisted of 32 dogs with a heart rate <110 bpm and <4 VPCs; this group was used as the referent category in the multivariable analysis. Group 2 consisted of 4 dogs with a heart rate <110 bpm and ≥4 VPCs. Group 3 consisted of 31 dogs with a heart rate ≥110 bpm and <4 VPCs. Group 4 consisted of 9 dogs with a heart rate ≥110 bpm and ≥4 VPCs.
The significant effect of treatment in prolonging the time to the primary endpoint persisted after adjustment for the influence of other variables (). The hazard ratio from the multivariable analysis for the effect of treatment implies that dogs in the placebo group were 3.3 times more likely than those in the pimobendan group to reach the primary endpoint first. The hazard of reaching the primary endpoint approximately doubled for every 0.1 unit increase in normalized systolic diameter. Dogs with ≥4 VPCs and an above median heart rate were approximately 7 times more likely to reach the primary endpoint first.
Explanatory multivariable model. HR, hazard ratio; CI, confidence intervals
Paired Holter data were available for 70 dogs: 37 in the pimobendan group, and 33 in the placebo group. All repeat Holters were recorded between 20 and 56 days after initiation of treatment. The median time to the 2nd Holter recording was 30 days (IQR 27–38.5 days) for the pimobendan group and 33 days for the placebo group (IQR 29–38 days). The time to 2nd recording was not significantly different for the groups (P = .33). Reasons for missing data were as follows: 1 pimobendan dog had sudden cardiac death; the other missing pimobendan dog had absent data because of technical difficulties at baseline; 1 placebo dog had sudden cardiac death; 1 placebo dog had developed CHF; 1 placebo dog did not have a 2nd Holter ECG until over 1 year after initiating treatment; and 1 placebo dog had absent data because of technical difficulties. In neither group was there a significant change in VPCs after treatment [comparison of log10 (total VPC + 1) pimobendan group P = .93, placebo group P = .33] (). Three of the pimobendan dogs were considered to have complex ventricular arrhythmias before treatment and 5 to have complex ventricular arrhythmias afterward (the original 3 dogs and 2 additional dogs). Six of the placebo dogs were considered to have complex ventricular arrhythmias before treatment and 6 to have complex ventricular arrhythmias afterward. (The 6 dogs with complex arrhythmias at the 2nd examination consisted of 4 of the original 6 dogs and 2 dogs not considered to have complex arrhythmias on the 1st examination. Of the 2 dogs considered to have complex arrhythmias at the original examination, but not at the 2nd examination, 1 dog had died and the other was considered not to have a complex arrhythmia on the 2nd Holter examination.) None of the dogs received any antiarrhythmic treatment between Holter examinations.
Fig. 5 The number of VPCs on the 24-hour ECG before and approximately 1 month after the initiation of treatment in the group receiving pimobendan (A) and the group receiving placebo (B). Differences were not significant in either group. VPC, ventricular premature (more ...)
Paired echocardiographic data were available for 73 dogs: 38 in the pimobendan group, and 35 in the placebo group. Reasons for failure to obtain a 2nd echocardiographic examination were, in all cases, that the dogs had already reached the primary endpoint. All repeat echocardiographic examinations were conducted between 20 and 56 days after the initiation of treatment. The median time to 2nd examination for the group receiving pimobendan was 30 days (IQR 27–37 days) and for the group receiving placebo was 34 days (IQR 30–38.5 days); the difference in time to the 2nd recording for the 2 groups was not significant (P = .18). The changes in LVIDS and LVIDD from the 1st echocardiographic examination to the 2nd examination are illustrated for both groups in . The median change in LVIDS in the pimobendan group was −4 mm [−5.0 to −2.48 mm]. The median change in LVIDS in the placebo group was 0 mm [−2.0 to 2.7 mm] (P < .001 for the comparison between groups). The median change in LVIDD in the pimobendan group was −3.1 mm [−5.0 to −0.07 mm]. The median change in LVIDD in the placebo group was 0.9 mm [−1.6 to 3.7 mm] (P < .001 for the comparison between groups).
Fig. 6 The change in left ventricular diameter occurring within the first 2 months of the study for dogs in which paired observations were available (73 dogs, 38 receiving pimobendan, and 35 receiving placebo). The change in systolic diameter is illustrated (more ...)
Sotalol treatment was administered to 8 dogs in total, one of which also received mexiletine. Sotalol treatment was introduced to 2 dogs in the placebo group 130 days and 146 days after randomization. Sotalol treatment was introduced to 6 dogs in the pimobendan group, a median of 242.5 days (range 53–935 days) after randomization.
Results of the exploratory multivariable analysis examining the influence of LVIDSN, change in LVIDS, and treatment on the time to the primary endpoint are illustrated in .
Exploratory multivariable model. LVIDSN, normalized left ventricular internal diameter in systole at baseline; HR, hazard ratio; CI, confidence intervals; ΔS, change in left ventricular diameter
None of the suspected adverse reactions or potential complications resulted in the withdrawal of a dog from the study ().
Suspected adverse drug reactions (not leading to withdrawal in 76 dogs)