Letter by Simonato and Altmayer Regarding Article, “Dog Ownership and Survival: A Systematic Review and Meta-Analysis”

To the Editor:

We have read with interest the article “Dog Ownership and Survival: A Systematic Review and Meta-Analysis” by Kramer et al¹ published in the October issue of Circulation: Cardiovascular Quality and Outcomes. This work has garnered media attention, with over 120 news reports and >48 000 downloads of the article. This is understandable given the popularity of dogs and the certain interest of laypeople in this topic. Nevertheless, we think that it is fraught with methodological weakness, serving as a case example of the danger of unadjusted and biased associations.

The authors report on 10 studies with a 24% risk reduction for all-cause mortality for dog owners as compared to nonownership (relative risk [RR], 0.76 [95% CI, 0.67–0.86]). However, when the selected studies are examined more closely, it is possible to ascertain that this association is possibly spurious. For instance, the authors included a Norwegian study by Torske et al,² which supposedly demonstrates an RR of 0.57 (95% CI, 0.54–0.6). This number is not reported anywhere in the original article, which reports only a hazard ratio (HR) of 0.99 (95% CI, 0.91–1.09) adjusted for age and sex, among other models. The number used by Kramer et al¹ can only be obtained by extracting the unadjusted number of deaths in dog owners (1586/9082) versus those of nonowners (11 112/31 368). However, it is important to note that on the original study dog owners were on average 4.5 years younger than nonowners and it would be reasonable to assume younger people on average die less than their older counterparts.

A similar issue takes place regarding the inclusion of a study by Ding et al.³ The authors of this original study report that dog owners were “younger, of lower educational level and occupational social class, and were more likely to be in a married/partnered/cohabited relationship, live with others, be economically active, reside in rural areas, smoke, report more drinking, walking, and [physical activity].”³ In one of the multivariable models included in the article, the authors have found an increased risk of all-cause of mortality in dog owners. Yet in the final model adjusted for age, sex, marital status, social class, employment status, educational attainment, living situation, alcohol, smoking, and longstanding illness, there was no significant association of all-cause mortality and dog ownership, with a HR of 1.03 (95% CI, 0.98–1.09). Kramer et al¹ used the unadjusted RR of dog owner death (1894/15 177) versus that of nonowners (6275/36 006), yielding the reported RR of 0.75 (95% CI, 0.71–0.78).

Finally, the study of Chowdhury et al⁴ must also be considered. In this study, a total of 4039 patients were selected after responding to a survey. In this article, the percentage of dog owners in the group of patients that have ever owned a pet is unknown. Only the percentage of current dog owners (60%) is reported. Kramer et al¹ included in the meta-analysis the unadjusted RR of all-cause mortality for having been a pet owner ever in one’s life, a number not reported in the original paper: RR 0.77 (95% CI, 0.67–0.89; pet owners: 796/2694; nonowners: 162/387). The original analysis is far more nuanced. Chowdhury et al⁴ include several different models, the most important of them being the adjusted all-cause mortality of having been a pet owner ever with an HR of 0.86 (95% CI, 0.73–1.02, P=0.09). No association was found with all-cause mortality and being a current pet owner in adjusted models. Still, the authors did find a protective relationship of being a current pet owner and cardiovascular mortality in a model adjusted for sex, age, education, marital status, pretreatment blood pressure, serum total cholesterol, HDL (high-density lipoprotein), diabetes mellitus, smoking status, body mass index, glomerular filtration rate, physical activity, randomized treatment group, and on-treatment blood pressure (HR, 0.74 [95% CI, 0.57–0.96]; P=0.02). However, there was no significant protective relationship in cardiovascular death for currently owning dogs (HR, 0.84 [95% CI, 0.60–1.19]; P=0.34), a trend for cats (HR, 0.63 [95% CI, 0.40–1.01]; P=0.05), and a significant relationship for non-cat, non-dog pets (HR, 0.62 [95% CI, 0.39–0.98]; P=0.04). The weakness in using RR when the more appropriate measure is the HR given the time component of the analyses should be noted.

We can understand and appreciate the difficulties involved in performing such a meta-analysis, and we congratulate the authors on attempting to bring some clarity to this important scientific question. Nevertheless, we do not think that these difficulties justify unadjusted comparisons. Dog ownership most likely is at best a marker of increased potential for physical activity, in the same way that having a gym membership or owning a treadmill would be and, at worst, a marker of younger age, which would be the key confounder here. We invite the authors to repeat their analysis using adjusted measures. We think that an adjustment for age would be sufficient to alter the findings.

Footnotes