Does Dog Ownership Really Prolong Survival?
Many households own a dog, and dog owners are more likely to walk and to meet physical activity guidelines, compared with non-dog owners.1,2 Other benefits, usually reported in cross-sectional studies, include improved mental wellbeing and reduced cardiovascular risk factors.3 The evidence on dog ownership to date was summarized by the American Heart Association (2013) as “probably having some causal role…. in reducing cardiovascular risk.”4
In October 2019, Kramer et al5 published a meta-analysis examining dog ownership and survival using 9 prospective epidemiological studies. This meta-analysis reported a 24% decreased risk of all-cause mortality among dog owners compared with non-dog owners. The protective effect was even stronger for the 3 studies that specifically looked at the risk of cardiovascular events among dog owners.5 The article was supported by an Editorial that outlined potential prevention mechanisms of dog ownership mediated through increased physical activity and effects on stress and blood pressure reduction.3
The authors calculated the ratio of deaths to the population at risk in those exposed and unexposed to dog ownership. The study reported they could only conduct an analysis of pooled unadjusted rate ratios5 (second last paragraph, p7). The Cochrane Collaboration recommends that unadjusted and adjusted estimates both be reported in meta-analyses, as the latter adjusts for important known confounders, and may produce different (risk) estimates, compared with unadjusted meta-analyses.6,7 We initially focused attention on the 6 population studies with estimates of all-cause mortality risk in the Kramer article.5 We calculated adjusted hazard ratios from these articles and re-did this meta-analysis to see if the evidence on dog ownership and mortality remained consistent. We extracted estimates from the articles that adjusted for the maximum number of covariates available, as recommended by the Cochran Collaboration6 (Table 1). We chose the identical random-effects meta-analysis methods5 namely the DerSimonian-Laird Method and the Cochran Q test and I2 values to assess heterogeneity between studies and used the Metagen package in R (R Foundation for Statistical Computing, Vienna, Austria). Where possible, the hazard ratios were extracted rather than the risk ratio, as the hazard ratio accounts for not only the occurrence of an event but also the timing of the event.
| Source, sample size | Adjusts for |
|---|---|
| Population-based studies | |
| Gillum et al,8 N=11 394 | Age, sex, race, SES, health status, activity level, healthy behaviors, and other risk factors |
| Chowdhury et al,9 N=4039 | Age, sex, marital status, education, blood pressure, cholesterol, serum HDL, history of diabetes, smoking, BMI, eGFR, physical activity, treatment group, and on-treatment blood pressure |
| Mubanga et al,10 N=3 432 153 (+34 202 twins) | Sex, marital status, number of children at home, population density, area of residence, region of birth, income, and latitude |
| Torske et al11, N=25 031 | Age and sex |
| Ding et al,12 N=59 352 | Age, sex, marital status, social class, employment, education, living circumstances, alcohol, smoking, and illness |
| Sørensen et al,13 N=275 184 | Age, sex, education, income, and marital status (through matching) |
| Individuals with prior cardiovascular disease | |
| Friedmann et al,14 N=96 | None |
| Friedmann et al,15 N=424 | None |
| Friedmann et al,16 N=460 | None. We included the unadjusted hazard ratio from this article; this differs from the calculated relative risk included in Kramer 2019 |
We present our adjusted meta-analysis for all-cause mortality (Table 2, uppermost two rows, and Figure [A]) and reproduce the original analysis (Figure [B]).5 Compared with the original analysis (unadjusted relative risk 0.76 [95% CI, 0.67–0.86]), we found a different picture using adjusted estimates (Figure [A], 4 of the 5 adjusted hazard ratios8,9,11–13 showing a nonsignificant effect, and the only significant effect coming from Mubanga10). Our adjusted pooled estimate from the 6 population-based studies was nonsignificant, effect size17 of 0.95 (0.85–1.05). In our re-analysis, the 3 studies by Friedmann14–16 in people with existing cardiovascular disease (CVD) show that dog ownership remains significantly associated with survival (relative risk [RR] 0.39 [95% CI, 0.20–0.77]), but we note that no adjusted estimates were available. In contrast to the original meta-analysis which used the unadjusted RR (0.49), we used the hazard ratio (0.60). Overall, the adjusted RR for the association between dog ownership and survival based on all of these 9 articles combined was not significant (Figure [A], RR=0.93 [0.83–1.03]).
| Model type | Adjustment | Studies included | Mubanga weight | Pooled effect |
|---|---|---|---|---|
| Random effects (Figure [B]) | Unadjusted (Kramer) | All | 17% | 0.76 (0.67–0.86) |
| Random effects (Figure [A]) | Adjusted | All | 19% | 0.93 (0.83–1.03) |
| Fixed effects | Unadjusted | All | 82% | 0.72 (0.71–0.73) |
| Fixed effects | Adjusted | All | 63% | 0.86 (0.84–0.87) |
| Fixed effects | Adjusted | All except Mubanga (2017) | 0% | 0.96 (0.93–0.98) |
| Random effects | Adjusted | All except Mubanga (2017) | 0% | 0.97 (0.90–1.04) |
| Fixed effects | Adjusted | All and an additional new 8 hypothetical smaller studies | 46% | 0.88 (0.87–0.89) |
| Random effects | Adjusted | All and an additional new 8 hypothetical smaller studies | 11% | 0.94 (0.88–1.01) |
Figure. A, Updated meta-analysis of the adjusted associations between dog ownership and the risk of all-cause mortality (showing adjusted ES [effect size]14). B, Original meta-analysis5 of the association between dog ownership and the risk of all-cause mortality. Reprinted from Kramer et al5 with permission. Copyright ©2019, Wolters Kluwer Health, Inc.
Further issues relate to the choice of fixed or random-effects meta-analysis.7 Random-effects models assume underlying true effect sizes vary across cohorts due to participants from different populations with different levels of potential confounders, such as physical activity levels or health status. For random effect models, studies of different sample sizes tend to have more similar weights. While in fixed-effect models, studies were weighted in proportion to their sample sizes (see Table 1 for cohort sample sizes). To address this, we conducted 6 additional meta-analyses on these data (Table 2). Pooled estimates in the fixed-effects models were statistically significant but substantially influenced by the one very large Swedish study (which contributed 92% of all participants across all population studies used here10), although the adjusted estimated attenuated the effect towards the null. Excluding this study showed further attenuation, which was still marginally significant only in the fixed-effects model (RR=0.96). To demonstrate the effect of the large single Swedish study,10 we hypothetically modeled if the results would change if in future, there were an additional 8 smaller new epidemiological studies, and the effects would persist as significant only in the fixed-effects model (RR=0.88).
In summary, our initial conclusion was different to the significant 24% risk reduction reported in the original meta-analysis.5 Our adjusted meta-analysis found a statistically nonsignificant 7% risk reduction in the association between dog ownership and all-cause mortality. There is still a protective association among those with preexisting CVD, but this is limited to 3 small serial studies by the same author with unadjusted estimates.14–16 Overall, for all 9 studies combined, the adjusted association remains nonsignificant. One major debate is around the choice of models and, given the undue weighting to the single Swedish study in fixed-effects models, these associations remained protective; removing the Swedish study, or using random-effects models attenuated or removed this association.
A more recent examination of pet ownership and CVD outcomes18 showed a nonsignificant RRadj of 0.99 (0.91–1.08), and for all CVD, RRadj was 0.95 (0.84–1.07). Subgroup analyses did tend to suggest lowered CVD risk estimates among pet owners, but risks for myocardial infarction and stroke did not differ by pet ownership.18 For the 3 small and possibly selected studies on people with CVD14–16 the association remains significant although attenuated slightly by our revised hazard ratio estimate. The recent analysis,18 in combination with the original study findings5 suggest there still may be some cardiovascular benefit associated with dog ownership, but the data do not support an overall benefit.
The original conclusion of the Kramer article provided positive evidence for dog ownership and achieved the second highest Altmetric research impact score ever for this journal (>2071; Altmetric.com, April 2020). However, including unadjusted estimates may overestimate risk reduction benefits. It is important to adjust for confounders, as shown in the effects of dog ownership on health, as adjusted estimates attenuate or remove significant associations in these studies, resulting in a slightly more nuanced conclusion. Other methodological considerations are the limitations of pooling hazard ratios and relative risks together19 and the issue that the covariates adjusted for were not identical across studies. These are methodological concerns for many meta-analyses and do not substantively affect the findings of this revised meta-analysis.
It is likely that our nonsignificant finding may be closer to the true pooled estimate. However, we cannot be certain that our findings reflect a true absence of effects of dog ownership on health or whether they are due to methodological limitations in these studies (eg, lack information about dog characteristics such as breed, age, caretaking/interactions with owners; influences of very large single studies; single measurement of dog ownership [exposure] with no consideration of ownership timeline, and serial dog walking behavior measures10). Further debate around the models used suggest that random effects are generally used, as they reduce the effects of undue weighting given to individual studies in fixed-effects models.7 Although positive effects of dog ownership are a hoped-for conclusion, especially among dog owners, the original results should be treated with caution. Considering that large randomized controlled trials on dog ownership and long-term health outcomes/survival are difficult to conduct,20 further well-designed prospective cohort studies collecting comprehensive information are needed to better characterize the epidemiological evidence that dogs influence longevity, overall and cardiovascular health and wellbeing.
Acknowledgments
All authors contributed to the conceptualization, design, and interpretation of the article; Dr Bauman wrote the draft, all commented and redrafted parts of the article. Dr Owen performed the revised meta-analyses.
Disclosures
Drs Bauman, Stamatakis, Torske, Owen, and Krokstad declare that they have emotionally vested interests in the topic, as between them, they are the devoted owners of 5 dogs, and Dr Torske is a veterinarian. The other author reports no conflicts.
Footnotes
References
- 1.
Christian HE, Westgarth C, Bauman A, Richards EA, Rhodes RE, Evenson KR, Mayer JA, Thorpe RJ . Dog ownership and physical activity: a review of the evidence.J Phys Act Health. 2013; 10:750–759. doi: 10.1123/jpah.10.5.750CrossrefMedlineGoogle Scholar - 2.
Westgarth C, Christley RM, Jewell C, German AJ, Boddy LM, Christian HE . Dog owners are more likely to meet physical activity guidelines than people without a dog: an investigation of the association between dog ownership and physical activity levels in a UK community.Sci Rep. 2019; 9:5704. doi: 10.1038/s41598-019-41254-6CrossrefMedlineGoogle Scholar - 3.
Kazi DS . Who is rescuing whom?: dog ownership and cardiovascular health.Circ Cardiovasc Qual Outcomes. 2019; 12:e005887. doi: 10.1161/CIRCOUTCOMES.119.005887LinkGoogle Scholar - 4.
Levine GN, Allen K, Braun LT, Christian HE, Friedmann E, Taubert KA, Thomas SA, Wells DL, Lange RA ; American Heart Association Council on Clinical Cardiology; Council on Cardiovascular and Stroke Nursing. Pet ownership and cardiovascular risk: a scientific statement from the American Heart Association.Circulation. 2013; 127:2353–2363. doi: 10.1161/CIR.0b013e31829201e1LinkGoogle Scholar - 5.
Kramer CK, Mehmood S, Suen RS . Dog ownership and survival: a systematic review and meta-analysis.Circ Cardiovasc Qual Outcomes. 2019; 12:e005554. doi: 10.1161/CIRCOUTCOMES.119.005554LinkGoogle Scholar - 6.
Higgins JPT, Green S ,edsCochrane Handbook for Systematic Reviews of Interventions Version 5.1.0, The Cochrane Collaboration. Updated 2011. www.handbook.cochrane.org. Combining studies: https://handbook-5-1.cochrane.org/chapter_13/13_6_2_2_combining_studies.htm.Google Scholar - 7.
Owen KB, Torske M, Bauman A . Letter by Owen et al Regarding article, “dog ownership and survival: a systematic review and meta-analysis”.Circ Cardiovasc Qual Outcomes. 2020; 13:e006906LinkGoogle Scholar - 8.
Gillum RF, Obisesan TO . Living with companion animals, physical activity and mortality in a U.S. national cohort.Int J Environ Res Public Health. 2010; 7:2452–2459. doi: 10.3390/ijerph7062452CrossrefMedlineGoogle Scholar - 9.
Chowdhury EK, Nelson MR, Jennings GL, Wing LM, Reid CM ; ANBP2 Management Committee. Pet ownership and survival in the elderly hypertensive population.J Hypertens. 2017; 35:769–775. doi: 10.1097/HJH.0000000000001214CrossrefMedlineGoogle Scholar - 10.
Mubanga M, Byberg L, Nowak C, Egenvall A, Magnusson PK, Ingelsson E, Fall T . Dog ownership and the risk of cardiovascular disease and death – a nationwide cohort study.Sci Rep. 2017; 7:15821. doi: 10.1038/s41598-017-16118-6CrossrefMedlineGoogle Scholar - 11.
Torske MO, Krokstad S, Stamatakis E, Bauman A . Dog ownership and all-cause mortality in a population cohort in Norway: the HUNT study.PLoS One. 2017; 12:e0179832. doi: 10.1371/journal.pone.0179832CrossrefMedlineGoogle Scholar - 12.
Ding D, Bauman AE, Sherrington C, McGreevy PD, Edwards KM, Stamatakis E . Dog ownership and mortality in England: a pooled analysis of six population-based cohorts.Am J Prev Med. 2018; 54:289–293. doi: 10.1016/j.amepre.2017.09.012CrossrefMedlineGoogle Scholar - 13.
Sørensen IK, Bidstrup PE, Rod NH, Rühling T, Johansen C . Is dog ownership associated with mortality? A nationwide registry study.Eur J Public Health. 2018; 28:1169–1171. doi: 10.1093/eurpub/cky164CrossrefMedlineGoogle Scholar - 14.
Friedmann E, Katcher AH, Lynch JJ, Thomas SA . Animal companions and one-year survival of patients after discharge from a coronary care unit.Public Health Rep. 1980; 95:307–312MedlineGoogle Scholar - 15.
Friedmann E, Thomas SA . Pet ownership, social support, and one-year survival after acute myocardial infarction in the Cardiac Arrhythmia Suppression Trial (CAST).Am J Cardiol. 1995; 76:1213–1217. doi: 10.1016/s0002-9149(99)80343-9CrossrefMedlineGoogle Scholar - 16.
Friedmann E, Thomas SA, Son H . Pets, depression and long term survival in community living patients following myocardial infarction.Anthrozoos. 2011; 24:273–285. doi: 10.2752/175303711X13045914865268CrossrefMedlineGoogle Scholar - 17.
Farvid MS, Ding M, Pan A, Sun Q, Chiuve SE, Steffen LM, Willett WC, Hu FB . Dietary linoleic acid and risk of coronary heart disease: a systematic review and meta-analysis of prospective cohort studies.Circulation. 2014; 130:1568–1578. doi: 10.1161/CIRCULATIONAHA.114.010236LinkGoogle Scholar - 18.
Yeh TL, Lei WT, Liu SJ, Chien KL . A modest protective association between pet ownership and cardiovascular diseases: a systematic review and meta-analysis.PLoS One. 2019; 14:e0216231. doi: 10.1371/journal.pone.0216231CrossrefMedlineGoogle Scholar - 19.
Sutradhar R, Austin PC . Relative rates not relative risks: addressing a widespread misinterpretation of hazard ratios.Ann Epidemiol. 2018; 28:54–57. doi: 10.1016/j.annepidem.2017.10.014CrossrefMedlineGoogle Scholar - 20.
Powell L, Edwards KM, McGreevy P, Bauman A, Podberscek A, Neilly B, Sherrington C, Stamatakis E . Companion dog acquisition and mental well-being: a community-based three-arm controlled study.BMC Public Health. 2019; 19:1428. doi: 10.1186/s12889-019-7770-5CrossrefMedlineGoogle Scholar