Study (publication year) (n-38) Effect size (95% CI) Effect size (95% CI) Weight (%) Study (publication year) (n=45) Effect size (95% CI) Effect size Weight (95% CI) (%) Saito (2021) 2.05 (1.52, 2.73) 1.55 Yu (2022) Saito (2021) Naito (2021) Lennartsson (2021) 1.30 (1.12, 1.50) 2.78 Moreno-Tamayo (2022) Beller (2022) 1.26 (1.17, 1.36) 3.45 Barnes (2022) 2.54 (1.59, 4.08) 0.79 Lennartsson (2021) Crowe (2021) 1.38 (1.21, 158) 2.91 Youm (2021) Ward (2021) Youm (2021) Gilmour (2020) Gilmour (2020) Gronewold (2020) Kraav (2020) Hoogendijk (2020) Kassandra (2019) Sakurai (2019) Laugesen (2018) Laugesen (2018) Smith (2018) Baller (2018) Elovainio (2017) 1.37 (1.04, 1.81) 1.65 Ward (2021) 1.53 (0.82, 2.84) 0.50 Schutter (2021) 1.30 (1.10, 1.50) 2.69 Gao (2021) 1.40 (1.20, 1.60) 2.81 Gao (2021) 1.47 (1.09, 1.97) 1.53 Gao (2021) 1.09 (1.02, 1.15) 3.57 Gao (2021) 1.06 (0.92, 1.21) 2.87 Gao (2021) 1.43 (1.36, 1.50) 3.64 Gao (2021) 1.34 (0.64, 2.81) 0.37 Gao (2021) 1.70 (1.10, 2.60) 0.92 Gao (2021) 1.60 (0.83, 2.90) 0.50 Crowe (2021) 1.29 (1.10, 1.51) 2.66 Kraav (2020) 1.21 (1.00, 1.46) 2.36 Novak (2020) 1.26 (1.20, 1.33) 3.62 Novak (2020) Chan (2015) 1.28 (0.90, 1.82) 1.23 Lara (2020) Drageset (2013) 1.02 (0.96, 1.08) 3.57 Hoogendijk (2020) Eng (2002) Pantell (2013) Pantell (2013) Yang (2013) Yang (2013) Tilvis (2012) Tilvis (2012) Saito (2012) Stringhini (2012) Stringhini (2012) Barkman (2004) Barkman (2004) Keller (2003) Greenfield (2002) Greenfield (2002) Maier (1999) Seeman (1993) Seeman (1993) 1.62 (1.29,2.02) 2.05 Gilmour (2020) 1.75 (1.38, 2.23) 1.92 Gilmour (2020) 1.53 (1.26, 1.87) 2.28 1.45 (1.20, 1.75) 2.37 1.03 (0.87, 1.23) 2.51 Wang (2018) 1.00 (0.87, 1.15) 2.85 Stringhini (2018) 1.19 (1.02, 1.39) 2.70 Beller (2018) 1.18 (0.99, 1.40) 2.51 Olaya (2017) 1.18 (0.90, 1.53) 1.73 2.70 (1.17, 6.23) 0.30 3.64 (0.72, 18.58) 0.08 1.25 (0.97, 1.61) 1.82 Julsing (2016) 1.20 (0.97, 1.48) 2.16 0.99 (0.71, 1.39) 1.30 0.83 (0.56, 1.22) 1.07 1.11 (0.99, 1.25) 3.08 Chan (2015) 0.99 (0.66, 1.51) 0.97 1.42 (0.87,2.32) 0.75 Seeman (1993) Seeman (1993) Seeman (1993) Seeman (1993) Kaplan (1988) 2.40 (1.35, 3.07) 0.98 1.28 (0.79,2.03) 0.80 Steptoe (2013) 1.89 (1.18, 3.04) 0.79 Drageset (2013) 1.78 (1.05, 3.03) 0.66 Tilvis (2012) 2.00 (1.18, 3.39) 0.67 Tilvis (2012) Schoenbach (1986) Schoenbach (1986) Yu (2022) Moreno-Tamayo (2022) Barnes (2022) Luo (2022) Komnej (2022) Kotozaki (2022) Kotozaki (2022) 1.50 (1.00, 2.20) 1.05 Parissinotto (2012) Luo (2012) 1.40 (0.80, 2.30) 0.66 1.22 (1.18, 1.25) 3.73 Holwerda (2012) Holwerda (2012) 1.24 (1.10, 1.38) 3.11 Tilvis (2012) 2.47 (1.91, 3.18) 1.81 Tilvis (2011) 1.33 (1.15, 1.54) 2.78 Shiovitz-Eara (2010) 1.62 (1.14, 2.29) 1.24 Stak (2005) Jeong (2022) Jeong (2022) Overall (1=77.8%, P=0.000) 1.38 (104,1.83) 1.49 (1.02, 2.19) 4.50 (1.90, 9.90) 1.61 Pitkala (2004) 1.09 Maier (1999) 0.30 Panninx (1997) T T 1.20 (0.50, 2.80) 1.32 (1.26, 1.39) 0.28 Grand (1990) 100.00 0.5 1.0 18.6 Weights are from random effects analysis. 1.01 (0.98, 1.04) 1.17 (1.05, 1.29) 3.34 (1.67, 6.71) 2.45 2.26 0.49 1.57 (1.28, 2.93) 1.01 1.20 (0.74, 1.95) 0.83 1.01 (0.43,2.36) 0.35 1.26 (0.94, 1.68) 1.44 1.06 (0.99, 1.14) 1.19 (0.97, 1.46) 2.37 1.82 1.13 (0.92, 1.39) 1.81 0.74 (0.49, 1.11) 1.02 1.37 (0.96, 1.94) 1.21 1.11 (0.82, 1.49) 1.41 1.10 (0.83, 1.46) 1.47 1.58 (1.03, 2.41) 0.98 1.03 (0.65, 1.62) 0.89 1.48 (1.28, 1.72) 2.08 1.11 (1.03, 1.18) 1.32 (0.77,2.28) 2.37 0.71 1.64 (0.98, 2.76) 1.02 (0.94, 1.12) 1.06 (0.92, 1.22) 0.76 2.32 2.11 Conde-Sala (2020) O'Silleabhin (2019) Henriksen (2019) 0.90 (0.80, 1.20) 1.83 1.10 (0.90, 1.30) 1.92 1.19 (1.04, 1.37) 2.12 Elovainio (2017) Tanskanen (2016) Tabue Teguo (2016) Holwerda (2016) Holwerda (2016) Ellwardt (2015) Stessman (2014) Stessman (2014) Stessman (2014) 1.18 (1.07, 1.29) 2.30 0.90 (0.70, 1.15) 1.62 1.00 (0.80, 1.40) 1.48 1.23 (1.10, 1.38) 2.22 0.98 (0.86, 1.10) 2.19 1.00 (0.61, 1.64) 0.81 0.99 (0.93, 1.06) 2.38 1.06 (1.01, 1.12) 2.41 1.11 (1.00, 1.23) 2.26 1.40 (0.85, 2.31) 0.79 1.13 (0.97, 1.32) 2.05 0.99 (0.85,1.16) 2.05 1.02 (0.99, 1.06) 2.44 1.38 1.34 (0.99, 1.82) 1.06 (0.54, 2.10) 0.50 1.10 (0.69, 1.77) 0.86 0.84 (0.56, 1.27) 1.02 Jylh (1989) Jylh (1989) Overall (1 = 91.1%, P=0.000) 0.54 1.00 Weights are from random effects analysis. 0.92 (0.78, 1.09) 1.99 0.96 (0.90, 1.06) 2.33 1.39 (1.15, 1.68) 1.89 1.16 (1.01, 1.33) 2.13 1.45 (1.11, 1.88) 1.55 1.07 (0.99, 1.17) 2.33 1.30 (1.04, 1.62) 1.74 1.01 (0.87,1.17) 2.08 1.18 (0.99, 1.42) 1.93 1.17 (1.02, 1.33) 2.15 1.83 (1.71, 1.87) 2.42 1.30 (0.80, 1.90) 0.96 1.16 (0.99, 1.39) 1.98 1.28 (1.14,1.44) 2.21 1.06 (1.00, 1.12) 2.40 1.42 (0.81, 2.50) 0.67 1.02 (0.75, 1.40) 1.35 1.17 (0.79, 1.74) 1.06 1.14 (1.08, 1.20) 100.00 6.71 Fig. 2 Association of SI or loneliness and risk of all-cause mortality. a, Association of SI and risk of all-cause mortality. b, Association of loneliness and risk of all-cause mortality. The pooled effect estimates were generated using a random-effects model. The statistical tests were two-sided. n indicates the number of studies included. Only the first author of each study is listed. Each solid diamond represents the effect size of each study. The horizontal bars represent the 95% CIs for each study effect size. The hollow diamond represents the overall estimated effect and its 95% CI. The data are presented as HRs (95% CI). Article https://doi.org/10.1038/s41562-023-01617-6 Articles on SI and mortality identified through database searching (n = 9,328): PubMed (n=2,193) Embase (n = 4,133) Web of Science (n = 3,002) Articles on loneliness and mortality identified through database searching (n = 5,030): PubMed (n = 1,141) Embase (n = 2,126) Web of Science (n = 1,763) Excluded by title and abstract review on SI (n = 9,210) and loneliness (n = 4,966) SI was not an independent variable (n = 27) Mortality was not the outcome variable (n = 14) Duplicated reports (n = 17) Insufficient information, different forms of outcome data, unable to extract the effect size (n = 7) The subjects were patients with HIV and HIV-infected/uninfected older veterans (n = 2) The subjects were elderly critically ill patients in intensive care units with a history of social isolation exposure (n = 1) Full-text reports retrieved for detailed evaluation on SI (n = 118) or on loneliness (n = 64) Full-text articles excluded on SI (n = 68) or on loneliness (n = 13) Articles included in the meta-analysis on SI and mortality (n = 50) or on loneliness and mortality (n = 51) Loneliness was not an independent variable (n = 1) Mortality was not the outcome variable (n = 3) Duplicated reports (n = 3) Insufficient information, different forms of outcome data, unable to extract the effect size (n = 6) Duplication: articles on both SI and loneliness and mortality (n = 18) Manually retrieved articles (n = 7) Articles included in the meta-analysis (n = 90) Fig.1|Studies included in the present review. The retrieval processes shown found 90 studies that met the selection criteria and were relevant to this meta-analysis. a significant difference was observed only for country, between the US and non-US countries (P=0.039 Supplementary Table 12) There P=0.296) (Fig. 4, Supplementary Fig. 6 and Supplementary Table 11). The subgroup analysis of Sland cancer mortality did not reveal sources b Study (publication year) (n=8) Effect size (95% CI) Effect size (95% CI) Weight (96) Ward (2021) Kraav (2020) Novak (2020) Novak (2020) Stringhini (2018) Elovainio (2017) Julsing (2016) Eaker (1992) Olsen (1991) Olsen (1991) Overall (1 = 77.3%, P=0.000) 0.58 1.00 1.28 (0.77, 2.12) 7.44 1.12 (1.01, 1.24) 18.89 1.52 (0.78, 2.96) 5.09 2.25 (1.14, 4.45) 4.93 1.44 (1.19, 1.74) 16.25 9.20 0.95 (0.88, 1.02) 19.45 1.18 (0.58, 2.39) 4.65 4.00 (1.80, 9.20) 3.71 1.70 (1.03, 2.81) 7.53 1.09 (0.79, 1.49) 12.07 1.30 (1.09, 1.54) 100.00 Weights are from random effects analysis. Fig. 3 | Association of SI or loneliness and risk of cardiovascular or circulatory system disease mortality. a, Association of SI and risk of cardiovascular or circulatory system disease mortality. b, Association of loneliness and risk of cardiovascular or circulatory system disease mortality. The pooled effect estimates were generated using a random-effects model. The statistical tests were two-sided. n indicates the number of studies included. Only the first author of each study is listed. Each solid diamond represents the effect size of each study. The horizontal bars represent the 95% Cls for each study effect size. The hollow diamond represents the overall estimated effect and its 95% CI. The data are presented as HRS (95% CI). Nature Human Behaviour Article https://doi.org/10.1038/s41562-023-01617-6 Study (publication year) Effect size (n=13) (95% CI) Effect size (95% CI) Weight (%) Fleisch Marcus (2017) Elovainio (2017) Ward (2021) 1.15 (0.92, 1.43) 2.82 1.32 (1.22, 1.43) 21.76 0.99 (0.65, 1.49) 0.80 Pinheiro (2021) 1.20 (0.99, 1.44) 3.91 Kassandra (2019) 1.07 (0.77, 1.49) 1.26 Kassandra (2019) 1.25 (1.17, 1.34) 29.82 Kraav (2020) 1.07 (0.95, 1.21) 10.08 Yang (2013) 1.30 (0.87, 1.95) 0.84 0.86 (0.59, 1.25) 0.97 Eng (2002) Berkman (2004) Kassandra (2019) Kassandra (2019) Koh-Bell (2021) 3.60 (0.99, 13.01) 0.08 0.94 (0.68, 1.30) 1.31 1.26 (1.17, 1.36) 24.24 0.76 (0.28, 2.06) 0.14 Yang (2013) Luo (2022) Overall (1 = 42.8%, P = 0.040) 0.86 (0.55, 1.34) 0.69 1.02 (0.73, 1.41) 1.27 1.22 (1.18, 1.27) 100.00 0.28 1.00 13.00 Study (publication year) (n = 3) Effect size (95% CI) Effect size (95% CI) Weight (%) Ward (2021) Kraav (2020) Elovainio (2017) Overall (1 = 0.0%, P=0.952) 0.68 1.00 1.53 1.02 (0.68, 1.53) 3.47 1.09 (0.96, 1.24) 34.91 1.09 (0.99, 1.20) 61.62 1.09 (1.01, 1.17) 100.00 Fig. 4 | Association of SI or loneliness and risk of cancer mortality. a, Association of SI and risk of cancer mortality. b, Association of loneliness and risk of cancer mortality. The pooled effect estimates were generated using a fixed-effects model. The statistical tests were two-sided. n indicates the number Nature Human Behaviour of studies included. Only the first author of each study is listed. Each solid diamond represents the effect size of each study. The horizontal bars represent the 95% Cls for each study effect size. The hollow diamond represents the overall estimated effect and its 95% CI. The data are presented as HRs (95% CI). a Study (publication year) (n=15) Effect size (95% CI) Effect size (95% CI) Weight (96) Naito (2021) Elovainio (2017) Ward (2021) Heffner (2011) Kassandra (2019) Kassandra (2019) 1.30 (1.13, 1.50) 9.42 1.24 (1.17, 1.31) 13.57 1.19 (0.75, 1.88) 2.11 2.66 (1.03, 6.85) 0.56 1.51 (1.16, 1.97) 4.94 1.29 (1.21, 1.38) 13.18 Kraav (2020) 1.06 (0.97, 1.16) 12.18 Stringhini (2012) 1.66 (1.22,2.26) 4.00 Yang (2013) 1.56 (1.16, 2.10) 4.22 Eng (2002) 1.37 (0.91, 2.08) 2.52 Berkman (2004) 1.29 (0.16, 10.41) 0.12 Kaplan (1988) Kaplan (1988) Kassandra (2019) Kassandra (2019) Stringhini (2012) 1.80 (0.87, 3.72) 0.92 1.72 (0.77, 3.84) 0.76 1.49 (1.14, 1.96) 4.79 1.47 (1.37, 1.57) 13.08 1.40 (0.77, 1.54) 3.35 Yang (2013) 1.47 (1.11, 1.95) 4.54 Luo (2022) 1.28 (1.01, 1.62) 5.72 1.34 (1.25, 1.44) 100.00 Overall (1 = 63.3%, P=0.000) 0.16 1.00 10.40 Weights are from random effects analysis.