Table 1.

Summary of main findings from included studies across depression, anxiety, and ‘core mental health’

Author Year Country	Quality ratinga	Sample size and characteristics	Measures (predictor and outcome)	Length of follow-up	Resultsb  +  + , + , -	Statistical analysis and main results
Depression ‘pure onset’
Beutel (2018) Germany	***	N = 10,036 Age: 40–74	Loneliness 1 = no loneliness or distress; 2 = slight; 3 = moderate; 4 = severe loneliness) Depression: PHQ-9 (> 10)	5 years	+  +	Baseline loneliness associated with onset of depression at 5 years: Model 1 (excluding PHQ-9 at baseline model): loneliness aOR 2.01 (1.479, 2.71), Model 2 (including PHQ-9 at baseline, i.e. those with subclinical depression): loneliness aOR 1.55 (1.14, 2.10)
Conde-Sala (2019) Spain	***	N = 22 268 Older adults over 65	Loneliness: Hughes et al. scale Depression: 12-item EURO-D	2 years	+  +	Loneliness at baseline significantly associated with new incidence of clinically depressive symptoms at follow-up Overall across cohort: Incidence (vs no depression) aOR 1.63 (1.62, 1.64)
Green (1992) UK	*	N = 1070 Older people over 65	Loneliness: single item Depression: AGECAT diagnosis	3 years	+	New depression group significantly more lonely at baseline (chi-squared 16.98, p < 0.0005) New depression group significantly more lonely at baseline (chi-squared 16.98, p < 0.0005) Odds of loneliness at baseline in ‘new depression’ group significantly different (aOR 1.82)
Harris (2006) England	**	N = 809 Older adults over 65	Loneliness: 1-item Depression: 15-item Geriatric Depression Scale	2 years	+  +	After adjusting for confounders, loneliness at baseline predicted onset of depression; Sometimes lonely: aOR 2.8 (1.3–5.8) Often/always lonely aOR 9.3 (2.9–30.3)
Kraav 2021 (Finland)	****	N = 2339 Middle-aged men	Loneliness 11-item ICD-10 depression	23.5-year	+  +	Loneliness predicting new onset of depression; HR 1.04 (1.02–1.06)
Prince (1998) UK	***	N = 538 Older people over 65	Loneliness: single item Depression (SHORT-CARE)	1 year	+  +	Risk of new onset depression higher in people who often felt lonely RR 3.6 (2.0–6.4)
Sjoberg (2013) Sweden	***	N = 245 (1901 birth cohort) N = 310 (1930 birth cohort) Older people (recruited aged 70)	Loneliness: single item Depression: DSM-IV diagnosis	5 years	+  +	Baseline loneliness associated with onset of depression at 5 years aOR 3.81 (1.10–13.20); aOR 2.83 (1.23–6.39)
Smalbrugge (2006) Netherlands	***	N = 218 Older people over 55 (48% under 80)	Loneliness: De Jong Gerveld loneliness scale Depression: Geriatric Depression Scale	6 months	–	Loneliness NOT associated with onset of depression Unadjusted OR 0.07 (0.01–0.57)
Stessman (2014) Israel	**	N = 340 (1990 recruited cohort) N = 705 (1998 recruited cohort) Two cohorts: (1) Age 70–78, (2) 78–85	Loneliness: single item Depression: brief symptoms inventory	7 years	- ( +)	‘Never lonely’ vs ‘any loneliness’ NOT associated with new depression: Age 70–78 aOR 0.61 (0.137–2.68); Age 78–85 aOR 1.61 (0.8–3.25) However, categorising as ‘Never/rarely lonely’ vs ‘often or very often’ at age 78 predicts new depression at 85. aOR 2.42 (1.18–4.9)
Other depression studies
Beller (2021) 14 European countries	**	N = 40 797 57% female, mean age 68	Loneliness: UCLA 3-item scale Depression: EURO-D scale	4 years	+  +	Loneliness correlated with later depression Fully adjusted b = 0.377 (0.297–0.456) p < 0.001 Stratified by gender: women B 0.480 (0.375–0.585) Men: 0.181 (0.057–0.306)
Cacioppo (2006) USA	***	Older people, aged 50–68 N = 212 CHASRSc cohort (baseline 2002)	Loneliness: UCLA loneliness scale Depression: CES-D (minus loneliness item)	3 years	+  +	Latent growth curve modelling Baseline loneliness (year 1) predicts subsequent depression (coefficient 1.40, SE 0.55, p < 0.05)
Cacioppo (2010) USA	***	Older people, aged 50–68 N = 229 CHASRSC cohort (baseline 2002)	Loneliness: UCLA loneliness scale Depression: CES-D (minus loneliness item)	Annual follow-up for 5 years	+  +	Significant 1-year cross-lagged effect of loneliness on depressive symptoms B = 0.18 (0.09–0.30) across 5 years Loneliness stable over time
Domenech-Abella (2019) Ireland	***	N = 5066 Community-dwelling adults aged 50 years and older	Loneliness: 5-item UCLA loneliness Depression: Composite International Diagnostic Interview-Short Form	3 waves 5–6 years	+  +	Loneliness at wave 2 predicted depression at wave 3: aOR 1.22 (1.15–1.30)
Groarke (2021) UK	***	N = 1958 Age range 18–87	Loneliness: UCLA 3-item Depression: PHQ-9	4 months	+  +	Loneliness at t1 associated with depression at T2 0.523 (p < 0.001)
Goosby (2013) USA	**	N = 10,564 Nationally representative sample of high school students (18 +)	Loneliness (items combined from CES-D, not validated) Depression: CES-D	2 years	+  +	Loneliness associated with new onset of depression aOR 1.41 (1.35–1.50) unadjusted OR 1.45, p < 0.001 Parental support across baseline and follow-up moderates the depression between loneliness and depression aOR 1.25 (1.14–1.42)
Johansson (2021) Sweden	**	N = 1836 University students	Loneliness: UCLA 3-item Depression: Depression and anxiety stress scale (DASS-21)	6 months	_	Loneliness association with depression at follow-ups Depression FU1: -0.79 (-1.23—0.34) FU2: -1.23(-1.71- -0.74)
Krendl (a) (2021) USA	**	N = 34 White female university students	Loneliness: UCLA 3-item Depression: PHQ-8	3–6 months	_	B 0.33 for loneliness predicting change in depression score, did not reach significance
Krendl (b) (2021) USA	**	N = 93 Older adults	Loneliness: UCLA 3-item Depression: PHQ-8	6–9 months	+  +	Greater loneliness predicted greater increases in depression score from Time 1 to Time 2 (= 0.89,.PHQ SD = 2.52), r(86) = 0.22, p = .045
Lee (2021) UK	****	N = 9432 Older adults	Loneliness: UCLA 3-item Depression: CES-D	12 years	+  +	For binary depression outcome: aOR univariable 1.93 (1.82–2.04); adjusted for sociodemographic factors: aOR 1.72 (1.62–1.82). fully adjusted aOR 1.28 (1.21—1.35)
Lim (2011) Singapore	*	N = 2799 Older people, mean age 66 (55+)	Loneliness: single item Depression: Geriatric Depression Scale Score	2 years	+  +	Loneliness significant predictor of higher depression scores after 2 years aOR 1.39, B = 0.33, SE 0.36, p = 0.03
Luoma (2015) Finland	***	N = 329 Mothers (recruited first trimester) Mean age 26.6	Loneliness: single item Postnatal depression (Edinburgh Postnatal Depression Score)	16 to 17-year follow-up	+  +	Group-based modelling identified a four cluster model was best for predicting depression trajectories (‘high stable’/’intermittent’/low stable’ and ‘very low’) Feeling lonely associated with high stable depression symptoms. aOR 2.1 (1.0–4.2) p 0.041
Okruszek (2021) Poland	**	N = 511 Young adults	Loneliness: UCLA-20 Depression: GHQ-30	2 weeks	_	Path analysis. Path from baseline loneliness to depression 0.05, not significant
Richardson (2017) UK	**	N = 454 (2 cohorts combined) University students Mean age 19.9	Loneliness: 3-item UCLA loneliness scale CES-D	2 cohorts followed up over 12–14 months	+	Baseline loneliness correlated with depression (time 2 r = 0.51, time 3 r = 0.48, time 4 r = 0.42) p < 0.001 After accounting for demographics and baseline scores, loneliness predicted depression only at T4 (β = 0.14, p < 0.05)
Theeke (2007) (thesis)	****	N = 13 812 Older people (50+) HRS cohortd	Loneliness: single item from within CES-D Depression (CES-D 7 items)	3 waves 2002–4	+	Never lonely vs chronically lonely: mean difference in depression score 1.55 (error 0.03 p < 0.005) Briefly lonely vs chronically lonely mean difference in depression score 0.72 (error 0.04) p < 0.005
Vicente (2014) Portugal	**	N = 83 Older people (institutionalised; average age 79.5)	Loneliness: UCLA loneliness scale Depression: GDS	2 years	–	Those whose depression scores worsened over time (including people who had no depression at baseline and people who had depression) had higher loneliness scores at baseline, but did not reach statistical significance
Luo (2012) USA	***	N = 2101 Older people, mean age 67 Subset of HRS cohortd	Loneliness: 3-item UCLA loneliness scale Depression: CES-D minus ‘I feel lonely’ and sleep items	2 years	+  +	Significant 2-year cross-lagged effect of loneliness on depression B = 0.132, p < 0.001
Xerxa (2021) USA	***	N = 1420 Young adult (19,21,25,39)	Loneliness: parent- reported and child-reported. Item on child and adolescent psychiatric assessment Anxiety: Young Adult Psychiatric Assessment	21 years (four time points)	+	Model adjusted for sociodemographics and childhood psychiatric problems: Childhood-reported loneliness depression aOR 1.86 (0.73–4.71) p 0.190; Model 2 parent-reported loneliness depression aOR 0.91 (0.32–2.57) p 0.861
Anxiety
Domenech-Abella (2019) Ireland	***	N = 5066 Community-dwelling adults aged 50 years and older	Loneliness: 5-item UCLA loneliness scale Composite International Diagnostic Interview-Short	5–6 years	+  +	Loneliness at wave 2 predicted anxiety at wave 3: aOR 1.60 (1.10–2.34)
Flensborg-Madsen (2012) Denmark	***	N = 4497 Adults (mean age 44.9)	Loneliness: single item Anxiety disorder presence (ICD8/10)	13 years	+  +	Multiple Cox regression analysis Women: ‘Yes’ vs ‘no’ lonely and being hospitalized with anxiety: HR 2.01 (1.31–3.06) ‘in doubt’ vs ‘no’ HR 1.14 (0.64–2.01) Men: ‘Yes’ vs ‘no’ lonely and later being hospitalized with anxiety: HR 2.34 (1.34–4.09) ‘in doubt’ vs ‘no’ HR 2.03 (1.19–2.63)
K rendl (a) (2021)	**	N = 34 University students	Loneliness: UCLA 3-item Anxiety: PHQ-8	3–6 months	+	Baseline loneliness and anxiety at follow-up. beta 0.28 anxiety p < 0.05
Johansson (2021) Sweden	**	N = 1836 University students	Loneliness: UCLA 3-item Anxiety: Depression and anxiety stress scale (DASS-21)	6 months	_	Anxiety FU1 − 0.65 (− 0.98 to − 0.31) FU2 (− 0.90–1.27 to − 0.54)
Richardson (2017) UK	**	N = 454 (2 cohorts combined) University students Mean age 19.9	Loneliness: 3-item UCLA loneliness scale Depression: CES-DCES-D	2 cohorts followed up over 12–14 months	+	Baseline loneliness correlated with anxiety time 2 (r = 0.41, T3 r = 0.40, T4 r = 0.34) p < 0.001 After accounting for demographics, loneliness predicted anxiety only at T3 (β 0.15, p < 0.01)
Xerxa (2021) USA	***	N = 1420 Young adult (19,21,25,30)	Loneliness: parent- reported and child-reported. Item on child and adolescent psychiatric assessment Anxiety: Young Adult Psychiatric Assessment	21 years	+  +	Adjusted for sociodemographic factors and childhood psychiatric problems: Childhood-reported loneliness anxiety aOR 3.53 (1.55–8.04) p 0.002
Other mental health problems
Ahrens (2021) Germany	****	N = 526 Mean age 31	Loneliness: LON 3-item, based on UCLA 3-item Mental health dysfunction: GHQ-8. ‘Significant dysfunction’ taken to be 23/24	7–8 weeks	+  +	Beta coefficient: −0.42(− 0.58 to − 0.27) p < 0.001 (loneliness is higher with lower scores on the version of scale used (LON)
Antonelli (2021) Brazil	**	N = 1674 Age range 18–75	Loneliness: UCLA 3-item Suicidal ideation (4 categories) Have you had any thoughts about killing yourself?	1 month	+  +	aOR 2.12(1.06–4.24) p 0.033
Nuyen (2019) Netherlands	***	N = 4007 General population aged 18–64 Mean age: 44.3 years	Loneliness: De Jong Gerveld loneliness scale Common mental disorders: Composite International Diagnostic Interview (CIDI) version 3.0	Baseline: 2013–2015 3 year follow-up	+  +	After adjusting for covariates, loneliness at wave 2 predicted the onset of severe 12-month CMD at wave 3; aRRR 3.28 (1–54-7.02) After adjusting for covariates, loneliness at wave 2 did not predict onset of a mild-moderate 12-month CMD; aRRR 0.94 (0.50–1.17)
Richardson (2017) UK	**	N = 454 (2 cohorts combined) University students Mean age 19.9	Loneliness: 3-item UCLA loneliness scale Core mental health:’ CORE-GP’	2 cohorts followed up 12–14 months	+	After adjusting for demographics and baseline scores, loneliness predicted core mental health at T2 (β 0.11, p < 0.05)
Shaw (2021) UK	****	N = 502 536 Middle-aged adults	Loneliness: single item: ‘do you feel lonely?” ICD-10 for suicide and self-harm	9 years	+  + (self-harm) -(suicide)	Self-harm (all p < 0.001) Model adjusted for ALL sociodemographic factors all physical health factors, plus perceived social support and living arrangements men: a OR 1.74 (1.40–2.76); women 1.89 (1.57–2.28 Suicide—significant univariable association (men) 3.20 (2.35–4.36), but fully adjusted model not significant 1.39 (0.97–1.99) Fully adjusted model, women 0.92 (0.54–1.57)

^aDetailed quality ratings in supplementary material 3

^b++ loneliness associated with onset p < 0.05, adjusted; + loneliness associated with onset p < 0.05 unadjusted,—non-significant

^cBoth drawn from same larger Chicago Health, Ageing and Social Relations study, but different statistical approaches, and follow-up

^dHealth and Retirement Study