What cut loneliness most in trials

The kind of programme that cut loneliness most in randomized trials, how the other kinds did, and what 90 studies found about social isolation, loneliness and the risk of dying. Sources are linked throughout.

The short version

In 20 randomized trials pooled in a 2011 review, programmes that worked on how lonely people think about others, such as cognitive behavioural and reframing approaches, cut loneliness the most (Masi, 2011).
Support and companionship programmes cut loneliness a little. Online contact and social skills training, two trials each, showed no clear effect (Masi, 2011).
In 90 studies that followed 2,205,199 adults over time, socially isolated adults had a 32% higher risk of dying than adults who weren't isolated, and lonely adults a 14% higher risk than adults who weren't lonely (Wang, 2023).
These studies watched people's lives, so they can't show that isolation or loneliness causes early death.

What cut loneliness in trials

A 2011 review gathered 50 studies of programmes meant to reduce loneliness, published up to September 2009. 20 of them were randomized trials, where people were put into a programme or a comparison group by chance. 11 of those 20 included adults aged 60 or older (Masi, 2011).

The trials tested four kinds of programme. The number for each is the review's pooled effect size: the gap in loneliness scores between the programme group and the comparison group after the programme, in standard units. Below 0 means the programme group was less lonely (Masi, 2011).

What the programmes did	Trials	Effect on loneliness
Worked on how people think about others and about being alone, through cognitive behavioural, reframing and similar approaches	4	−0.60, the largest
Gave support or companionship, such as regular contact, support groups or a pet	12	−0.16, small
Gave more chances for contact, online or by email	2	−0.06, could be chance
Trained social skills	2	+0.02, could be chance

Only the first two showed a clear effect. The four kinds were tested in different trials with different people, so the table shows the size of each pooled result and doesn't settle an order among the last three (Masi, 2011). The authors suggest that reducing social isolation does not necessarily reduce loneliness.

Where to start

The programmes with the largest effect worked on thoughts, such as expecting to be rejected. Some were led by a counsellor or psychologist. The review doesn't say which approach suits a particular person.
Isolation and loneliness are different. You can be lonely among people and content alone, and more contact may not ease loneliness by itself.
Only 4 trials tested the approach with the largest effect, and the review searched studies up to 2009, so treat it as a lead to follow.

The review of 90 studies

Wang and colleagues combined 90 studies that followed 2,205,199 people aged 18 and over through time. They looked at social isolation, being cut off from other people, and loneliness, feeling alone, separately (Wang, 2023).

From a review of 90 studies that followed 2,205,199 adults over time.

In adults in general, socially isolated people were compared with people who weren't isolated, and lonely people with people who weren't lonely:

Socially isolated people had a 32% higher risk of dying from any cause, and lonely people a 14% higher risk (Wang, 2023).
Socially isolated people had a 34% higher risk of dying from heart and blood vessel disease (Wang, 2023).
Socially isolated people had a 24% higher risk of dying from cancer, and lonely people a 9% higher risk (Wang, 2023).

Among people who already had heart and blood vessel disease, isolated people had a 28% higher risk of dying from any cause. Among people with breast cancer, isolated people had a 51% higher risk of dying from any cause (Wang, 2023).

How far to trust these numbers

The 90 studies watched people's lives and didn't assign anyone to be isolated or lonely. They can't show that isolation or loneliness causes early death, or that changing either would lower anyone's risk by these amounts.
Illness can cut people off from others. An earlier pooled review found that how healthy people were at the start affected the results, and poor health may explain some of the link between isolation and dying (Holt-Lunstad, 2015).
That earlier review found 29% higher odds of dying for social isolation, 26% for loneliness and 32% for living alone, and no difference between measures of being isolated and of feeling isolated (Holt-Lunstad, 2015). The newer review's 32% and 14% came from separate analyses, so they don't show that isolation carries more risk than loneliness.
Many of the loneliness trials were small, from 22 to 644 people, and two kinds of programme had only two trials each. The non-randomized studies in the same review showed bigger effects than the randomized ones, which is why only the randomized trials are shown here (Masi, 2011).