Higher BMI Was Linked to Less Leisure-Time Activity Later in 22,696 Finnish Twins

A Finnish twin study followed body mass index and leisure-time activity across four assessments between 1975 and 2011. Higher BMI predicted less leisure-time activity years later, while the reverse direction held up far less consistently.

Leisure-time activity and body weight travel together across adult life, but which one moves the other has never been settled. A paper published on October 21, 2026 in the International Journal of Behavioral Nutrition and Physical Activity put three different causal methods to the same Finnish twin data and found the arrow pointing mostly in one direction.

The analysis used the Finnish Twin Cohort, a long-running Finnish study of twin pairs. It included 22,696 twins who were 18 to 50 years old at baseline, 52.4 percent of them women. Questionnaire assessments were gathered in 1975, 1981, 1990 and 2011, and 8,527 participants also had genetic data on file.

Rather than trust a single model, the authors ran three approaches in parallel: a random intercept cross-lagged panel model for the repeated measures, one-sample Mendelian randomisation using genetic instruments, and direction of causation twin models including an MR-DoC variant at each measurement point. Triangulation rests on a simple idea, that methods with different weaknesses should not fail in the same way.

The three lines of evidence converged on one direction. A higher BMI at an earlier assessment predicted lower leisure-time activity at a later one, and that pattern appeared in every method. Estimates running the other way, from leisure-time activity to later BMI, were smaller and did not line up across the three approaches.

Three Methods Asked the Same Question

Each of the three tools answers the causal question differently. The cross-lagged panel model separates stable between-person differences from within-person change, so it asks whether a person who gains weight relative to their own average then moves less than usual. The Mendelian randomisation analysis uses inherited variants as instruments, which are fixed at conception and therefore cannot be altered by later behaviour.

The twin models exploit something only twins offer. Identical and fraternal pairs share different proportions of their genes while sharing much of their early environment, which lets a model estimate how far a trait drives another trait rather than simply sitting alongside it.

None of the three removes every objection. Cross-lagged models still rely on the spacing of the assessments, Mendelian randomisation assumes the genetic variants influence the outcome only through the exposure, and twin models carry assumptions about how genes and shared environments combine. The case for triangulation is that an answer surviving all three is harder to explain away than an answer from any one of them.

Where the Weight-to-Activity Signal Was Strongest

The within-person estimates grew larger over time. BMI measured in 1981 predicted leisure-time activity in 1990 at -0.071, with a confidence interval from -0.106 to -0.036. BMI measured in 1990 predicted activity in 2011 at -0.149, with an interval from -0.190 to -0.108. Both are standardised estimates, so they describe shifts in standard deviations rather than minutes or kilograms.

The genetic analyses pushed the same way. MR-Egger estimates for BMI acting on leisure-time activity were -0.096 in 1990 and -0.238 in 2011. The twin models produced four estimates for that direction, ranging from -0.087 to -0.257, all of them negative and all with intervals excluding zero.

Why the Leisure-Time Activity to BMI Direction Stayed Unclear

The reverse path was the inconsistent one. In the twin models, estimates for leisure-time activity acting on BMI were negative, running from -0.029 to -0.132. But the cross-lagged panel model found no within-person association in that direction at all, so the three methods did not agree.

The authors offer reasons rooted in measurement rather than biology. Self-reported activity carries more error than a reported height and weight, which blurs any real effect. Converting questionnaire answers into metabolic equivalent scores adds its own uncertainty. And the genetic instruments available for leisure-time activity are weak compared with those for BMI, which limits what Mendelian randomisation can detect.

What a 36-Year Window Can Show

Spacing assessments across 1975, 1981, 1990 and 2011 is unusual, and it matters for interpretation. Long gaps let slow processes register, which is why the 1990 to 2011 estimate is the largest in the set. They also mean a great deal happens between measurements that no questionnaire captures: job changes, injuries, caregiving, and the ordinary accumulation of small habits.

The widening estimates are consistent with a feedback process rather than a single event. Carrying more weight makes vigorous recreation less comfortable, less of it follows, and the gap compounds across decades.

The baseline year is worth noting too. Participants were 18 to 50 years old at the first assessment and were followed to 2011, which means the final estimate covers people well into middle and later adulthood. A design that follows the same individuals that far captures changes a short trial never could, but it also means the findings describe one Finnish birth range rather than a universal rule.

How Leisure-Time Activity Was Defined

The exposure here was recreational movement reported on questionnaires, not total daily movement. That distinction matters because occupational and commuting activity can run in the opposite direction to recreational activity in a population, a pattern that has surfaced in other work on how activity environments relate to cardiometabolic outcomes.

Self-report also flattens intensity. A device-based study can separate a brisk hour from a gentle one, which is how accelerometer research has isolated the contribution of short bouts of vigorous activity to mortality risk. A questionnaire scored in metabolic equivalents cannot draw that line as sharply.

What This Means in Practice

Read carefully, the paper is less about whether exercise is worthwhile and more about where the resistance sits. If weight gain quietly erodes recreational movement over years, then programmes that treat low activity as a motivation problem may be addressing the wrong link in the chain.

For an individual, none of this changes the general guidance. It does suggest a few sensible moves:

• Treat a drop in recreational activity as a signal worth noticing rather than a character flaw, and look at what has become physically harder.
• Favour activities whose difficulty scales gently with body weight, such as cycling, swimming or resistance work, when walking or running feels punishing.
• Expect weight and activity to influence each other over years, not weeks, and judge progress on that timescale.

The evidence that movement pays off elsewhere remains strong, from work on leisure-time exercise and later dementia risk to trials testing how once-weekly interval walking affects body fat. This study simply says the causal traffic between weight and recreation is busier in one direction than the other.

Anyone deciding what to do with that should talk to a clinician or a qualified coach about their own situation, particularly if weight has changed recently or joint pain is limiting what they can do.