Why poor sleep and anxiety keep setting each other off
The core finding across several longitudinal studies is a feedback loop. In a large cohort of community-dwelling adults, people with disturbed sleep at the start were about 1.89 times more likely to develop anxiety a year later, while anxious people were about 1.20 times more likely to develop sleep disturbance [1]. The same two-way pattern appears in early adolescents followed over 1.5 years: changes in sleep quality went with changes in anxiety symptoms, and changes in anxiety went with changes in sleep quality [2].
That loop is not just a statistical artifact of asking people about both problems at once. When researchers track the same people across multiple time points, the earlier level of one problem still predicts the later level of the other, even after accounting for where each person started [1][2]. In practical terms, a bad night can make the next day feel more threatening, and a anxious day can make the next night harder—so each problem keeps the other alive.
What a cross-lagged design actually shows that a one-time survey cannot
A single survey can tell you that people with sleep problems are also anxious, but it cannot tell you which came first. A cross-lagged panel model uses at least two measurement waves and asks a sharper question: does sleep at time one predict anxiety at time two, after controlling for anxiety at time one—and does anxiety at time one predict sleep at time two, after controlling for sleep at time one? That is the design used in the adult cohort [1], the adolescent study [2], and the TRAILS adolescent cohort, which used a random-intercept cross-lagged panel model to separate stable differences between people from change within people over time [4].
The random-intercept version matters because it addresses a common objection: maybe some people are simply more reactive or more vulnerable, and that stable trait explains both problems. By separating between-person differences from within-person change, the TRAILS analysis tests whether a person's sleep and anxiety move together over time in a way that is not just a fixed personal characteristic [4]. That makes the bidirectional claim stronger than a simple correlation, though it still describes temporal prediction rather than proving a biological mechanism.
Does the loop look the same in everyone, and which direction is stronger?
The strength of the two directions is not necessarily equal. In the adult cohort, the path from sleep disturbance to anxiety was larger than the reverse path, leading the authors to conclude that poor sleep's effect on anxiety seems more significant and that targeting sleep may help break the cycle [1]. That is a claim about relative magnitude in one adult sample, not a universal ranking.
Age and context also shape what the loop looks like. In early adolescents, the bidirectional associations between sleep quality and anxiety did not differ between males and females [2]. In college students measured during and after collective isolation, anxiety and sleep disturbance were bidirectionally related, but anxiety did not predict later sleep disturbance once the isolation period ended—suggesting the loop can loosen when the external stressor lifts [5]. These differences are a reminder that the direction and strength of the relationship depend on who is being studied and what else is happening in their lives.
If the loop runs both ways, where should treatment start?
Because the relationship is bidirectional, treating only one side may leave the other side pulling the person back. The adult cohort's authors argue that timely sleep-focused intervention could help break the vicious circle and improve quality of life [1], while the adolescent study's authors suggest clinicians consider both sleep and anxiety together when planning treatment [2]. Neither set of results proves that treating sleep will cure anxiety, but they do suggest that sleep is a reasonable place to intervene—especially when sleep problems appear first or are more severe.
What the evidence does not yet settle is exactly how much of the loop is causal in a strict sense. Cross-lagged models show temporal prediction and are consistent with mutual causation, but they cannot rule out every shared underlying factor. The TRAILS random-intercept approach narrows that gap by focusing on within-person change [4], yet the question of which intervention breaks the cycle most effectively remains open.
Sources used in this answer
The bidirectional relationship between sleep disturbance and anxiety: Sleep disturbance is a stronger predictor of anxiety
In a large adult cohort, sleep disturbance predicted later anxiety more strongly than anxiety predicted later sleep disturbance, with cross-lagged models supporting mutual causation.
Longitudinal bidirectional effects between sleep quality and internalizing problems
In early adolescents followed over 1.5 years, sleep quality and anxiety symptoms predicted each other over time, with no difference between males and females.
Anxiety, sleep problems, and vigorous physical activity: bidirectional associations from early adolescence to early adulthood in Swedish adolescents
The study examined bidirectional associations among physical activity, sleep problems, and anxiety in Swedish adolescents using a cross-lagged panel model.
The bidirectional association between sleep problems and anxiety symptoms in adolescents: a TRAILS report
Using a random-intercept cross-lagged panel model in adolescents and young adults, the study assessed bidirectional temporal associations between sleep problems and anxiety symptoms.
Cross-lagged relationship between anxiety, depression, and sleep disturbance among college students during and after collective isolation
Among college students during and after isolation, anxiety and sleep disturbance were bidirectionally related, but anxiety did not predict later sleep disturbance after isolation ended.