Objectives Psychosocial working conditions are suggested risk factors for low-back pain, but it is unclear whether these associations are causal. The present study examined whether there are lagged and bidirectional associations between job strain and low-back pain and further controlled for unmeasured time-invariant confounding. Methods The study was based on four biennial waves of data from the Swedish Longitudinal Occupational Survey of Health (SLOSH), including 3084 men and women. Cross-lagged analyses using structural equation modeling (SEM) were conducted on job strain, a combination of high job demands and low control, and any as well as low-back pain severity (how much any problems affected the respondents life). Analogous SEM (dynamic panel) models with fixed effects were also fitted to remove confounding from time-invariant factors (such as non-observed individual and environmental factors, eg, genetics, childhood conditions, personality). Results The SEM models indicated bidirectional associations between job strain and any back pain over a 2-year time lag (β=0.21 and 0.19, P<0.05), when adjusting for a range of covariates. Job strain was also associated with an increase in low-back pain severity and vice versa. However, the SEM models with fixed-effects showed no statistically significant lagged relationships between job strain and any or low-back pain severity (β=-0.05 and β=0.00, respectively). Conclusions This study suggests that associations between job strain and low-back pain with a lag of years may be due to residual confounding by time invariant characteristics. Further studies are, however, needed to elucidate short-term relationships.