The study population included the participants to the EDS 2002 aged between 45 and 74 years, who were either employed or retired at the time of the survey, had participated in all three interviews and answered the questions on low back pain and the two questions on occupational exposure in the self-administered questionnaire. Individuals in employment who had not been working for health reasons for a period of several weeks at the time of the survey were excluded, as were those who had retired before the age of 55. We also excluded those aged 60 years or older who were still employed since age at retirement is rather low in France and those who are still active beyond 60 years represent a selected population.

The self-administered questionnaire included two questions on physical exposure at work: whether the work involved carrying heavy loads, and whether it involved tiring postures. For each of these two occupational strains, active workers indicated if they were currently exposed, if they had been exposed in the past but were no longer exposed, or if they had never been exposed. For retired subjects, the exposure, if any, was exposure during their active working life. A subject was classified as "exposed" to physical occupational strains if he or she had been exposed in the past, or was exposed at the time of the survey, to at least one of the two occupational strains.

Retirement status was assessed during the face-to-face interview.

The prevalence of low back pain for more than 30 days in the previous year was computed separately for men and women, for active and retired, according to 5-year age group and past/present occupational exposure. In this national study, weightings taking into account the study design were available. Both weighted and unweighted prevalences were calculated. Results presented here are weighted prevalences and their 95% confidence intervals, giving estimates for the whole population. Prevalences for ever exposed and never exposed subjects were compared within sex, age and retirement status subgroups. Comparison tests were based on the distribution of the estimates, with an assumption of normality for the weighted prevalences.

Concerning men (Figure 1) below the retirement age (aged 45-59), the prevalence of LBP was significantly higher in all three age groups among those currently or previously exposed to manual material handling and/or tiring postures (n = 1051) compared with those never exposed to these strains (n = 1183), respectively over 20% versus less than 11% (p < 0.0001 for each of the three age groups).

Among retired men (aged 55-74), the differences in LBP prevalence between those previously exposed (n = 748) and those never exposed in the past (n = 599) tended to become smaller with increasing age. No difference was observed for those aged 65 and over and, for the oldest age group (70-74 years) the prevalence was even slightly higher among unexposed men.

The patterns were quite similar for women (Figure 2). Among active women exposed to physical strains at work (n = 741), the prevalence of LBP was around 25% compared to around 15% for those never exposed (n = 1260). These differences, although less pronounced than those observed for men, were also significant in the three age groups (p < 0.02 in the three age groups). For retired women, the differences between those exposed in the past (n = 430) and those never exposed (n = 489) also tended to decrease with age, as indicated in Figure 2. In the 60-64 age group, the prevalence among exposed retired women was nonetheless significantly higher than that for their never exposed counterparts (p < 0.05).

Among the 55-59 years age group comprising both active and retired subjects, those retired more often declared that they suffered from LBP than their active counterparts, except for men classified as 'exposed' who had a similar prevalence whether they were retired or not.

Since this study is cross sectional, the temporal sequence between LBP and exposures is unknown and the possibility of bias must be considered. Individuals who suffered from LBP could have moved to less physically strenuous jobs prior to the survey. This selection effect would have lead to an overestimation of LBP in the 'unexposed' group if only occupational exposure at the time of the survey had been considered. To minimize this bias, active subjects were classified as "exposed" to occupational strains not only if they were exposed at the time of the survey, but also if they had been exposed in the past. It might be that some subjects had never been exposed because they suffered from LBP very early in their life, before the beginning of their work history, but this must be infrequent. A recall bias could however have occurred with those suffering from LBP having overestimated their exposure to occupational risk factors. Differential errors are expected to be limited since the questionnaire covered many areas of health. In addition, in France there is no specific category such as work-related low back pain, except in very special situations. Among older people, non differential errors could also occur.

Exposure to the occupational strains studied here occurs more often in the lowest occupational classes 16 known to have a shorter life expectancy 17. Older retired individuals who had been exposed in the past may therefore have been underrepresented in the present study. However since common LBP is associated with disability rather than with mortality this should not be an important source of bias. In addition, we did not consider people living in nursing home or retirement home but few people at these ages currently live in these situations in France 18. Active workers above 60 years of age were also not studied in order to minimize selection bias.

Exposures were self assessed which is the only option available when the entire occupational history is considered, especially in the general population. The questions used may be considered as not very specific. The level of exposure is not precisely known. However, rather simple questions about various aspects of the demand of physical work perform rather well as to reproducibility and validity in workers 19. We are not aware of comparable data on the assessment of the past occupational exposure of retirees. For working people the frequency of exposure to handling of heavy loads, was consistent with that found in a French national survey on working conditions which took place in 2003 20, although a precise comparison is difficult due to the age structure in our population. As far as we are aware, there is no such information for retirees within the French population.

Among men, missing data for occupational exposure were more frequent for farmers (compared with blue collar workers) and less common among those in management and intermediate occupations. Being older increased the probability of non response to these questions among active subjects; among retired individuals, the only group with a significantly increased frequency of missing information was the oldest. Among women, farmers also answered less often compared to women in other occupations and non response was significantly more frequent only among the oldest active women and increased with age among those retired. There was no difference in LBP status between those who answered these questions and those who did not in either group.

We wanted here to present descriptive data raising questions on the long term effects of occupational exposure on LBP rather than quantify these effects. For that reason, potential confounding factors, such as obesity, were not taken into consideration. Even though the prevalence of such factors could differ between subgroups, adjustments are not expected to modify the main results. Non-occupational physical activities were not considered. There is little information about physical activity after retirement, especially in relation with physical activity at work 212223. In one study, the largest gain in sport score was observed among those who were the least active at work, however it was not possible to conclude to gain or loss in physical activity in general 21. Does the activity performed by retired individuals differ according to the physical demand they experienced at work? And if it is the case, how does it differ? To our knowledge this topic has not been investigated.

Finally, with data from a cross sectional survey, comparisons between age groups might be due to a generation effect. However, this could not explain the main results which were based on comparisons between exposed and unexposed individuals within the same age group.

Our results do tend to indicate that early prevention in occupational field is of importance not only for short term effects, but also for long term effects when workers retire.

In the 55-59 years age group comprising both active and retired individuals, the figures for LBP prevalence are consistent with health playing a role in the decision to retire 24. However, the exposed men in this age group have a high prevalence irrespective of their situation, active or retired.

Previous studies have reported the long term effects of occupational exposure on low back pain. In Sweden low back pain was increased among Post Office pensioners, aged 71 to 75 years, who had been exposed to the manual handling of heavy loads for over twenty years 10. Several types of low back pain have been related to previous biomechanical strains within a sample of the Gazel cohort comprising both older active workers and 'young' pensioners 25. Manual shipyard workers were also found to suffer from musculoskeletal disorders two to three years after retirement, attributable to heavy physical workload during their active life 8. In the latest study, retired office workers were found to suffer from slightly more back symptoms than three years earlier. Locomotor impairment of the lower back was also associated with the duration of work at the coal face in miners retired for at least 10 years 11. In another study carried out in France with a 5 year follow-up of retirees from various occupational settings, the lifetime physical workload was associated with frequency and course of musculoskeletal pain at various sites of the body 9. A significant increase in pain prevalence after five years for some of the 'unexposed' subjects was also observed.

Many previous studies were based on small samples 8911. Other ones considered only young retirees 8, or included active workers 25. In the one dealing with low back pain among older retired subjects, the long term effects were observed only for those with the longest duration of exposure 10. The other studies focused on an outcome somewhat different including locomotion impairment 11 and pain at various sites 9. Our results are globally in accordance with these previous studies, with the advantage of being based on a large population sample.

In France, the legal age of retirement, which was 60 when the survey was performed, is rather low compared with other western countries. Workers close to retirement age were also often out of the labour market due to employment policies or for economic reasons. Hence, generalization of findings from France to other countries might be discussed. However, the fact that the effects of past physical occupational exposures do not disappear with retirement is probably a general result which would be observed also with an older age at retirement. Furthermore, being exposed at older age could also have specific consequences on the lower back. Considering that retiring later implies a longer duration of exposure, at least for a part of the workers, these aspects appear important to consider, especially as public policies are favouring increasing age at retirement.