The purpose of this study was to explore whether occupation is associated with total knee or hip replacement for OA in Icelandic men and women.
Male farmers differed significantly from other work classes, having a greatly increased likelihood for both TKR and THR for OA. In the results presented in Table , we used managers and professionals as the reference group since we deemed that these had the lightest physical workload. It is possible to motivate the choice of other work classes as a reference group so we tested this model with each of the different work classes as a reference group. When exploring the association with TKR, we found a significant OR in some cases for other occupations than male farmers, but only the farmers differed consistently from the other work classes. For THR the results were similar. Male fishermen had a significant OR for TKR when using managers and professionals (OR 3.3, 95% CI 1.3 to 8.4) or operators and unskilled labourers (OR 2.4, 95% CI 1.1 to 5.2) as a reference group. Farmers had a consistently and significantly high OR regardless of reference group, except when using fishermen as a reference group. This suggests that male fishermen are associated with having TKR.
For women we found no differences between the work classes for TKR or THR. In this generation there was a gender difference in physical workload at the farm. The male farmer would be mostly responsible for the heavier work and the women would have a lighter workload, so the physical workload of the male farmer is not comparable to the workload of the female farmer.
There was no significant difference in mean age at onset of symptoms within the different work classes. If such a difference had existed it might lead to an overrepresentation of the work class that had lower mean age at symptom start.
We found a very strong inheritance for the farming occupation. Since few parent-child relationships were found within the cohort after individuals younger than 60 years were excluded, we calculated the odds ratio for how occupation is passed on from parent to child for the cohort prior to age exclusion. This confirmed that the farming profession is to a great extent passed on from parents to their children. Of the father-son relationships that we found in our cohort prior to exclusion due to age, 83% of farmers were sons of farmers. This might in some part explain the association between TKR/THR and farming. One possible explanation is that children, especially the sons, who were raised on farms, participated in heavy manual labour at a younger age than urban children would, and this might increase the prevalence of OA in this group. There is extensive familial clustering of OA in Iceland and OA cases can be traced to a relatively low number of founders (ancestors) compared to controls [33
]. It is not unlikely that by chance these founders were farmers. Taking into account the strong inheritance of the farming profession, this might lead to an enrichment of genes associated with OA amongst the farmers of today. Because OA usually does not become evident until later age, it is not possible to truly determine the case or control status of the younger individuals. This limited our ability to establish the relative contribution of occupation and heredity for OA. A continued follow-up of this cohort may clarify this.
Previous studies on this topic have limitations, as does the present study. Some studies have been limited to a certain geographic area [8
] and a selection bias might occur if persons in certain occupations had moved to another area after receiving their diagnosis, for example, because they needed to change occupation or for better health care. Our study was based on the entire population in Iceland and data were from all hospitals in the country.
In several previous investigations workload was defined from certain tasks [5
] (for example, kneeling, squatting) rather than the occupational title. Participants in these studies were asked to grade how much they lifted, kneeled, squatted, and so on, during different periods of their life. In a study [34
] that compared the correlation between self-reported and observed tasks at work, the coefficient of determination (r2
) was as low as 0.15, meaning that in some of the cases 61% of study participants were not reporting correctly. If recall bias is added to this, the correlation is probably poorer. We chose to base our classification on job title which should be less prone to recall bias. However, our approach does not take into account the differences between different occupations that are classified into the same group or the differences that can be present within a given job title. Farming in Iceland has been very homogenous, all being single family farms with cattle, sheep or a mixture of both. Fishermen can also be considered a homogenous work class. Other work groups presented here are more heterogeneous as we had several subclasses combined in, for example, service and shop workers.
The definition of OA varies between different studies. Some studies define OA by radiological findings, with or without clinical symptoms, while others use joint replacement as definition. OA varies in its severity and radiological findings have poor correlation to the clinical presentation [18
]. Total joint replacement is generally done for patients with severe OA symptoms not managed satisfactorily by other interventions. In contrast to, for example, a definition based on radiographs only, a case definition based on joint replacement represents a significant disease burden. A limitation of this case definition is, however, the multiple influences beyond symptoms on the patient and health professional decision of joint replacement [35
]. A further limitation with our choice of case definition was that we surely had some false negatives in our control group, that is, some of the controls might after the end of the study develop OA that requires joint replacement. This would lead to a bias towards the null. To try to minimise this we chose to include only individuals that were 60 years of age or older at study entry. Because we used joint replacement as a definition for our cases, persons that were deemed too ill to have joint replacement are also classified as controls, even though their disease severity motivates that they should be classified as cases. This healthy patient selection bias is also towards the null.
An additional confounder is that persons with physically demanding occupations could be more at risk for joint trauma and develop their OA at a younger age [9
]. They might also experience greater problems with performing their work after the onset of OA and thus seek help earlier and be overrepresented as cases. We found no statistical difference in the age of onset of symptoms between the different work classes so this does not seem to be a problem in the current study. However, we cannot exclude the possibility of a healthy worker effect where only the most healthy survive within the trade (that is, young individuals with, for example, hip pain would be forced into another trade than farming or fishing). This would also lead to bias toward the null. It has been suggested that farmers are less willing to seek healthcare for musculoskeletal problems [37
], which would also lead to a bias toward the null.
Our analysis of recreational physical activity was limited by the fact that 25% of, the participants did not answer this question. No other question in our questionnaire had such a high percentage of missing values. We suggest that the most probable reason is that it was not clearly enough stated that those that had not been active in any recreation physical activity should state that fact. We nevertheless tested different models to compensate for our missing values. None had any significant impact. It is not surprising from the biological standpoint that the effects of recreational activity a few hours per week should be much less than the effects of the occupation eight hours per day, five days per week and in some cases (for example, farmers) much more than that.
Changes in access to joint replacements over time might be a possible confounder. More than 90% of our controls were born after 1920 and joint replacement had become readily available in Iceland after 1985, so we do not believe that many of our controls were in fact individuals in need of arthroplasty that were unable to have an operation due to poor availability.
The controls were not drawn from the source population which the cases were derived from, but were relatives of the cases. This was dictated by the fact that the current study is a part of a larger project studying the effect of genes and environment on the risk of OA in the Icelandic population. Any bias introduced by the control selection (first degree relatives), would likely be bias towards the null because it is more likely to find the same profession also within immediate family members than from the background population, that is, yielding an increased frequency of, for example, farmers or fishermen also in the control sample.
The association with hip osteoarthritis amongst male farmers has been established in several studies [7
], including one prospective study [23
]. One previously published study suggested an association between knee OA and farming [12
], while another found no such association [38
], although that study has not been ranked as being of high quality. To the best of our knowledge this is the first population-based study to suggest an association with both TKR and THR and farming in the same sample. Few studies have been done on female farmers.
The evidence for other professions is less compelling [11
]. We found no significant association for other professions, other than an increased likelihood for TKR due to knee OA for fishermen, when compared to more sedentary occupations. We found that for male service and shop workers with THR and male craft workers with TKR, the OR were borderline significant and in a larger sample these associations might reach statistical significance. For women, no OR was statistically significant, but for THR they were all less than 1. This might indicate that the reference group (managers and professionals) has an increased risk for THR in women.