The questionnaire was developed on the results of a literature review 12 . Previous literature (in English and German) was analysed, and relevant risk factors and confounding factors were included in the questionnaire. Hence the questionnaire contained questions about sociodemographic factors, relevant diseases, occupational history, and leisure-time activities. Participants were asked to describe every occupation, every sport, and every other leisure-time activity, and they were asked to indicate the respective duration (in years) and also the number of hours per day and per week. In the work analysis, the amount of different body postures (sitting, standing, walking, kneeling/squatting) as well as the prevalence of certain job characteristics (for example, climbing stairs, jumping, lifting/carrying of loads, time pressure) was assessed.

The telephone interview contained detailed questions on the frequency and duration of lifting and carrying for every occupational employment. This interview was conducted if daily lifting or carrying of loads was mentioned in the questionnaire by cases or controls in order to obtain more detailed information about the individual's work tasks.

The patients' history and the physicians' findings were documented in a patient record including information on general health status, as well as the condition of knee cartilage, meniscus, and ligaments (according to the International Cartilage Repair Society standard). This patient record was filled out by the orthopaedic surgeon treating the patient (cases only).

The inclusion criteria were as follows: age between 25 and 75 years, place of residence in the defined vicinity of the participating hospitals, and linguistic and cognitive ability to understand and fill out the questionnaire and to provide informed consent.

The case group's additional criteria were as follows: knee OA confirmed by either radiological diagnostics (≥grade II on the Kellgren and Lawrence scale 13 ) or findings from arthroscopy or open surgery (≥grade III on the Outerbridge scale 14 ). Further criteria for inclusion were: diagnosis of knee OA for no longer than 10 years; no previous fractures involving knee joints or injuries of the knee (ligament or cartilage injuries); and no inflammatory or reactive knee joint illnesses.

The control group's additional criteria were as follows: treatment for an accident due to an external cause (that is, not due to circulatory, metabolic, or neurological disorders), an accident that was not work-related, and no already existing physician diagnosis of knee OA.

In the present study, an OR of 2.5 (95% CI, 1.4 to 4.7) for accumulated kneeling and squatting >8,934 hours over life in women was calculated. In men, the OR for kneeling/squatting for 3,474 to 12,244 hours over life was 2.2 (95% CI, 1.2 to 3.8), and the OR for kneeling/squatting for >12,244 hours over life was 2.5 (95% CI, 1.4 to 4.3). These results indicate an effect of kneeling/squatting on the occurrence of symptomatic knee OA in both genders.

In 2005 Jensen calculated an individual exposure from the amount of knee-straining activities and the number of years in the trade within a collective of floor layers, carpenters and compositors. The ORs for knee complaints and radiographically determined knee OA were 3.0 (95% CI, 0.5 to 17.2) in the low-exposure group, 4.2 (95% CI, 0.6 to 27.6) in the medium-exposure group, and 4.9 (95% CI, 1.1 to 21.9) in the high-exposure group compared with the zero-exposure group 18 . D'Souza and colleagues reported on an analysis of the US national survey (Third National Health and Nutrition Examination Survey (NHANES III)) and used ergonomists' ratings of job categories to describe relationships between work activities and symptomatic knee OA 19 . A significant exposure-response relationship was found between symptomatic knee OA and kneeling in men but not in women. Within a German case-control study, the OR of having radiographically confirmed knee OA was 2.4 (95% CI, 1.1 to 5.0) within the group with cumulative exposure to kneeling and squatting >10,800 hours compared with unexposed subjects 20 .

To our knowledge, only one study investigating the dose-response relationship of cumulative kneeling or squatting and knee OA found no correlation 21 . In this study, however, the daily exposures of kneeling and squatting were asked dichotomously (>1 hour/day or ≤1 hour/day) and then multiplied by exposure years, so these results might be imprecise.

In sum, our results support the presumptions that there is a dose-response relationship between knee-straining work activities and symptomatic knee OA, and that this relationship exists also in women.

In the present study, an OR of 2.1 (95% CI, 1.1 to 4.0) could be derived in women for lifting and carrying of least 1,088 tons over life. This correlation was not significant in men.

In the study by D'Souza and colleagues mentioned above, a significant trend in heavy lifting and severe symptomatic knee OA was detected in both genders 19 . Coggon and colleagues calculated an OR of 1.7 (95% CI, 1.2 to 2.6) for regular lifting and carrying of loads >25 kg (men and women considered in common) 21 . In the study by Seidler and colleagues, lifting and carrying of loads was significantly associated with knee osteoarthritis 20 . The dose-response relationship between lifting and carrying of loads and knee OA was described with an OR of 2.0 (95% CI, 1.1 to 3.6) in the exposure group of 630 to <5,120 kg-hours over life, up to an OR of 2.6 (95% CI, 1.1 to 6.1) in the highest exposure group (>37,000 kg-hours over life) in men. Jensen also investigated the correlation between knee OA and lifting and carrying of loads in her review 8 . She concluded that there is moderate evidence of a dose-response relationship between the lifting and carrying of loads and knee OA.

Our results support the current position that there is moderate evidence of a dose-response relationship between the lifting and carrying of loads and symptomatic knee OA.

In the present study, neither in men nor in women could a correlation between jumping or climbing stairs and symptomatic knee OA be described. McAlindon and colleagues examined a subset of the Framingham Heart Study cohort 22 . They also did not detect effects of climbing stairs. Mounach and colleagues reported in their case-control study that climbing stairs >50 steps/day was associated with a decreased risk of knee OA (OR, 0.5; 95% CI, 0.3 to 0.9) 23 . In contrast, Cooper and colleagues reported an increased OR in people climbing >10 flights of stairs per day (OR 2.7, 95% CI, 1.2 to 6.1) 24 . Sandmark and colleagues described an increased OR (OR, 2.7; 95% CI, 1.7 to 4.1) for jumping in men, but not in women 25 . In the same study, a slightly increased but predominantly nonsignificant OR was described for climbing stairs in both genders. Manninen and colleagues referred to their results of a case-control study wherein climbing already at a medium level of exposure was associated with an increased risk of knee OA among men (OR 3.1; 95% CI, 1.3 to 7.5) 26 . Although in laboratory analyses Sahlström and colleagues identified that jumping down or climbing stairs and ladders revealed a significant increase in movement in the knee compared with normal walking 27 , the effect of these exposures on the knee cartilage remains unclear. Our results could not support either of these effects.

In the present study, a correlation between symptomatic knee OA and further work factors (piece-work, time pressure, hand-arm or whole-body vibration, manual handling of heavy tools, working in wetness, coldness, or heat) could not be found. Elsner and colleagues described significant associations between knee OA and some of the work factors just mentioned 28 . In men, hand-arm vibration (OR, 2.8; 95% CI, 1.2 to 6.4) as well as working under wet/cold conditions and/or draught (OR, 2.0; 95% CI, 1.2 to 3.8) were associated with knee OA, but not in women. In women, manual handling of heavy tools (OR, 6.1; 95% CI, 2.0 to 20.1) was associated with knee OA, but not in men. Sandmark and colleagues described a slightly increased but nonsignificant OR for vibration in men, but no effect in women 25 . To conclude, there seems to be low evidence for the effect of additional working factors on the knee, but few studies dealing with these topics are available. Our results do not support the results of Sandmark and colleagues 25 and of Elsner and colleagues 28 .

Of all the factors observed in the present study, the increase of the BMI correlated strongest in both genders. As stated above, compared with those with normal body weight, an OR up to 12.6 (95% CI, 4.4 to 35.9) in men with obesity grade II or obesity grade III and up to 17.6 (95% CI, 4.5 to 69.2) in women with obesity grade III was calculated. These findings are in compliance with common literature that describes obesity as a major risk factor in the occurrence of symptomatic knee OA.

Anderson and Felson calculated an OR for overweight (OR, 1.7; 95% CI, 1.1 to 2.8), for obesity grade I (OR, 4.8; 95% CI, 2.8 to 8.3), and for obesity grade II + III (OR, 4.5; 95% CI, 1.8 to 11.2) compared with normal body weight in men 29 . In women, the OR was also calculated for overweight (OR, 1.9; 95% CI, 1.2 to 2.9), for obesity grade I (OR, 3.9; 95% CI, 2.6 to 5.7), and for obesity grade II and obesity grade III (OR, 7.4; 95% CI, 5.2 to 10.5), compared with woman with normal weight.

A recent longitudinal study shows that, compared with subjects with a normal BMI, those who were obese (BMI 30 to <35 kg/m²) or very obese (BMI ≥35 kg/m²) were at an increased risk of incident knee OA (relative risk, 2.4 and 3.2, respectively; P for trend <0.001) 30 . Among others, the relevance of BMI was confirmed by Cooper and colleagues (OR, 3.3; 95% CI, 1.6 to 6.9 for BMI ≥25 kg/m² compared with those with BMI <25 kg/m² among both genders) 31 , by Dawson and colleagues (OR, 36.4; 95% CI, 3.1 to 432.0 for BMI ≥25 kg/m² compared with those with BMI <25 kg/m² among both genders) 32 , and by Liu and colleagues (OR, 10.5; 95% CI, 7.9 to 14.1 for BMI ≥25 kg/m² compared with those with BMI <25 kg/m² among both genders) 33 .

Hartmann and Seidel examined data from male construction workers 34 . They calculated the OR for overweight (OR, 1.2; 95% CI, 1.1 to 1.3), for obesity grade I (OR, 1.5; 95% CI, 1.3 to 1.7), for obesity grade II (OR, 1.6; 95% CI, 1.2 to 2.1), and for obesity grade III (OR, 1.8; 95% CI, 1.0 to 3.0) compared with men of normal weight. Liu and colleagues further reported that about 69% of the knee joint replacements in their study sample were to be assigned to overweight causes 33 .

According to the results of Wang and colleagues 35 , the risk of primary knee and hip joint replacement due to OA relates to both adipose mass and central adiposity. This relationship suggests that both biomechanical and metabolic mechanisms associated with obesity contribute to the risk of joint replacement, with stronger evidence at the knee rather than at the hip.

Our results support these existing results. We could clearly find a strong correlation between increasing BMI and symptomatic knee OA.

In the data from the present study, the existence of malalignment of the tibiofemoral joint was associated with symptomatic knee OA in women only (OR, 11.5, 95% CI, 4.7 to 28.7 compared with women without malalignment of the knee).

Malalignment of the knee has rarely found consideration in the relevant epidemiologic literature 11 . Schouten and colleagues published their results of a 12-year follow-up study in 1992 36 . Besides other factors, previous malalignment of the tibiofemoral joint (OR, 5.1; 95% CI, 1.1 to 23.1 compared with people without malalignments) was determined as a prognostic factor for development of knee OA. Greinemann wrote in his 1983 study among mine foremen that slight malalignment of the tibiofemoral joint did not promote knee OA 37 . A high position of the patella, however, might be an aggressive prearthritic deformity according to the results of that study. Unfortunately, the position of the patella was not assessed in the present study.

Our results support the findings of the current review by Tanamas and colleagues 11 , in which malalignment of the tibiofemoral joint was found to be an independent risk factor for the progression of symptomatic knee OA.

In both genders, knee OA within parents, brothers, or sisters was a significant predictor for symptomatic knee OA in the investigated person. The OR was 2.2 (95% CI, 1.4 to 3.4) in women and 2.4 (95% CI, 1.4 to 4.0) in men.

Cooper and colleagues described an OR for the heredity of knee OA of 2.7 (95% CI, 1.3 to 5.5) for both genders combined 24 . The influence of genetic factors for the development of knee OA was the focus of the work group of Spector and colleagues in several studies. In 1996 they published a study among female twins in which they prove a genetic effect for knee and hand OA 38 . The intraclass correlation of a radiographic OA score in identical pairs was 0.64 compared with nonidentical pairs (0.38). In 2004 Spector and MacGregor summarised their findings about the influence of genetic factors of OA derived from classic twin studies 6 . They indicated that the influence of genetic factors was 39 to 65% in knee/hand OA, 60% in hip OA, and 70% in spine OA. According to the authors, therefore, about one-half of OA can be explained by genetic factors 6 .

In our study, the assessment of genetic influences was conducted only by the question about knee OA within parents, brothers, or sisters. As described above, the OR clearly increased for both men and women in connection to genetic predisposition. Our results are in compliance with the results of the authors mentioned above.

The factor of smoking (here measured in package-years) was negatively associated with symptomatic knee OA in women (smoking >20 package-years). This phenomenon has been discussed several times in other studies.

Felson and colleagues described this phenomenon in an overview article 39 after they detected the negative association between smoking and knee OA when evaluating two large datasets derived from the First National Health and Nutrition Examination Survey (NHANES I) 29 and from the Framingham Heart Study 2 . It appears that smoking or some unidentified factor correlated with smoking modestly protects against the development of knee OA. As a possible explanation, Gullahorn and colleagues reported that, according to their study results, nicotine upregulates glycosaminoglycan and collagen synthetic activity of articular chondrocytes 40 . The findings about the correlation between smoking and knee OA were summarised by Elloumi and Kallel 41 . They concluded that smoking would have a modest protective effect against the development of OA. This protective effect would be widely supported by the anabolic activity that nicotine carries on the chondrocytes of the articular cartilage. Given the dangers associated with nicotine and smoking, however, one cannot recommend tobacco as a prevention factor for OA 41 . Our results are in compliance with the results of these authors.

In our study when looking at sports with a risk of unapparent knee trauma, cumulative sports (addition of hours over life of all these kinds of sports) was observed to be a relevant factor for symptomatic knee OA. The OR was 2.5 (95% CI, 1.3 to 4.6) in women performing ≥1,440 hours over life and was 2.6 (95% CI, 1.6 to 4.2) within men having performed ≥3,232 hours over life, compared with persons without any sporting activities.

Participation in physical activity is widely accepted to be associated with physical, psychological, and social benefits 42 . In the literature, few studies could be identified that investigated the correlation between sports history and the development of knee OA. In a review by Gross and Marti, the evidence of the correlation between OA and sports was described as moderate 5 . They concluded that very intensive sports exercise can lead to a low-grade increase in the risk for hip and knee OA (ball and strength sports in particular). In very active runners, the risk of OA also increases. The risk of OA at the weight-bearing joints (hip and knee) might be increased by extremely intensive and long-time sports activity, but might not be predominant in the majority of the population engaging in sports, since the amount of sports activity is lower. A recent review 9 detected that some studies had reported an association between physical activity and a risk for knee OA 43 44 45 , and that other studies had shown physical activity may have no effect 46 47 or may even protect the knee joint from degenerative changes 48 49 .

In contrast to our results, Manninen and colleagues referred to a comparative analysis of different kinds of sports and knee OA in a case-control study 50 . The OR for knee arthroplasty decreased to 0.9 (95% CI, 0.3 to 2.6) in men with a low number of cumulative exercise hours and to 0.4 (0.1 to 0.95) in those with a high number of cumulative exercise hours, with a history of no regular physical exercise as the reference. For the women, the corresponding ORs were 0.6 (0.3 to 0.93) and 0.6 (0.3 to 0.98). The authors concluded that recreational physical exercise was associated with a decrease in the risk of knee OA 50 . Urquhart and colleagues concluded that certain types of exercise had different effects on different people 9 . Rather than a uniform approach to the implementation of physical activity, individually tailored exercise programmes were needed to allow exercise to be carried out safely.

On the basis of our results, we support the necessity for further investigations on the relationship between physical exercise and symptomatic knee OA.

The principle strength of the present study lies in the high power of the dataset. Another strength is the extensive anamnesis, incorporating occupational factors as well as individual factors and leisure-time activities. Additionally, compared with other case-control studies in Germany 51 , the response rate is quite high (73.2% in cases and 65.4% in controls). Compared with international studies, however, the response rate is moderate and nonresponse bias may have influenced the results.