Article Text

Download PDFPDF

Extended report
Knee extensor strength and body weight in adolescent men and the risk of knee osteoarthritis by middle age
  1. Aleksandra Turkiewicz1,
  2. Simon Timpka2,3,
  3. Jonas Bloch Thorlund4,
  4. Eva Ageberg5,
  5. Martin Englund1,6
  1. 1Department of Clinical Sciences Lund, Orthopedics, Clinical Epidemiology Unit, Faculty of Medicine, Lund University, Lund, Sweden
  2. 2Genetic and Molecular Epidemiology Unit, Lund University Diabetes Center, Lund University, Malmö, Sweden
  3. 3Division of Women’s Health, Department of Medicine, Brigham and Women’s Hospital and Harvard Medical School, Boston, Massachusetts, USA
  4. 4Department of Sports Science and Clinical Biomechanics, University of Southern Denmark, Odense, Denmark
  5. 5Department of Health Sciences, Lund University, Lund, Sweden
  6. 6Clinical Epidemiology Research and Training Unit, Boston University School of Medicine, Boston, Massachusetts, USA
  1. Correspondence to Aleksandra Turkiewicz, Lund University, Faculty of Medicine, Department of Clinical Sciences Lund, Orthopedics, Clinical Epidemiology Unit, Remissgatan 4, Lund 22185, Sweden; aleksandra.turkiewicz{at}


Objectives To assess the extent to which knee extensor strength and weight in adolescence are associated with knee osteoarthritis (OA) by middle age.

Methods We studied a cohort of 40 121 men who at age 18 years in 1969/1970 underwent mandatory conscription in Sweden. We retrieved data on isometric knee extensor strength, weight, height, smoking, alcohol consumption, parental education and adult occupation from Swedish registries. We identified participants diagnosed with knee OA or knee injury from 1987 to 2010 through the National Patient Register. We estimated the HR of knee OA using multivariable-adjusted Cox proportional regression model. To assess the influence of adult knee injury and occupation, we performed a formal mediation analysis.

Results The mean (SD) knee extensor strength was 234 (47) Nm, the mean (SD) weight was 66 (9.3) kg. During 24 years (median) of follow-up starting at the age of 35 years, 2049 persons were diagnosed with knee OA. The adjusted HR (95% CI) of incident knee OA was 1.12 (1.06 to 1.18) for each SD of knee extensor strength and 1.18 (1.15 to 1.21) per 5 kg of body weight. Fifteen per cent of the increase in OA risk due to higher knee extensor strength could be attributed to knee injury and adult occupation.

Conclusion Higher knee extensor strength in adolescent men was associated with increased risk of knee OA by middle age, challenging the current tenet of low muscle strength being a risk factor for OA. We confirmed higher weight to be a strong risk factor for knee OA.

  • knee osteoarthritis
  • epidemiology
  • osteoarthritis

Statistics from


Knee extensor muscle weakness is well recognised in individuals with knee osteoarthritis (OA).1–5 The reported mechanisms behind this association include reduced physical activity due to joint pain leading to muscle atrophy and/or impaired muscle activation.4 However, the underlying mechanisms behind any association between muscle strength and incident knee OA are not fully understood. Knee extensor muscle weakness might precede the development of radiographic or symptomatic knee OA, but results are inconclusive. Segal et al6 reported a decreased risk of symptomatic, but not radiographic, knee OA in Multicenter Osteoarthritis Study participants with higher when compared with lower knee extensor strength. In the same cohort, higher muscle strength was reported to be a protective factor for knee OA when investigating persons with meniscal pathology only.7 Slemenda et al8 reported a weak association between low muscle strength and incident radiographic knee OA, but only in women. In other studies, weak or no association between low knee extensor strength and incident knee OA was found among young persons after anterior cruciate ligament reconstruction,9 10 while in a sample of persons having chronic knee pain, fewer single-leg rises, indicative of weak lower extremity strength, were associated with higher risk for radiographic OA.11 Importantly, in the literature, there is a paucity of studies on the role of lower extremity muscle strength in young individuals for knee OA development. There is also a lack of studies using population-based samples not enriched with pre-existing mild disease or with multiple knee OA risk factors as in most of the studies mentioned above. In addition, the common practice of defining strength exposure variables as a ratio of force by body weight might have resulted in finding spurious associations.12 Furthermore, possible mediators of the association, such as joint injury or occupational factors have not been evaluated. Therefore, the aims of this study were (1) to evaluate to what extent isometric knee extensor strength and body weight in adolescence are associated with the risk of knee OA by middle age in a comprehensive, population-based cohort of Swedish men and (2) to evaluate the role of knee injury and adult occupation behind an association between knee extensor strength and knee OA.


Data sources and study sample

We included 41 886 men, typically aged 18 years, who underwent a mandatory military conscription examination in Sweden between September 1969 and May 1970. The examination was performed at six centres including standardised physical and mental tests as well as evaluations by a physician. At that time, only a severe disability could be a reason for exclusion from the examination. The data were registered in the Swedish Military Service Conscription Register.

For the included men, we retrieved individual level data from the Swedish National Patient Register containing information about every hospitalisation (from 1987 to 2010), 1-day surgery (from 1997 to 2010) and specialist outpatient care visits (from 2001 to 2010). We included the date of visit and the diagnostic codes at the time of visit according to the International Classification of Diseases (ICD) system (ICD-9 up to year 1997 and ICD-10 from year 1998 and onwards). We further required the men to be alive and resident in Sweden by 1 January 1987, at the typical age of 35 years, corresponding to the start of the registration of diagnostic codes in the National Patient Register (figure 1).

Figure 1

Flow chart of the study. OA, osteoarthritis.

In addition, we collected information about the highest level of parental education at the time of conscription and about occupation of the conscripts themselves in adulthood from Statistics Sweden.13

The study was approved by the Regional Ethical Review Board in Lund, Sweden.

Isometric knee extensor strength

Isometric knee extensor strength was measured using a validated protocol as previously described.14 In brief, custom build chairs were used to measure knee extension in a sitting position (90° of hip flexion with the pelvis fixed by a strap to prevent concomitant extensions of the hip) with 90° knee flexion and arms crossed over the chest. Strength was measured using a dynamometer placed at the level of the lateral malleolus. The test was administered by trained personnel and performed at least three times with the highest value being recorded. Knee extensor strength in Newtons, multiplied by estimated shank length in metres (calculated as 0.246 times the body length), was used in the analysis.15


Body weight was measured to the nearest kg, and body height was measured to the nearest cm. Body mass index (BMI) was calculated and WHO thresholds of 18.5 and 25 were used to define underweight and overweight.16 Smoking status was self-reported in categories through questionnaires: non-smoker (reference category), 1–5 cigarettes per day, 6–10 cigarettes per day, 11–20 cigarettes per day and 21 or more cigarettes per day. Alcohol consumption was self-reported as frequency and amount of consumed beer, wine or sprit per week. We recalculated the intake as grams of pure alcohol consumed per week.17 The conscripts’ adult occupation was categorised into (1) farmers or mining or oil industry workers, (2) production workers and machine operators, (3) physically demanding service job workers (police, fire department, hotel, household chores and similar) or (4) other (reference category) based on the ISCO-68 international classification. The parental highest level of education (maximum from mother and father) was categorised into two categories: high school or less (reference category) and higher education.

We defined knee injury as (at least one) diagnostic code of dislocation of knee (ICD-9 code: 836, ICD-10 code: S83), knee contusion (924B, S80.0), fracture of patellae (822, S82.0) or internal derangement of knee (717, M23) during the follow-up period.

Definition of knee OA

We defined incident knee OA as the first record of knee OA registered in inpatient or specialist care between the year 1987 (typical age 35 years) and 2010 (typical age 59 years) with either the ICD-9 code 715 or the ICD-10 code M17. The validity of the diagnostic codes in Sweden has previously been reported to be high.18–20 The positive predictive value of a knee OA diagnosis in a random sample of one of the largest regions in Sweden (Skåne, 1.3 million inhabitants) with respect to American College of Rheumatology clinical and radiographic criteria or radiographic knee OA (equivalent to Kellgren-Lawrence grade 2 or higher) was 88%.19

Statistical analysis

For descriptive purposes, we divided the study sample into four groups using quartiles of knee extensor strength. For our primary analyses, we used Cox proportional hazards regression models. We evaluated the proportional hazards assumption using tests and visual inspection of the zero-slope of Schoenfeld residuals. Follow-up for all included men was from 1 January 1987 (National Patient Register inception) until knee OA diagnosis, death, emigration or 31 December 2010, whichever occurred first (figure 1). The main exposure of interest was knee extensor strength in Nm included as a continuous variable. We performed a crude analysis and an analysis adjusted for pre-specified confounders, that is, parental education status (categorical), smoking (categorical), alcohol consumption (continuous) and body weight and height (both continuous). We evaluated the association between body weight and incident knee OA in a second model, adjusted for all the above covariates apart from knee extensor strength.

The percentage of persons who emigrated or died during the study period was low (9%), and results from an analysis ignoring time to event were essentially the same as from the Cox regression analysis. Thus, for the formal mediation analysis, we used logistic regression where we excluded persons who emigrated. The potential mediators evaluated were knee injury during follow-up and adult occupation. Only knee injuries diagnosed before knee OA were included. We used the method for decomposition of effects in nonlinear probability models developed by Karlson, Holm and Breen,21 as implemented in the Stata command ‘khb’.22 The analysis was adjusted for the same confounders as those included in the Cox model. The statistical analysis was performed using Stata 14, StataCorp 2015 software.

Sensitivity analyses

We performed additional analyses where persons diagnosed with OA (joint not specified), rheumatoid arthritis (RA) or meniscal/cartilage pathology at the time of conscription were excluded.

We repeated our main analysis of the association between muscle strength and knee OA when excluding persons in the lowest or highest percentile of weight (ie, we included only those with weight between 49 and 95 kg to avoid potential bias due to non-positivity). Third, to evaluate the impact of our knee OA definition on the results, we expanded the definition to also include diagnoses of knee OA registered within primary care. These data were retrieved from regional healthcare registries (years 2000–2010) of two large regions in Sweden (Skåne and Västra Götaland) for a subset of 10 391 persons and yielded 24% more persons with knee OA. However, the date of diagnosis was not available and, thus, a logistic regression model was used for this analysis.

Finally, to evaluate if the results may reflect the propensity of individuals to seek healthcare, we repeated the analysis with hip OA as the outcome, as we expected not to find an association between knee extensor strength and hip OA.


Four per cent of persons were excluded from the analysis sample due to no exposure measurement, emigration or death before the follow-up start in 1987 (figure 1). These persons were on average heavier, consumed more alcohol and were more likely to have at least one parent with higher education. The mean (SD) knee extensor strength was 234 (47) Nm, and the mean (SD) body weight was 66 kg (9.3 kg). Their mean (SD) BMI was 21.0 (2.6), whereof 14% and 7% were underweight and overweight, respectively, according to the BMI thresholds of 18.5 and 25. The most common musculoskeletal disease diagnosed at the conscription examination was flat foot causing no or light reduction in function (in 7% of the men). Less than 1% of the men were diagnosed with OA, RA or meniscal/cartilage pathology at the conscription examination (table 1).

Table 1

Descriptive characteristics of the study sample divided into groups by quartiles of knee extensor strength

The mean (SD) follow-up time was 22.8 (3.8) years with a maximum of 25 years. Among the 40 117 included men, we identified 2049 incident knee OA cases, yielding an incidence rate of 22.4 (95% CI 21.5 to 23.4) per 10 000 person-years and an incidence proportion of 5% over 25 years (table 2). The mean age at the time of knee OA diagnosis was 53 years.

Table 2

Incidence of knee osteoarthritis (OA) by knee extensor strength quartiles

The crude HR for incident knee OA per 47 Nm (1 SD) of knee extensor strength was 1.26 (95% CI 1.20 to 1.31). After adjustment for body weight and height, the HR was 1.11 (95% CI 1.06 to 1.16). After additional adjustment for smoking, alcohol consumption and parental education, the results remained essentially the same with an HR 1.12 (95% CI 1.06 to 1.17) (figure 2). Higher body weight was associated with a higher risk of knee OA, with an HR of 1.18 (95% CI 1.15 to 1.21) per 5 kg body weight.

Figure 2

Survival function, estimated from the Cox regression model, at quartiles of quadriceps strength, adjusted for weight, height, parents’ education, smoking and alcohol consumption. OA, osteoarthritis.

Investigating the mediating role of knee injury during follow-up on the association between knee extensor strength and knee OA, the ratio of total effect of knee extensor strength to direct effect was 1.13. In other words, 13% of the total effect of knee extensor strength on knee OA was mediated by knee injury. Similarly, in the mediation analysis of adult occupation, the ratio of total effect to direct effect of knee extensor strength on knee OA was 1.02. Assessing both mediators in one model yielded similar results, where 15% of the effect of knee extensor strength was mediated by knee injury and occupation. On the absolute scale, the probability of having incident knee OA in this study sample was higher by 0.48 (95% CI 0.24 to 0.71) percentage points for a 1 SD of knee extensor strength. This was reduced to 0.42 (95% CI 0.19 to 0.66) percentage points after controlling for adult knee injury.

In the sensitivity analysis excluding 267 persons with OA, RA or meniscal/cartilage pathology diagnosed at the conscription examination, results remained similar (HR of knee OA per 1 SD of knee extensor strength was 1.11, 95% CI 1.06 to 1.17). Exclusion of persons with extreme weight had no essential impact on the results, HR of 1.11 (95% CI 1.06 to 1.17). When including knee OA diagnosed within primary care (in a subset of 10 391 persons), results were similar with an OR of 1.09 (95% CI 0.99 to 1.19). The association of knee extensor strength with hip OA adjusted for weight, height, parents’ education, smoking and alcohol consumption was 1.04 (95% CI 0.97 to 1.12).


In this population-based study of young men, we found that higher knee extensor strength in adolescence was associated with an increased risk of incident knee OA by middle age. In absolute terms, the risk of incident knee OA over the follow-up time increased with 0.5% per 47 Nm higher knee extensor strength. Furthermore, 15% of this effect was estimated to be mediated by adult knee injury and adult occupation. Although the increase in risk was small, these results challenge the current belief of reduced muscle strength as a risk factor for knee OA.

Knee extensor weakness is well documented in individuals with knee OA and may often be secondary due to knee pain and/or reduced physical activity level.3 However, knee extensor weakness has also been suggested to be an independent risk factor for the development of knee OA.23 These results have been fairly consistent irrespective of the knee OA definition used, as recently reported in a systematic review.23 However, these previous studies consisted of mainly middle-aged or elderly persons in cohorts enriched with risk factors for knee OA including frequent knee pain. In contrast, our study was population-based and included virtually all Swedish young men who were typically aged 18 years at the time of conscription, 35 years at the start of follow-up and 59 years at the end of follow-up. Thus, our study is less prone to sample selection bias than previous studies.6 11 24–26 Furthermore, in a previous cohort study of persons without musculoskeletal problems at baseline, the authors reported that higher knee extensor strength divided by body weight was associated with lower risk of knee OA in women. Their analysis models were adjusted also for BMI (kg/m2),27 which implies a complicated function of body weight (with both weight and the inverse of weight included in the model). This makes interpretation of their findings and comparisons with other studies challenging. Furthermore, common practice of dividing muscle strength by body weight may lead to finding a protective effect of strength through inadequate control for the effect of body weight.12 In our study, the relationship between knee extensor strength and risk of knee OA was roughly linear and persisted also after adjustment for several confounders and for body weight. We found body weight to be the strongest risk factor for knee OA, which is in accordance with previous evidence in younger adults.28 29 Importantly, the absolute increase in risk due to higher knee extensor strength was rather small (~0.5% per one SD of knee extensor strength) and, thus, our results should not be interpreted as a discouragement of strengthening excises or physical activity in general.

The possible mechanisms behind an association between poor muscle strength and development of knee OA have mostly been hypothesised to be due to a protective role of strong muscles acting as shock absorbers and stabilisers of the joint. The interplay of muscle strength and other risk factors may be different in young adults when compared with middle-aged and elderly persons who may have several risk factors in addition to muscle weakness, that is, overweight or obesity. In a formal mediation analysis in the present work, we estimated the role of knee injury and adult occupation in the association between knee extensor strength and knee OA. Knee injury is the strongest known risk factor for the development of knee OA.30 In our study, 13% of the increase in risk of knee OA associated with higher knee extensor strength was attributed to the role of knee injury. A limitation is that we were not able to identify knee injuries in young adulthood (before the age of 35 years), when such acute injuries often occur, or knee injuries that did not lead to a healthcare visit. Thus, our result is most likely an underestimation of the true role of knee injury behind the observed association between knee extensor strength in adolescence and knee OA by middle age. This reinforces the need for the prevention of joint injuries in young physically active individuals.31

There are some limitations related to the data sources that need to be acknowledged. First, we did not have any measurement of muscle strength during the follow-up. One hypothesis, still to be evaluated, may be that active persons with high muscle strength in adolescence, who later become less active and lose muscle strength, may be those at high risk of knee OA. Also, only knee OA diagnosed within inpatient or outpatient specialist care was included. Thus, persons with knee OA who have not sought healthcare, for example, those perceiving mild symptoms without substantial limitations in daily life, have not been captured. However, including knee OA diagnosed in primary care in a subset of our study sample yielded similar results as our main analysis. One could further speculate that our results reflect a higher propensity for healthcare seeking in persons with higher muscle strength, for example, due to higher body weight and overweight/obesity-related issues. This, however, is unlikely to have influenced our results, as we did not find a similar association between knee extensor strength and incident hip OA. Similar constraints apply for possible misclassification of knee injury. Finally, only men were included in our study sample and, thus, we cannot draw any conclusions about women. Also, in the 1970s,  many countries had compulsory military training and were in this respect similar to Sweden. Today, however, only a minority of countries have compulsory military service; thus, the generalisability to new generations of men may be uncertain.


In this large population-based sample of young men, we found evidence of increased risk of incident knee OA during middle age when having higher knee extensor strength or higher body weight, even after taking into account the role of adult knee injury and occupation, as well as several other confounders based on individual-level data. We could not confirm the previously reported increase in risk of knee OA in persons with knee extensor weakness. However, the vast majority of previous studies included older participants selected on the basis of having multiple risk factors for knee OA and, thus, are not directly comparable with our study sample. Thus, our results indicate that low muscle strength in young adulthood may not be a risk factor for the development of knee OA in men, while high body weight is.


We would like to acknowledge the personnel involved in the collection of the conscription data and all the participating men.



  • Contributors Conception and design of the study: ME and AT. Data collection: ME and ST. Statistical analysis: AT. Data interpretation: all authors. Drafting of the manuscript: AT. Critical revision of manuscript and acceptance of the final version: all authors.

  • Funding We would like to acknowledge the support from the Swedish Research Council, Kock Foundations, Österlund Foundation, Crafoord Foundations, the Swedish Rheumatism Association, the Faculty of Medicine Lund University and Region Skåne.

  • Competing interests None declared.

  • Ethics approval The Regional Ethical Review Board in Lund, Sweden.

  • Provenance and peer review Not commissioned; externally peer reviewed.

  • Data sharing statement The data are not available for sharing because of privacy and integrity issues protected by the Swedish law.

Request Permissions

If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.