Occupational Strain as a Risk for Hip Osteoarthritis

Abstract

Background

Multiple epidemiological studies have revealed an association between occupational physical strain and the risk of developing hip osteoarthritis.

Methods

To determine the association between the lifting and carrying of heavy loads or other physically demanding work and the risk of hip osteoarthritis (HOA) or total hip replacement (THR), we systematically searched the literature for primary studies on the effects of exposure to physical strain and meta-analytically reviewed the results that were amenable to comparisons across studies. We separately assessed studies that had hip pain as an endpoint.

Results

5 cohort studies and 18 case–control studies were found suitable for inclusion. The lifting of heavy loads increases the risk of HOA or THR: exposure doubles the risk in men (relative risk [RR] 2.09, 95% confidence interval [1.4; 3.1]) and increases it by roughly 40% in women (RR 1.41 [1.0; 1.9]). Physically demanding work consisting of a combination of activities of various kinds (dealing with heavy loads, heavy manual work, or prolonged walking and standing) increases the risk by roughly 150% in men (RR 2.46 [1.3; 4.8]) and 40% in women (RR 1.38 [0.9; 2.2]). Hip pain was also reported more commonly in the exposed groups.

Conclusion

The studies are moderately to highly heterogeneous. An association exists between years of lifting heavy loads or other kinds of physical strain on the job and the risk of developing osteoarthritis of the hip. The greater the exposure, the greater the risk. The evidence base for risk assessment in women is currently inadequate.

Osteoarthritis of the hip (HOA) can cause severe pain and impairs quality of life. The burden on the healthcare system is considerable, and HOA has a negative effect on employment (1). While there are no robust data on the prevalence of HOA specific to Germany, a representative study (e1) found a lifetime prevalence of 2% for both men and women in Europe as a whole. International epidemiological studies show that the prevalence of HOA in women is 1 to 3% in the age group 45 to 75 years—depending on region and diagnostic criteria—and rises steeply thereafter. In men up to 75 years of age the reported prevalence of HOA varies between 2.5 and 8.5% (2– 4). Alongside predisposing factors, the risk is increased by hip defects, age, overweight, and injury (5, 6).

The guidelines stipulate diagnosis of HOA, also known as coxarthrosis, on the basis of the triad of joint pain, impaired mobility, and radiological findings. Internationally, the most commonly used systems are the American College of Rheumatology criteria (7), the Kellgren and Lawrence radiological score (8), and function-related scoring systems such as the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (9) and the Harris Hip Score (10). Radiological signs of arthrosis, e.g., narrowing of the joint space, do not always go hand in hand with pain and functional impairment. Only around 25 to 45% of cases of hip pain are attributable to HOA (11, 12). The overall importance of HOA will increase with the aging of the population (13).

Aim of study

The published reviews (14, 15) conclude that there is a basic link between occupations involving physically demanding work and the development of HOA. Owing to the nature of the searches and the restriction to certain types of study design, the numbers of studies included in these reviews were relatively small. Some of the searches were carried out more than 10 years ago. Meta-analyses and statements on the dose–response relationship are conspicuous by their absence. Therefore, we decided to carry out a new systematic survey of the literature with an expanded search string and including all types of studies. Study quality was assessed in uniform fashion and the data for men and women were analyzed separately. Our aim was to identify ways of preventing occupational HOA.

Methods

A systematic survey of the databases Medline, CENTRAL, Embase, and HSE-Line was carried out in March 2015. The search strategy was based on a combination of MeSH terms and text words as recommended by Mattioli in 2010 (16). We also conducted a manual search of the tables of contents of occupational medicine journals, the abstracts of international orthopedic conferences, and the references of all studies selected for inclusion. The inclusion criteria are listed in Box 1. At the beginning of the study the protocol was registered in PROSPERO (registration number CRD42015016894).

BOX 1. Inclusion criteria.

• Population

  • Adult men and women, currently or previously employed

• Exposure

  • Precise data on lifting of loads and physically demanding tasks

• Endpoints

  • Osteoarthritis of the hip on clinical or radiological examination

  • Implantation of a hip joint prosthesis (total hip replacement)

  • Also: Self-reported hip pain or impairment of mobility

• Study design

  • Cohort studies

  • Case–control studies

  • Cross-sectional studies, provided task or occupation performed for at least 10 years

  • Studies published in German or English from 1990 onward

Study selection

Two independent reviewers scrutinized and evaluated titles and abstracts (AB, AF, DK, MF), extracted data, and assessed the quality of the included studies (AB, DK, SU, UBA). In the case of disagreement, a third reviewer (AS) was consulted.

Quality assessment

The quality criteria were developed separately for cohort studies and case–control studies on the basis of the Newcastle–Ottawa Quality Assessment Scale (17) and the Cochrane Handbook (18). The assessment categories are shown in Box 2. A detailed description of the quality criteria can be found in eTables 1 and 2.

BOX 2. Quality criteria.

• Cohort studies (max. score 19 points)

  • Selection (max. 5 points) Representativeness of cohorts, selection of nonexposed probands, baseline participation rate, osteoarthritis of hip excluded at baseline

  • Exposure (max. 4 points) Validity and accuracy of data acquisition, same method for exposed and nonexposed probands

  • Comparability, confounding (max. 4 points) Adjustment for age, sex, body mass index; full report of hypotheses, risk estimators separate for men and women

  • Diagnosis of hip osteoarthritis (max. 6 points) Validity of diagnosis, blinding, duration and quality of follow-up

• Case–control studies (max. 15 points)

  • Selection of cases and controls (max. 6 points) Validity of diagnosis, representativeness of cases and controls, response rate

  • Comparability, confounding (max. 4 points) Adjustment for age, sex, body mass index; full report of hypotheses, risk estimators separate for men and women

  • Exposure (max. 5 points) Validity and accuracy of data acquisition, blinding to diagnosis, same method for cases and controls

eTable 1. Quality criteria for cohort studies.

I. Selection (max. 5 stars)	Assessment
1. Representativeness of the cohorts
Representative of the occupational group concerned or of the general population. whichever is applicable	*
Selected group
No or inadequate description of cohort derivation
2. Selection of members of the groups with no or low exposure
Drawn from the same population segment as the exposed group; one star awarded for representativeness of cohort	*
Drawn from the same population segment as the exposed group. but no star awarded for representativeness of cohort
Drawn from another source
No or inadequate description of derivation of nonexposed cohort
3. Participation rate at baseline
Response rate >80% or participation rate 50%; nonresponder analysis shows comparability of responders and nonresponders	**
Response rate ≥ 50% to 80% without nonresponder analysis	*
Response rate < 50%
Not reported
4. Baseline: probands with hip osteoarthritis or hip pain
Not included in study or excluded from analysis	*
Data not acquired or not considered in analysis
II. Exposure (max. 4 stars)
1. Validity and reproducibility of data acquisition (methods)
Reliable. reproducible data acquisition (e.g.. measurement at workplace. valid documentation)	**
Structured interview/questionnaire or registry data	*
Written self-report
No or inadequate description of data acquisition
2. Same method of data acquisition for exposed and nonexposed subjects
Yes	*
No
3. Accuracy of exposure data
Quantitative data on exposure (frequency and/or duration. e.g.. weight of load. hours driving a tractor)	*
Only occupation or activity specified. without information on concrete tasks such as those mentioned in the line above
III. Comparability of exposure groups. analysis (max. 4 stars)
1. Comparability of cohorts on basis of designs or analysis (adjustment. subgroup analysis. or other suitable methods. e.g.. stratification or matching)
Analysis after adjustment, matching, or stratification for age, sex, and body mass index or other important confounders (with a coherent concept, i.e., the second star for adjustment for other important confounders was awarded only if the adjustment is sufficiently justified, e.g., by description as confounders or change in estimate)	**
Analysis after adjustment. matching. or stratification for age and sex (if relevant)	*
No adjustment. no matching. no stratification
2. Have all the planned comparative hypotheses mentioned in the methods section been reported with quantifiable data?
Yes	*
No (e.g.. only significant differences are reported. or only p-values are given)
3. Have effect estimators been calculated separately for men and women?
Yes. or only one sex was investigated	*
No
IV. Endpoint (max. 6 stars)
1. Validity and reproducibility of the diagnosis hip osteoarthritis
If one of the following conditions is met, two stars are awarded: diagnosis by statement of hip pain + clinical examination with measurement of mobility/function + radiological imaging with clearly defined criteria as stipulated in guidelines (7, 8; measurement of joint space width), or implantation of a hip prosthesis (total hip replacement, THR), or on waiting list for THR	**
If one of the following conditions is met, one star is awarded: diagnosis from registry data or measurement of mobility/function by means of a complex scale with subscales for pain assessment, mobility, and function—e.g., Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), Harris Hip Score (HHS), Hip Disability and Osteoarthritis Outcome Score (HOOS)—or statement of hip pain or impaired mobility + criteria of hip osteoarthritis according to guidelines on clinical examination without radiological imaging, or statement of hip pain or impaired mobility + radiological finding without clinical examination, or diagnosis of hip osteoarthritis by an orthopedist or rheumatologist without documentation of the radiological or clinical findings, or solely radiologically confirmed diagnosis of hip osteoarthritis	*
Simple form of self-reported hip pain and mobility impairment (e.g., visual analog scale [VAS], inquiry about hip pain and dysfunction on most days of previous month) or patients’ self-report of diagnosis by an orthopedist/surgeon or medical diagnosis (not by an orthopedist or rheuma - tologist) without documented criteria
2. Diagnosis of hip osteoarthritis with blinding to exposure
Yes. or influence of exposure on documentation of outcome can be practically excluded (substantiation required)	*
No
Not reported
3. Was the follow-up long enough for the endpoint to occur?
Yes (duration ≥ 10 years)	*
No
4. Adequate follow-up
Information as to outcome is available for at least 95% of all probands in the cohort	**
It is improbable that the rate of study drop-outs suffices to produce bias: follow-up rate ≥ 80% or nonresponder analysis shows comparability of responders and nonresponders	*
Follow-up rate <80% and no nonresponder analysis. or nonresponder analysis shows that responders and nonresponders are not comparable
No description. or differences in follow-up rates >15% between exposed and nonexposed groups

Statistical analysis

To estimate the risk of HOA we compared, in each study, the highest and the second highest exposure category with the lowest exposure category. The individual studies reported odds ratios (OR), relative risks (RR), or hazard ratios (HR) with 95% confidence intervals (95% CI). Owing to the low prevalence of HOA, these estimators of effect do not differ widely from one another, so we interpreted all relative effect estimators as estimated relative risks. The effect estimators of the individual studies were merged with the random effects model in meta-analyses to produce pooled RR for the various endpoints (STATA version 14, metan procedure).

We performed separate meta-analyses for studies on the lifting/carrying of loads and studies on physically demanding work, with separate evaluation of data for men, women, and sex-undifferentiated groups. Some studies investigated several of these categories and were therefore included in various subanalyses. Four studies in which pain was the sole diagnostic criterion were not included in the meta-analyses but were evaluated separately.

The absolute risks of HOA (19) were calculated from the pooled RR and the lifetime prevalences of 1.9% for men and 2.1% for women (e1). The heterogeneity of the effect estimators was interpreted substantively and described using the I² statistic (18).

Results

A total of 3419 titles and abstracts were scrutinized and 247 full texts were read. After exclusion of all publications that did not fulfill our criteria, five cohort studies (eTable 3) and 18 studies with evaluation according to case–control status (eTable 4) were included (Figure).

eTable 5. Risk of developing osteoarthritis of the hip* owing to the lifting of heavy loads: joint evaluation of men and women. comparison of highest and second highest exposure categories with lowest exposure category.

Study	No. of cases/ controls Age (years)	Comparison	QS	Effect estimator [95% CI]
Highest vs. lowest exposure category
Allen 2010 (e12) (CCS)	38/183 63.6 ± 10.5	Loads ≥ 50 kg. >10 x per week vs. <10 kg	7	OR 1.88 [1.21; 2.92]
Coggon 1998 (e8) (CCS)	73/47 45–91	Loads ≥ 50 kg more than 10 x per week for at least 10 years vs. <10 kg	12	OR 1.80 [1.1; 2.9]
Rubak 2014 (e7) (according to Rubak 2010 [e24]) (CCS)	126/85 41–69	>50 TY vs. 0 TY	14	OR 1.37 [1.06; 1.77]
Sobti 1997 (e11) (CCS)	23/984 70–75	>20 years with mean load weight >25 kg vs. 0 years	10	OR 1.14 [0.65; 2.00]
Yoshimura 2000 (e13) (CCS)	13/4 45. 63.8 ± 10.9	Loads >50 kg vs. <10 kg	9	OR 4.1 [1.1; 15.2]
Heterogeneity I2 = 26%		Increase in risk from 20 per 1000 to 31 per 1000 [24; 40]	RRpooled: 1.55 [1.2; 2.0]
Second highest vs. lowest exposure category
Allen 2010 (e12) (CCS)	75/501	Loads 20–59 kg. >10 x per week vs. <10 kg	7	OR 1.63 [1.15; 2.30]
Coggon 1998 (e8) (CCS)	57/52 45–91	Loads >20 kg. 10 x per working day for 13–24 years vs. <10 kg	12	OR 1.10 [0.67; 1.8]
Rubak 2014 (e7) (according to Rubak 2010 [e24]) (CCS)	605/979 41–69	10–50 TY vs. 0 TY	14	OR 1.36 [1.09; 1.68]
Sobti 1997 (e11) (CCS)	26/748 70–75	1–20 years with mean load weight >25 kg vs. 0 years	10	OR 1.65 [0.97; 2.80]
Yoshimura 2000 (e13) (CCS)	28/19 45. 63.8 ± 10.9	Loads 25–50 kg vs. <10 kg	9	OR 1.5 [0.7; 3.0]
Heterogeneity I2 = 0%		Increase in risk from 20 per 1000 to 28 per 1000. 95% CI: [24; 34]	RRpooled: 1.41 [1.2; 1.7]

CCS, Case–control study; CI, confidence interval; OR, odds ratio;

QS, quality score; RR, relative risk; RRpooled, pooled relative risk; TY, ton year (1 TY = lifting of 1 t in the course of 1 year)

* Total hip replacement or clinically/radiologically diagnosed osteoarthritis of the hip

eTable 4: Case–control studies that used data on physical activity and loads to assess exposure.

Study (reference) Country Study type	Study population Total no. of participants Response Cases/controls Sex Age Inclusion and/or exclusion criteria	Description of cases and controls Diagnostic criteria for hip OA Recruitment period	Exposure documentation Instrument Documented duration of exposure Activities/occupational groups Categories	Comparisons / adjustment / confounders	Quality criteria Selection (6) / comparability (4) / exposure (5) / total (15) Conflict of interest and/or funding
Allen 2010 (e12)** USA* CSS with evaluation by cases and controls	Samples (n=5138) from first and second follow-up of a cohort study (JoCo OA), adults >45 years from six communities in Johnston County. 933 men, 1796 women Cases: radiological OA 847, controls 1683 Cases: radiological OA + symptoms 294, controls 2212 Response 60% Exclusion criteria: In occupation <1 year	Cases: Radiological OA (Kellgren grade ≥ 2) and/or symptom in same hip (pain, stiffness on most days) Controls: Hip X-ray Kellgren grade 0 or 1 and/or no hip symptoms 1999-2004 and 2003-2004	Interview-supported questionnaire: Documentation of 10 activities in occupations performed for longest times (walking, carrying >10 pounds, sitting etc.) Lifetime exposure to eight activities (standing heavy labor >50% of work, standing light labor >50% of work, lifting 10/20/50 kg ≥ 10×/week, etc.)	Multivariate logistical regression: exposed versus nonexposed Adjustment: Age, sex, smoking, BMI, accidents, race, old knee injuries, housework	2/ 3/2/ 7 Conflict of interest: n.d. Funding: Center for Disease Control and Prevention/Association of Schools of Public Health
Coggon 1998 (e8) England CCS	Patients from two health districts 210 men (mean age 68 years), 401 women (mean age 71 years) Response: cases 84%, controls 58% Exclusion criteria: Hip fracture, Perthes disease, hip dysplasia, other secondary hip OA	Cases: All patients >45 years on waiting list for THR Controls: Patients from same general practice (exclusion criteria: previous hip interventions) 1993-1995	Interview-supported questionnaire All jobs >1 year, documentation of loads with frequency and weight, also Sitting, kneeling, etc.	At least 10 years of handling loads: <10 kg vs. 10-24 kg, 25-49 kg, > 50 kg (>10×/week) Matching for age and sex	5/4/3/12 Conflict of interest: n.d. Funding: Arthritis and Rheumatism Council of Great Britain
Croft 1992 (e5) England CCS **	All outpatients of the North Staffordshire urology department who underwent i.v. urography 787 (353 cases, 434 controls) Response 69% cases, 68% controls Age 60-75 years Exclusion criteria: Deceased, too ill to interview, X-ray image not evaluable	Cases: THR or joint space <2.5 mm, severe cases <1.5 mm Controls: Joint space >3.5 mm in both hips 1982-1987	PC interview on lifetime occupational history, information on activities (e.g., lifting loads >25 kg, without details of frequency)	OR for exposure vs. nonexposure, e.g., handling loads >25 kg for 1-19 years and >20 years vs. <1 year Adjustment: Age, hospital	2/3/4/9 Conflict of interest: n.d. Funding: n.d.
Cvijetic 1999 (e16) Croatia CSS with evaluation by cases and controls	Random sample of population of Zagreb (total pop. not reported): 678 participants Response rates not stated No. of cases/controls: n.d. 298 men, 292 women Age > 45 years Exclusion criteria: Rheumatoid arthritis, gout	Cases: OA Kellgren grade ≥ 2 Controls: All other probands of sample 1981-1982	Interview + questionnaire on occupational history Four categories: 1) Mostly sedentary jobs 2) Standing >80% of time 3) >80% non-sitting positions (walking, slight physical strain, lifting loads of up to 5 kg) 4) High physical strain (frequent walking, lifting loads of over 5 kg)	Comparison for each category: years of work 20-29 and >30 vs. years of work <20 Adjustment: Age, sex, BMI	4/4/3/11 Conflict of interest: n.d. Funding: n.d.
Elsner 1995 (e10) Germany CCS	Patients of an orthopedic practice: 134 men, 86 women Response cases: 61% Response controls: n.d. Age: n.d. Exclusion criteria: Hip dysplasia, inflammatory disease, obesity, trauma	Cases: Patients with hip symptoms and any, even discreet, radiological signs of hip OA (no closer definition) Controls: Symptom-free patients of a general and ophthalmological practice (no X-rays) 1989-1993	Questionnaire with questions on activities (in years) e.g., lifting and carrying 5-20 kg and >20 kg “in your life”	Activities carried out “always or frequently” vs. “rarely or practically never” Adjustment: Age	1/2/2/5 Conflict of interest: n.d. Funding: Hans Böckler Foundation
Heliovaara 1993 (e22) Finland CSS with evaluation by cases and controls	Representative sample (n=8000) of the Finnish population from residency registries Response: 7217 (90%), not stated separately for cases/controls Men (3637) and women (4363) >30 years 369 cases	Cases: Hip OA diagnosed by study physician on basis of medical history, symptoms, and clinical findings (sensitivity and mobility of hips, no more detailed information) Controls: All other probands of sample 1977-1980	Questionnaire on occupational history Generation of a sum index of “physical stress at work” (max. 5 points), with one point each awarded for lifting and carrying heavy loads, unfavorable body posture, repetitive movements, etc., in each case in the occupation performed for the longest time (duration not specified)	Sum index “physical stress at work”: 0 points vs. 1/2/3/4-5 points Adjustment: Age, sex, BMI, accidents	4/ 3/ 2/ 9 Conflict of interest: n.d. Funding: n.d.
Jacobsen 2004* (e23) Denmark CSS with evaluation by cases and controls	Proband from third follow-up Copenhagen City Heart Study (random sample) 1397 men (23-93 years), 2289 women (22-92 years) 2949 cases, 1202 controls Exclusion criteria: Surgery of fracture of hip, congenital hip disorder, rheumatoid arthritis	Cases: >3 positive responses in questionnaire on musculoskeletal symptoms and joint space width <2mm Controls: <4 positive responses in questionnaire on musculoskeletal symptoms 1991-1994	CCHS questionnaire: Activity (e.g., sitting, lifting and carrying >20 kg), duration of activity, frequency and weight of load documented; individual results not published	OR range 0.7-1.0 for type and duration of activity, including daily lifting; no individual data Adjustment: None	3/1/3/7 Conflict of interest: n.d. Funding: Research Board National University Hospital of Rigshospitalet, Danish Rheumatism Association, etc.
Keila-Kangas 2011 (e1) Finland CSS with evaluation by cases and controls	Sample from health survey 2000 3110 men (mean age 51 years), 6446 women (mean age 53 years) Response 82% Exclusion criteria: n.d.	Cases: THR, previous diagnosis on basis of convincing findings or mobility restrictions on clinical examination (e.g., internal rotation) and typical symptoms of hip OA. Controls: All participants 2000-2001	Interview: Five occupations carried out for longest time, cumulative years for working with loads >20 kg more than 10x per working day	Prevalence calculations for hip OA for years of load handling OR nonexposure vs. exposure for 1-12 years, 13-24 years and >24 years with loads >20 kg Adjustment: Age, BMI, smoking, trauma	5/4/4/13 Conflict of interest: None Funding: Finnish Work Environment Fund and Farmers’ Social Insurance Institution
Lau 2000 (e9) Hong Kong CCS	Patients from seven of the eight hospitals in Hong Kong 138 cases, 414 controls Cases: 30 men, 108 women Controls: 90 men, 324 women Age not stated Response: n.d. Exclusion criteria: Fracture, rheumatoid arthritis, dysplasia, other secondary hip OA	Cases: Hospitalized patients with hip OA Kellgren grade >3 or THR in the previous 3 years Controls: From general practice clinics in same region 01/1998-12/1998	Interview-supported questionnaire: Information on occupation performed for longest time: lifting, walking, kneeling, etc. with frequency, duration, and weight	Lifting of loads >10 kg never vs. 1-10×/week, >10× week Matching: Age, sex Adjustment: BMI, hip accidents	1/3/3/7 Conflict of interest: n.d. Funding: Health Services Research Committee of Hong Kong
Pope 2003 (e15) United Kingdom CCS	Sample of patients from two general practices Cases: 128 men, 224 women Controls: 1518 men, 1484 women Response 61% Exclusion criteria: n.d.	Cases: Statement of hip pain for at least 24 h in previous months and positive responses regarding hip pain in questionnaire Controls: All others Recruitment period: n.d.	Questionnaire covering all occupations carried out for >1 year: information on standing, sitting, walking, lifting of loads >23 kg (frequency of lifting not specified)	Multivariate analysis Nonexposure vs. loads >23 kg 1-12 years, > 13 years Adjustment: Age, sex	2/2/3/7 Conflict of interest: n.d. Funding: n.d.
Riyazi 2008 (e17) Netherlands CCS	Probands from the GARP study (Genetics, Arthrosis and Progression) – patients with hip OA from rheumatological-orthopedic practices and general medical practices in five big cities. 1874 probands contacted in writing; not known how many thereof were cases; 382 cases recruited, made up of 312 women and 70 men Response rate for cases not stated 345 controls, response 71% Age 40 – 70 years Exclusion criteria: n.d.	Cases: Diagnosis by a physician according to ACR criteria + femoral or acetabular osteophytes or joint space narrowing or THR Controls: Random sample of population of same region (telephone list, no checking whether controls had hip OA)	Questionnaire: Occupational designations documented, then professionals classified occupations as physically or mentally demanding activities (according to the Schellart system). Occupations classified as physically strenuous (lifting heavy objects, using heavy tools while standing) were, for example, agriculture, construction, and industry (duration not specified)	Physically demanding work vs. ?? (n.d., only result reported) Age matching (±5 years)	2/ 3/ 3/ 8 Conflict of interest: None Funding: Netherlands Organization for Scientific Research
Roach 1994 (e2) USA CCS	Members of US Army from Chicago area treated in a Department of Veterans’ Affairs Hospital Response: n.d. 99 cases, 233 controls Only men Mean age - cases: 69.3 years; controls: 67.6 years Exclusion criteria: occupational injuries, polyarthropathy, fractures of lower extremity, minimal joint space <1.5 mm	Cases: X-ray of hip due to symptoms (assessment by ARA criteria, Kellgren grade 3+4) or THR (from radiological database), Controls: Outpatients with i.v. pyelography 01/1989-06/1990	Questionnaire: Number of years in five categories of occupation: (1) light standing work, (2) sitting work, (3) heavy standing work, (4) work involving kneeling or crouching, (5) walking work Heavy work = categories 3, 4, and 5 with at least 15 years’ exposure Light work = categories 1 and 2 with at least 15 years’ exposure Intermediate work = all other probands	Heavy labor and moderately heavy labor vs. light work. For heavy labor vs. light work (15-24 years, 25-34 years, >34 years) Adjustment: Age, BMI, recreational activities	2/ 4/2/ 8 Conflict of interest: n.d. Funding: n.d.
Rubak 2014* (e7) Denmark CCS	Registry study (registries of residents and patients) All Danish men and women born between 1925 and 1964 who had been in work in the previous 10 years 861 case–control sets for women and 915 sets for men Mean age: men 64.3 years, women 64.7 years Response 71% Exclusion criteria: n.d.	Cases: First THR by primary hip OA Controls: Persons without THR 2005-2006	Questionnaire on occupations performed: cumulative load per day and frequency of lifting of loads >20 kg estimated by means of a job exposure matrix. Calculation of ton-years (= 1 t lifted per day or daily lifting of loads > 20 kg/10x day for a year) Whole-body vibration, standing	Lifting and carrying: nonexposure vs. ton-years vs. <10, 10-20, 20-115 ton-years Matching: Age, sex	5 / 4 / 5/14 Conflict of interest: n.d. Funding: Danish Ministry of Health and Prevention
Sobti 1997 (e11) England CSS with evaluation by cases and controls	Retired postal workers born 1919-1923 Age 70-75 years 2389 men, 1531 women 1736 controls, 1503 cases Response rate 78% Exclusion criteria: Questionnaire not fully completed	Cases: Two categories: (1) THR; (2) hip pain and stiffness in previous months Controls: n.d. Recruitment period: n.d.	Questionnaire: Information on all occupations carried out for >1 year. For average working day: lifting of loads >25 kg, climbing stairs, etc.	Regression model: Comparison, for example, of years in which loads >25 kg were lifted on an average working day. No exposure vs. 1-20 years, >20 years Adjustment: Sex	5/2/3/10 Conflict of interest: n.d. Funding: n.d.
Vingard 1991* (e4) Sweden CCS	All men in catchment area of four large hospitals in Stockholm 253 cases, 392 controls Response: cases 96%, controls 77% Age: 50-70 years Exclusion criteria: Malformations, sequelae of poliomyelitis, rachitis, injuries of trunk or lower extremity	Cases: First THR due to idiopathic OA (hospitals contacted) Controls: Random sample of study population 1984-1986	Telephone interview and questionnaire on lifetime occupational history. Questions on body postures, driving of vehicles (each in hours per week) and how many kilograms were lifted each week. Cumulative value from beginning of working life to diagnosis in years. For each exposure, classification into three categories: Low (no exposure + under 5%) Intermediate and high (each 50% of rest)	Intermediate and high exposures were compared with low exposures, e.g., lifting events >40 kg Adjustment: Age, BMI, sports activities	5/ 4/4/ 13 Conflict of interest: None Funding: National Institute of Occupational Health and Swedish Work Environment Fund
Vingard 1992 (e3) Sweden CCS **	Men receiving an invalidity pension from the Stockholm district (registry) Age: 46-65 years 140 cases/298 controls Response rates: n.d. Exclusion criteria: psychic disorders, trauma, rheumatic diseases, congenital malformations	Cases: Invalidity pension owing to hip OA Controls: Random sample of population 1979, 1980, 1981, 1984	Interviews on occupational history followed by Nordic Occupational Classification. Exposure was classified as low, intermediate, and high. Survey of work history up to 1981	Calculation of RR for receiving invalidity pension due to hip OA: intermediate and high exposure vs. low exposure Adjustment: Age, sex	4/2/4/10 Conflict of interest: n.d. Funding: Swedish National Social Insurance Board and Swedish Work Environment Fund
Vingard 1997 (e6) Sweden CCS	All women aged 50-70 years from five districts in Sweden Initially 255 cases/334 controls Response: 230 cases (90%)/273 controls (82%) Exclusion criteria: Every other form of arthritis and trauma of lower extremity, controls with known hip disease	Cases: THR due to primary hip OA (from National Register of Total Hip Replacement), diagnosis based on defined protocol with radiograph and clinical examination Controls: From residents‘ registration offices of study regions, matched for age 1991-1994	Interview and questionnaire on occupational history: cumulative exposure in hours between ages of 16 and 50 years. Each type of exposure was classified into three categories: Low (lowest 25%) Medium (25-50%) High (upper 25%)	Comparison of high and intermediate exposure vs. low exposure for physically strenuous activity, heavy carrying, etc. Adjustment: Age, BMI, sports activities	6/ 4/ 3/ 13 Conflict of interest: n.d. Funding: Swedish Council for Work–Life Research
Yoshimura 2000 (e13) Japan CCS	Patients from the registries of five hospitals in two cities Cases/controls: each 103 women and 11 men Response: cases 90%, controls 88% Exclusion criteria: Rheumatoid arthritis, congenital hip disease, Bekhterev disease	Cases: >45 years, on waiting list for THR Controls: From residents‘ registration offices (exclusion criterion: THR) Recruitment period: 1 year	Questionnaire on principal occupation: handling of loads >10 kg during an average working week. Also kneeling, walking, etc.	Principal occupation (mean duration 23 years) nonexposure vs. 10 kg, 25 kg, 50 kg Matching: Age, sex, region Additional adjustment: knee pain, mean age at the first job	5/2/2/9 Conflict of interest: n.d. Funding: n.d.

* Several publications related to one study, here summarized

** Study contains evaluation of both occupations and activities (see both eTables 3 and 4)

CCS, Case–control study; n.d., no data; OA, osteoarthritis; OR, odds ratio; CSS, cross-sectional studies; THR, total hip replacement

Figure.

Flow chart of literature survey

Participants in primary studies

The 23 studies selected for inclusion were mostly carried out in Europe (n = 19), predominantly in the Scandinavian countries, the Netherlands, or the UK.

In the five cohort studies the duration of follow-up varied from 2 to 22 years. The participants were mostly of working age (20 to 70 years). Their mean age at the time of investigation was usually over 45 years. The proportions of men and women varied widely among the occupational groups. Four studies (e2– e5) included only men, and in one study (e6) all the participants were women. All of the studies included in our analysis adjusted for age and sex, and 11 of the case–control studies and four cohort studies also adjusted for body mass index or other significant confounders.

Exposure

Eleven case–control studies and one cohort study investigated the exposure effect for the lifting of loads. In five of these case–control studies (e1, e7– e10) men and women were analyzed separately. Two of the case–control studies included only men (e4, e5), while one was restricted to women (e6). Five of the case–control studies evaluated men and women together (e7, e8, e11– e13). Two of the case–control studies and one cohort study diagnosed HOA on the basis of pain and were therefore analyzed separately, because pain is not a reliable diagnostic criterion (e11, e14, e15). The degree of exposure was determined by means of questionnaires or interviews: the loads ranged from 5 kg (e16) to more than 40 kg (e4). One study (e17) could not be included in the meta-analysis owing to a lack of sufficient data on categories of exposure.

Five case–control studies and three cohort studies did not describe the loads lifted or carried in concrete terms but instead defined criteria for “physically demanding work.” The definition of physically demanding work usually comprised a combination of various characteristics of activity, e.g., handling loads, heavy manual activity, and long periods of standing or walking. A sedentary occupation was generally defined as light manual labor. Some studies classified occupations as involving light or heavy physical strain. Five studies investigated the risk for men (e2, e3, e16, e18, e19), two were restricted to women (e18, e19), and one did not differentiate by sex (e20). Two cohort studies used pain as diagnostic criterion (e14, e21).

Diagnosis

The studies included in meta-analysis employed various criteria for the diagnosis of HOA:

• Implantation of a hip joint prosthesis (total hip replacement, THR) (e4, e6– e8, e11, e13, e18, e19)

• Clinical and radiological diagnostic criteria or THR (e1, e5, e9, e15, e17)

• Exclusively clinical or radiological diagnostic criteria (e3, e10, e12, e16, e20, e22, e23).

Quality assessment

The five cohort studies selected for inclusion scored 12 to 15 quality points on a scale of 0 to 19 (eTable 3).

eTable 3: Cohort studies that used data on physical activity and loads to assess exposure.

Study (reference) Country FU	Study population Baseline: No. of participants, Response Recruitment period Age Sex	Exposure documentation Duration of exposure Activities	Endpoint documentation Confirmation of diagnosis of hip OA, Time of diagnosis Lost to FU Exclusion of probands with hip OA/pain at BL	Comparisons / adjustment / confounders	Quality criteria Selection (5) / exposure (4) / comparability (4) / endpoint (6) / total (19) Conflict of interest and/or funding
Flugsrud 2002 (e19) Norway FU: 0.1-9.5 years (average 9.0 years)	Participants in “National Health Screening Service“ (on cardiovascular disease) in three districts of Norway Baseline 1977-1983 Compilation of data from cardiovascular screening registry and Norwegian arthroplasty registry 56 818, men and women (response rate BL 91%, 24 997 men and 25 037 women 34-59 years (mean age 46.6 years)	Questionnaire on physical activity at work (in previous year): self-assessment as sedentary (e.g., office work), intermediate (e.g., light industry), moderately heavy (e.g., construction), heavy (e.g., agriculture)	First hip replacement FU 1989-1998 Exclusion of probands with THR at beginning of FU or secondary hip OA	RR for sedentary versus intermediate, moderately heavy, heavy work Adjustment: Age, sex, BMI, leisure exercise, smoking	4/ 2/ 4/ 4/ 14 Conflict of interest: n.d. Funding: OrtoMedic Charnley Fellowship
Juhakoski 2009 (e20) Finland FU: 20-23 years	8000 men and women (representative sample of Finnish population; Mini-Finland Health Examination Survey) 30-70 years (mean age 42 years) Response BL: 90% 1978-1980	Questionnaire + interview on physical strain at work (reference value for duration of exposure not stated), six groups: Light sedentary activity, other sedentary activity, light standing physical activity, light to moderately heavy labor, heavy manual labor, very heavy manual labor	Diagnosis of hip OA by clinical examination according to protocol + symptoms or only by symptoms Follow up 2000-2001 in framework of “Health 2000 Survey”: 1286 previous participants were invited to FU, 909 (70%) took part in FU Exclusion of probands with hip OA at BL	OR for light sedentary activity vs. other groups Adjustment: Age, sex, BMI, smoking	5/2/3/2/12 Conflict of interest: none Funding: Mikkeli Central Hospital
Ratzlaff 2011* (e21) Canada FU: 2 years	127 424 men and women, members of “Canadian Association of Retired Persons” approached (e-mails and newsletters); 4258 participants Response rate 3.3 % Age 45-85 years (mean 61.6 years) 2005	Questionnaire on lifetime working history, calculation of CPFI, i.e., product of lifetime specific activities (hours) x body weight (kg) x typical peak hip joint loading (kg h/year), “peak joint force” from literature review	Self-reported hip OA (questionnaire with specific items and pain scale—verification by means of validity study on 100 probands, kappa 0.65) Documentation of outcomes at BL and FU Exclusion of probands with hip OA at BL FU 2006 and 2007: 2918 FU participants (68% response)	HR for lifetime exposure from CPFI quintiles 2-5 vs. quintile 1 Adjustment: Age, sex, accidents	2/ 4/ 3/ 1/ 10 Conflict of interest: none Funding: Canadian Institute of Health Research, encouragement to take part in study lottery with prizes of $1500
Rubak* 2013 (e18) Denmark FU: ≥ 10 years	All persons born in Denmark between 1925 and 1964 (Danish Civil Registration System) 2 522 349 men and women Mean age at beginning of follow up: women: 48.2 years (SD 9.9), men 49.1 years (SD 10.5) Registry study 2007	At least 10 years full-time work Data from Supplementary Pension Fund Register Classification of work according to industry exposure matrix (IEM) 0=minimal strain, 1=moderate strain, 2= high strain; “point-years” calculated from no. of years‘ work, added to IEM score for all types of work (in analogy to pack-years)	First THR due to primary hip OA, Data from “National Patient Register” from “Classification of Surgical Procedures” FU: 1996-2006 Exclusion of probands with hip OA at BL	OR for cumulative physical strain at work (point-years) 0-5, 5-15, 15-25, 25-35,35-86 vs. persons who have never worked in industry or in occupations involving moderate or high strain Adjustment: Age, sex, SES	5/2/3/5/15 Conflict of interest: n.d. Funding: Danish Ministry of Health and Prevention
Tüchsen 2003 (e14) Denmark FU: 5 years	Random sample from Central Population Register BL 1990 with 9653 participants, age 18-59 years Response 90%	Questionnaire survey of physically demanding work, heavy lifting (>20 kg), sedentary work Each category classified as “seldom or never” and “≥ 25% of working time”; no reference duration stated	Questioned about hip pain (in previous 12 months) Interview FU: 1995, only participants who were in work in 1990. n= 5001 (84% response) Exclusion of probands with hip OA at BL Total of 223 incident case observed	Logistic regression, OR for heavy lifting and physically demanding work for ≥ 25% of working time versus seldom or never Adjustment: Age, sex, BMI, smoking, and the various categories against each other	5/3/3/1/12 Conflict of interest: n.d. Funding: n.d

*Several publications related to one study, here summarized

BL, Baseline; CPFI, cumulative peak force index; FU, follow-up, HR, hazard ratio; n.d., no data; OA, osteoarthritis, OR, odds ratio;

RR, relative risk; SD, standard deviation; SES, socioeconomic status; THR, total hip replacement

Three population-based studies met all the quality criteria with regard to selection of participants. Two studies inquired about the weight of the loads lifted or carried. One frequently occurring limitation was joint analysis of men and women. The diagnosis was established by valid means in two studies, while in all other studies it rested on information provided by the probands.

The 18 included case–control studies scored 5 to 14 quality points on a scale of 0 to 15 (eTable 4). Seven studies attained at least 5 of a possible 6 points for “selection and representativeness” of the cases and control subjects. Seven studies scored maximum points for “comparability of cases and controls”. Only one study (e7) scored maximum points for determination of exposure. This study was based on assessment of exposure using a job exposure matrix (JEM). Four studies (e1, e7, e9, e11) inquired about load weight and the duration and frequency of lifting/carrying events, permitting estimation of the cumulative dose.

Study results

In men, lifting loads of the highest exposure category doubled the risk of HOA/THR (RR 2.09 [1.4; 3.1]). The second highest exposure category was associated with a 1.3-fold increase in risk (RR 1.35 [0.9; 1.9]). All studies found an increase in the risk of HOA/THR with increasing exposure. Based on a mean lifetime prevalence of HOA of 1.9% (19 per 1000), the risk in the highest exposure category rose to 40 per 1000 (95% CI: [27; 59]) (table 1). In the second highest exposure category the risk of HOA/THR went up to 26 per 1000 [18; 36]. Therefore, a positive dose–response relationship was demonstrated both at the level of the seven studies included and at the level of the pooled estimators (table 2).

Table 1. Risk of developing osteoarthritis of the hip*¹ owing to the lifting of heavy loads in men: comparison of highest and lowest exposure category.

Study	No. of cases/ controls Age (years)	Comparison	QS	Effect estimator [95% CI]
Highest versus lowest exposure category
Coggon 1998 (e8) (CCS)	60/36 45–91	Loads ≥ 50 kg more than 10 × per week for at least 10 years vs. <10 years	12	OR: 3.20 [1.60; 6.50]
Croft 1992 (e5) (CCS)	26/85 60–75	Loads >25 kg for more than 20 years vs. <1 year exposure	9	OR: 2.50 [1.10; 5.69]
Elsner 1995 (e10) (CCS)	61/34 adults*2	Loads >20 kg. always/often vs. seldom/ practically never	5	OR: 1.10 [0.64; 2.10]
Kaila-Kangas 2011 (e1) (CCS)	23/3110 30–97	Loads >20 kg. >10 × per working day. for more than 24 years vs. 0 years	13	OR: 2.30 [1.2; 4.3]
Lau 2000 (e9) (CCS)	8/6 n.d.	Loads ≥ 50 kg at least 10 × per week vs. 0 × per week	7	OR: 9.60 [2.2; 42.2]
Rubak 2014 (e7) (CCS)	249/343 41–69	>20 TY (1 TY = daily lifting of 1 t for 1 year) vs. 0 TY	14	OR: 1.35 [1.05; 1.74]
Vingård 1991 (e4) (CCS)	n.d. 50–70	25 200–930 600 lifting events (loads >40 kg) up to age 49 years vs. 0–1187 lifting events	13	RR: 2.4 [1.5; 3.83]
Heterogeneity I2 = 66%		Increase in risk from 19 per 1000 to 40 per 1000; 95% CI: [27; 59]		RRpooled: 2.09 [1.4; 3.1]
					Increase in risk with high exposure

CCS. Case–control study; CI. confidence interval; n.d.. no data; OR. odds ratio;

QS. quality score; RR. relative risk; RR_pooled. pooled relative risk; TY. ton year

*¹ Total hip replacement or clinically/radiologically diagnosed osteoarthritis of the hip; *² age range not specified

Table 2. Risk of developing osteoarthritis of the hip*¹ owing to the lifting of heavy loads in men: comparison of second highest and lowest exposure category.

Study	No. of cases/ controls Age (years)	Comparison	QS	Effect estimator [95% CI]
Second highest versus lowest exposure category
Coggon 1998 (e8) (CCS)	37/31 45–91	Loads 25–49 kg more than 10 × per week for at least 10 years vs. <10 years	12	OR: 1.90 [0.90; 3.90]
Croft 1992 (e5) (CCS)	14/106 60–75	Loads >25 kg for 1–20 years vs. <1 year exposure	9	OR: 1.20 [0.50; 2.89]
Elsner 1995 (e10) (CCS]	69/46 adults*2	Loads 5–20 kg. always/often vs. seldom/ practically never	5	OR: 1.00 [0.60; 1.80]
Kaila-Kangas 2011 (e1) (CCS)	10/3110 30–97	Loads >20 kg. >10 × per working day. for 13–24 years vs. 0 years	13	OR: 2.20 [0.8; 5.9]
Lau 2000 (e9) (CCS)	5/4 n.d.	Loads ≥ 50 kg 1–10 × per week vs. 0 × per week	7	OR: 8.50 [1.6; 45.3]
Rubak 2014 (e7) (CCS)	153/303 41–69	10 TY to <20 ty vs. 0 ty	14	OR: 0.89 [0.67; 1.17]
Vingård 1991 (e4) (CCS)	n.d. 50–70	1188–25 199 lifting events (loads >40 kg) up to age 49 years vs. 0–1187 lifting events	13	RR: 1.38 [0.85; 2.40]
Heterogeneity I2 = 66%		Increase in risk from 19 per 1000 to 26 per 1000; 95% CI: [18; 36]		RRpooled: 1.35 [0.9; 1.9]
Increase in risk with high exposure

CCS, Case–control study; CI, confidence interval; n.d., no data; OR, odds ratio;

QS, quality score; RR, relative risk; RRpooled, pooled relative risk; TY, ton year

*¹ Total hip replacement or clinically/radiologically diagnosed osteoarthritis of the hip; *² age range not specified

In women, lifting loads increased the risk of HOA/THR in the two highest exposure categories by about 40% compared with the lowest exposure category (RR 1.41; [1.0; 1.9] and RR 1.40 [0.9; 2.2] respectively). The risk of developing HOA thus increased from 21 per 1000 [21; 40] to ca. 30 per 1000 [19; 46] (table 3).

Table 3. Risk of developing osteoarthritis of the hip*¹ owing to the lifting of heavy loads in women: comparison of highest and second highest exposure categories with lowest exposure category.

Study	No. of cases/ controls Age (years)	Comparison	QS	Effect estimator [95% CI]
Highest versus lowest exposure category
Coggon 1998 (e8) (CCS)	13/11 45–91	Loads ≥ 50 kg more than 10 × per week for at least 10 years vs. <10 years	12	OR: 1.10 [0.49; 2.46]
Elsner 1995 (e10) (CCS)	18/11 adults*2	Loads >20 kg. always/often vs. seldom/practically never	5	OR: 1.90 [0.80; 4.80]
Kaila-Kangas 2011 (e1) (CCS)	16/3 446 30–97	Loads >20 kg. > 10 × per working day. for more than 24 years vs. 0 years	13	OR: 1.20 [0.7; 2.1]
Lau 2000 (e9) (CCS)	21/29 k. A.	Loads ≥ 50 kg at least 10 × per week vs. 0 × per week	7	OR: 2.90 [1.50; 5.60]
Rubak 2014 (e7) (CCS)	25/23 41–69	>20 TY vs. 0 TY	14	OR: 1.0 [0.7; 1.4]
Vingård 1997 (e6) (CCS)	82/n.d. 50–70	>44 089 lifting events (weight unspecified) up to age 50 years vs. 0–20 328 lifting events	13	OR: 1.50 [0.90; 2.50]
Heterogeneity I2 = 46%		Increase in risk from 21 per 1000 to 44 per 1000; 95% CI: [21; 40]	RRpooled: 1.41 [1.02; 1.94]
Second highest versus lowest exposure category
Coggon 1998 (e8) (CCS)	20/21 45–91	Loads 25–49 kg more than 10 × per week for at least 10 years vs. <10 years	12	OR: 0.90 [0.49; 1.66]
Elsner 1995 (e10) (CCS)	37/27 adults*2	Loads 5–20 kg. always/often vs. seldom/practically never	5	OR: 2.00 [1.00; 4.10]
Kaila-Kangas 2011 (e1) (CCS)	8/3 446 30–97	Loads >20 kg. 10 × per working day. for 13–24 years vs. 0 years	13	OR: 3.8 [1.7; 8.1]
Lau 2000 (e9) (CCS)	10/18 n.d.	Loads ≥ 50 kg. 1–10 × per week vs. 0 × per week	7	OR: 2.00 [0.89; 4.60]
Rubak 2014 (e7) (CCS)	33/40 41–69	>10 to 20 TY vs. 0 TY	14	OR: 0.81 [0.61; 1.1]
Vingård 1997 (e6) (CCS)	n.d./n.d. 50–70	20 329–44 089 lifting events (weight unspecified) up to age 50 years vs. 0–20 328 lifting events	13	OR: 1.10 [0.71; 1.71]
Heterogeneity I2 = 74%		Increase in risk from 21 per 1000 to 29 per 1000; 95% CI: [19; 46]	RRpooled: 1.40 [0.9; 2.2]

CCS, Case–control study; CI, confidence interval; n.d., no data; OR, odds ratio;

QS, quality score; RR, relative risk; RRpooled, pooled relative risk; TY, ton year

*¹ Total hip replacement or clinically/radiologically diagnosed osteoarthritis of the hip; *² age range not specified

The five studies that did not evaluate men and women separately (e7, e8, e11– e13) also confirm that higher exposures are associated with an increased risk of HOA (etable 5). The lifting and carrying of heavy loads also increases the risk of developing hip pain (RR 1.47 [1.1; 1.9]) (e11, e15). The moderate to high heterogeneity displayed by all risk estimators can be attributed to differences in study design and the variation in the weight of the loads concerned. Sensitivity analyses based on high-quality studies (scoring at least 12 of 15 points for quality) confirm the increase in the risk of developing HOA with increased exposure (e1, e4, e6– e8).

In men, physically demanding work increased the risk of HOA/THR 2.5-fold in the highest strain category (RR 2.46 [1.3; 4.8]) and twofold in the second highest strain category (RR 1.97 [1.3; 3.1]) (etable 6). The increase in risk was somewhat lower for women: 38% in the highest (RR 1.38 [0.9; 2.2]) and 30% in the second highest strain category (RR 1.30 [0.9; 2.0]) (etable 7). Studies that evaluated men and women together confirm these findings (etable 7). The two cohort studies (e14, e21) that investigated hip pain found that activities involving physical strain increased the risk of pain (RR 1.77 [1.3; 2.4]).

eTable 6. Risk of developing osteoarthritis of the hip* owing to physically demanding work in men: comparison of highest and second highest exposure categories with lowest exposure category.

Study	Participants Age (years)	Comparison	QS	Effect estimator [95% CI]
Highest vs. lowest exposure category
Case–control studies
Cvijetić 1999 (e16)	50/50 62.5 ± 10.3	High physical demands (>80% walking. standing. loads >5 kg) for >30 years vs. <20 years	11	OR: 1.22 [0.98; 2.46]
Roach 1994 (e2)	54/97 68.2 ± 6.4/ 67.7 ± 7.1	Hard work while standing. kneeling/walking for at least 15 years vs. light work (sitting/standing)	8	OR: 2.4 [1.3; 4.3]
Vingård 1992 (e3)	50/23 46–65	High personal exposure to strain (calculated from years in occupation and four load categories from slight to high) in past 20 working years vs. low exposure	10	OR: 12.4 [6.7; 23.0]
Case–control studies: Heterogeneity I2 = 92%			RRpooled: 3.74 [2.6; 5.5]
Cohort studies
Flugsrud 2002 (e19)	6202/6420 34–59	Physically demanding work (e.g.. forestry. farming; n.d. on duration of exposure) vs. sedentary work	14	RR: 2.1 [1.5; 3.0]
Rubak 2013 (e18)	88 131/21 407 49.1 ± 10.5	35–86 PY vs. 0 PY	15	OR: 1.33 [1.17; 1.53]
Cohort studies: Heterogeneity I2 = 82%			RRpooled: 1.63 [1.1; 2.5]
Second highest vs. lowest exposure category
Case–control studies
Cvijetić 1999 (e16)	92/92 62.5 ± 10.3	High physical demands (>80% walking. standing. loads >5 kg) for 20–29 years vs. <20 years	11	OR: 2.46 [0.51; 4.54]
Roach 1994 (e2)	27/60 68.2 ± 6.4/ 67.7 ± 7.1	Moderately hard work (not sitting. no hard work) for at least 15 years vs. light work (sitting/standing)	8	OR: 1.9 [1.0; 3.8]
Vingård 1992 (e3)	70/128 46–65	Moderate personal exposure to strain (calculated from years in occupation and four load categories from slight to high) in past 20 working years vs. low exposure	10	OR: 4.1 [2.4; 7.1]
Case–control studies: Heterogeneity I2 = 92%			RRpooled: 2.95 [2.0; 4.4]
Cohort studies
Flugsrud 2002 (e19)	5972/6420 34–59	Intermediate manual labor (e.g.. postman. construction worker) vs. sedentary work	14	RR: 1.7 [1.1; 2.4]
Rubak 2013 (e18)	171 795/21 407 49.1 ± 10.5	25–34 PY vs. 0 PY	15	OR: 1.27 [1.11; 1.48]
Cohort studies: Heterogeneity I2 = 33%			RRpooled: 1.36 [1.1; 1.7]
Total: Heterogeneity I2 = 79%		Increase in risk from 19 per 1000 to 26 per 1000 [21; 32]	RRpooled: 1.97 [1.3; 3.1]

CI, Confidence interval; n.d., no data; OR, odds ratio; PY, point years (years of employment × strain score from industry exposure matrix);

QS, quality score; RR, relative risk; RRpooled, pooled relative risk

* Total hip replacement or clinically/radiologically diagnosed osteoarthritis of the hip

eTable 7. Risk of developing osteoarthritis of the hip* owing to physically demanding work in women: comparison of highest and second highest exposure categories with lowest exposure category.

Study	Participants Age (years)	Comparison	QS	Effect estimator [95% CI]
Highest vs. lowest exposure category
Cvijetić 1999 (e16) (CCS)	49/49 63.5 ± 11.0	High physical demands (>80% walking. standing. loads >5 kg) for >30 years vs. <20 years	11	OR: 1.41 [0.78; 1.97]
Flugsrud 2002 (e19) (C)	1 368/3 423 34–59	Physically demanding work (e.g.. forestry. farming) vs. sedentary work	14	RR: 2.1 [1.3–3.3]
Rubak 2013 (e18) (C)	13 850/35 635 48.2 ± 9.9	35–86 PY vs. 0 PY	15	OR: 1.01 [0.88; 1.16]
Heterogeneity I2 = 80%		Increase in risk from 19 per 1000 to 31 per 1000 [21; 48]	RRpooled: 1.38 [0.9; 2.2]
Second highest vs. lowest exposure category
Cvijetić 1999 (e16) (CCS)	115/115 63.5 ± 11.0	High physical demands (>80% walking. standing. loads >5 kg) for 20–29 years vs. <20 years	11	OR: 2.34 [0.66; 2.93]
Flugsrud 2002 (e19) (C)	5 112/3 423 34–59	Intermediate: work with frequent walking and load lifting/carrying (e.g.. postman. construction worker) vs. sedentary work	14	RR: 1.4 [0.9; 2.0]
Rubak 2013 (e18) (C)	46 674/21 407 48.2 ± 9.9	25–34 PY vs. 0 PY	15	OR: 0.99 [0.88; 1.10]
Heterogeneity I2 = 72%		Increase in risk from 21 per 1000 to 23 per 1000 [17; 32]	RRpooled: 1.30 [0.9; 2.0]

C, Cohort study; CCS, case–control study; CI, confidence interval; OR, odds ratio; PY, point years (years of employment × strain score from industry exposure matrix); QS, quality score; RR, relative risk; RRpooled, pooled relative risk

* Total hip replacement or clinically/radiologically diagnosed osteoarthritis of the hip

Owing to the widely varying definitions of “physically demanding work,” the heterogeneity of the exposure effects for this form of exposure was greater than in studies of the effects of lifting loads. The study with a low-threshold definition of physically demanding work (e16) showed no increase in effects with increasing exposure, whereas very pronounced exposure effects were found in some of the studies with higher thresholds (e3). For this form of exposure too, analysis of only high-quality studies confirms that the risk increases with the level of exposure.

Discussion

The studies included in our analyses show that both the lifting of heavy loads and physically demanding work in general increase the risk of HOA. This finding is in agreement with previously published reviews (14, 15, 20) reporting a moderate to high increase in the risk of the development of HOA for “heavy lifting and carrying” and “physically demanding work,” whereby these earlier publications described the exposure effects purely in narrative form. Our study involved additional meta-analytic pooling of the effects observed in comparable studies.

All analyses show that men are at greater risk of developing HOA. In all categories considered, the increase in risk is lower for women than for men. This is probably due to methodological inadequacies of the studies concerned. Most of the studies did not cover occupations where women predominate (e.g., nursing). Frequently the high exposure categories included only small numbers of women.

The frequency of lifting was usually not reported in absolute terms; rather, a minimum number of lifting events (e.g., 10 per working day) was defined. It can be assumed that men lift heavy loads more often per working day than women and thus have greater exposure. A sufficiently reliable statement on the effects for women cannot be derived from the data of the studies we analyzed. Studies with hip pain as the sole diagnostic criterion (e11, e14, e15, e21) also show an increase in risk with increasing exposure.

Dose–response relationship

Examination of the effect estimators for the highest and second highest exposure categories shows that the risk of HOA increases with the weight of loads that are lifted and carried. A precise dose–response relationship could not be calculated because of the pronounced heterogeneity of data acquisition.

Four studies (e1, e7– e9) permit estimation of the minimum dose for loads. In men, the risk of developing HOA is increased after ca. 20 years’ regular lifting of loads weighing around 20 kg. For loads of 50 kg, the risk increases after only 10 years. In cumulative terms, lifting of at least 3000 to 5000 t is necessary to raise the risk of HOA to any significant extent. The weight thresholds seem be lower for women. However, the studies do not permit specification of relevant cumulative loads for women. Men in occupations involving physically demanding work have an increased risk of HOA after ca. 15 to 20 years (e2, e4).

Limitations

The precision of the effect estimators is reduced by the heterogeneity of the exposure. Not all studies reported the duration of exposure. Investigations into memory distortion (recall bias) with regard to occupational exposure have shown that patients tend to overestimate their past exposure, particularly in respect of lighter loads (<5 kg), shorter work tasks (<2.5 min), and recent exposure. In the studies we analyzed, the duration of exposure was usually over 10 years and the loads mostly exceeded 20 kg, or exposure was derived indirectly from the type of occupation. In general the risk of HOA tended to be overestimated by recall bias. However, we believe that recall bias probably had only a slight impact owing to the nature of the studies included. The data on the effects in women are insufficient.

Total hip replacement is often used as a study endpoint. However, the relative frequency of this operation differs widely among countries owing to variation in healthcare structures (21). Therefore, studies of THR patients do not reflect the total incidence of HOA. Despite these limitations, the consistently comparable results of the cohort and case–control studies indicate that our findings are robust.

Summary

For men, our data demonstrate a positive association between long-term lifting and carrying of heavy loads, or physically demanding work in general, and the risk of developing osteoarthritis of the hip. For women, we take the view that more research is needed.

On the basis of our findings, recommendations for preventive measures can be proposed. Loads of 20 kg or more should not be lifted without mechanical assistance. With the goal of detecting signs of HOA as early as possible, preventive occupational medicine should include examination of the hip after no more than 15 to 20 years in a relevant occupation. Since the limited available data do not show any meaningful effect of training and exercise on the progression of HOA in the occupational context (22), any measures taken should aim at reducing the amount of strain. Workers who need to change their job should take advantage of the occupational rehabilitation programs offered by health insurance providers and pension insurance funds. Osteoarthritis of the knee is legally defined in Germany as an occupational disease (code no. 2112) with a cause–effect relationship comparable to that which is demonstrable for HOA. Thus, together with compensation in individual cases, section 3 of the Ordinance on Occupational Diseases (Berufskrankheiten-Verordnung, BKV) provides for technical aids and occupational reorientation.

THE CLINICAL PERSPECTIVE. What can the company physician and family physician do?

• Early diagnosis of hip osteoarthritis*¹

  History

  • Hip pain (localized in groin, radiating pain)

  • Morning stiffness (over 30 min but under 60 min)

  • Impaired mobility of hip in daily life

  Examination

  • Limping, one side shorter than the other

  • Internal rotation <15°

  • Inguinal pressure and trochanter percussion pain

  Radiography

  • Joint space between femoral head and acetabulum <2 mm

  • Osteophytes and sclerosis

• Preventive occupational medicine*²

  Medical check up offered by employer for workers performing tasks associated with greatly elevated health risks for the musculoskeletal system, e.g., the lifting, holding, carrying, pushing, or dragging of heavy objects

  Preventive examination by employee request

  • In the presence of symptoms that are otherwise connected with the task concerned

  • The employer must offer the prevention scheme; participation is voluntary. Only with the employee’s agreement is the outcome communicated to the employer.

• Assistance and rehabilitation

  Those affected should take advantage of the support services provided for by law as early as possible, in order to alleviate their symptoms and adapt their workplace accordingly. In addition to the prescription of physiotherapy, the following options exist:

  • Prescription of rehabilitative exercise and functional training off budget, via the health insurance provider

  • Out-of-hospital preventive measures for employees with early functional impairments, via the pension insurance fund

  • Outpatient and inpatient medical rehabilitation via the pension insurance fund with subsequent participation in the aftercare program IRENA

  • Support in the workplace, via the pension insurance fund: funding for technical aids, occupational rehabilitation

  • Support for the disabled, via the local integration agency

*¹ According to the criteria of the American College of Rheumatology (7) and the radiological scoring systems described by Kellgren and Lawrence (8) and Croft (e5)

*² As stipulated by the Ordinance on Occupational Health Care (Verordnung zur arbeitsmedizinischen Vorsorge, ArbMedVV)

Key Messages.

• Both the lifting of heavy loads and physically demanding work increase the risk of osteoarthritis of the hip (HOA).

• In all the studies evaluated, the effect increased with increasing exposure. Precise delineation of a dose–response relationship was precluded by the heterogeneity of data acquisition.

• In all analyses the effects were lower for women than for men. There are methodological reasons for this. Women were clearly underrepresented in the studies evaluated. Occupations that involve physical strain and are typically carried out by women (e.g., nursing) were not included. Further research is required.

• Only four studies permit dose estimation with regard to loads. For men, the risk of developing HOA is increased after ca. 20 years’ regular lifting of loads weighing around 20 kg. For loads of 50 kg, the risk increases after only 10 years.

eTable 2. Quality criteria for case–control studies and cross-sectional studies.

I. Selection (max. 6 stars)	Assessment
1. Are the cases adequately defined?
If one of the following conditions is met, two stars are awarded: diagnosis by statement of hip pain + clinical examination with measurement of mobility/function + radiological imaging with clearly defined criteria as stipulated in guidelines (7, 8; measurement of joint space width), or implantation of a hip prosthesis (total hip replacement, THR), or on waiting list for THR	**
If one of the following conditions is met, one star is awarded: diagnosis from registry data or measurement of mobility/function by means of a complex scale with subscales for pain assessment, mobility, and function—e.g., Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC), Harris Hip Score (HHS), Hip Disability and Osteoarthritis Outcome Score (HOOS)—or statement of hip pain or impaired mobility + criteria of hip osteoarthritis according to guidelines on clinical examination without radiological imaging, or statement of hip pain or impaired mobility + radiological finding without clinical examination, or diagnosis of hip osteoarthritis by an orthopedist or rheumatologist without documentation of the radiological or clinical findings, or solely radiologically confirmed diagnosis of hip osteoarthritis	*
Simple form of self-reported hip pain and mobility impairment (e.g., visual analog scale [VAS], inquiry about hip pain and dysfunction on most days of previous month) or patients’ self-report of diagnosis by an orthopedist/surgeon or medical diagnosis (not by an orthopedist or rheumatologist) without documented criteria
2. Representativeness of the cases
Continuous or randomly selected and obviously representative case series from defined study basis	*
Possibility of selection bias or inadequate description of case identification
3. Selection of controls
General population or same population as cases. if a star was awarded for representativeness of cases	*
Same population as cases. but no star awarded for representativeness of cases
Hospitalized controls
No adequate selection or selection not described
4. Nonresponse rate
Response rate >80%. or participation rate 50% + nonresponder analysis shows comparability of responders and nonresponders	**
Response rate ≥ 50%. and difference in response rate between case and control groups does not exceed 25%	*
Nonresponder analysis shows comparability of responders and nonresponders
Not described or conditions not stated
II. Confounding. comparability of exposure groups. analysis (max. 4 stars)
1. Comparability of cases and controls on basis of design or analysis (adjustment. subgroup analyses. or other suitable methods: stratification. matching)
Analysis after adjustment, matching, or stratification for age, sex, and body mass index or other important confounders (with a coherent concept, i.e., the second star for adjustment for other important confounders should be awarded only if the adjustment is sufficiently justified, e.g., by description as confounders or change in estimate)	**
Analysis after adjustment. matching. or stratification for age and séx (if relevant)	*
No adjustment. no matching. no stratification
2. Have all the planned comparative hypotheses mentioned in the methods section been reported with quantifiable data?
Yes	*
No (e.g.. only significant differences are reported. or only p-values are given)
3. Have effect estimators been calculated separately for men and women?
Yes. or only one sex was investigated	*
No
III. Exposure (max. 5 stars)
1. Validity and reproducibility of data acquisition (methods)
Reliable. reproducible data acquisition (e.g.. measurement at workplace. valid documentation)	**
Structured interview/questionnaire or registry data	*
Written self-report
No or inadequate description of data acquisition
2. Accuracy of exposure data
Quantitative data on exposure (frequency and/or duration. e.g.. weight of load. hours driving a tractor)	*
Only occupation or activity specified. without previously mentioned information on concrete tasks
3. Documentation of exposure with blinding as to outcome
Yes	*
Influence of outcome on documentation of exposure can be practically excluded (substantiation required)	*
No
Not reported
4. Same method of data acquisition for cases and controls
Yes	*
No

eTable 8. Risk of developing osteoarthritis of the hip* owing to physically demanding work in men and women: comparison of highest and second highest exposure categories with lowest exposure category.

Study	Participants Age (years)	Comparison	QS	Effect estimator [95% CI]
Highest vs. lowest exposure category
Allen 2010 (e12) (CCS)	65/405 63.6 ± 10.5	Heavy labor (standing for >50% of working time) vs. <50% of working time	10	OR: 1.39 [1.01; 1.91]
Juhakoski 2009 (e20) (C)	11/178 63 ± 8. 50–94	Physically demanding work vs. light sedentary work	12	OR: 6.7 [2.3; 19.5]
Heliovaara 1993 (e22) (CCS)	27/421 n.d.	Sum index of manual labor strain 4–5 vs. sum index 0	9	OR 2.7 [1.7; 4.4]
Heterogeneity I2 = 82%		Increase in risk from 20 per 1000 to 37 per 1000 [28; 48]	RRpooled: 1.85 [1.4; 2.4]
Second highest vs. lowest exposure category
Allen 2010 (e12) (CCS)	63/560 63.6 ± 10.5	Light work (standing for >50% of working time) vs. <50% of working time	10	OR: 0.74 [0.54; 1.01]
Juhakoski 2009 (e20) (C)	13/178 63 ± 8. 50–94	Intermediate manual labor vs. light sedentary work	12	OR: 3.1 [1.2; 8]
Heliövaara 1993 (e22) (CCS)	62/825 n.d.	Sum index of manual labor strain 3 vs. sum index 0	9	OR 2.7 [1.8; 3.9]
Heterogeneity I2 = 93%		Increase in risk from 20 per 1000 to 26 per 1000 [20; 34]	RRpooled: 1.31 [1.0; 1.7]

C. Cohort study; CCS. case–control study; CI. confidence interval; n.d.. no data; OR. odds ratio; QS. quality score; RR_pooled. pooled relative risk

* Total hip replacement or clinically/radiologically diagnosed osteoarthritis of the hip