39 Work 35 (2010) 39–48 DOI 10.3233/WOR-2009-0956 IOS Press Biomechanical shoulder loads and postures in light automotive assembly workers: Comparison between shoulder pain/no pain groups F.A. Seamana, W.J. Alberta,∗ , N.R.E. Weldona, J. Crolla and J.P. Callaghan b a b Faculty of Kinesiology, University of New Brunswick, Fredericton, NB, Canada Department of Kinesiology, University of Waterloo, Waterloo, ON, Canada Received 23 July 2008 Accepted 8 January 2009 Abstract. Background: The number of workplace shoulder compensation claims resulting from musculoskeletal disorders (MSDs) has decreased slightly in recent years, however the median number of days off work remains unchanged, which suggests an increase in injury severity . Little information is available regarding cumulative shoulder exposures, and there is no information on their impact on shoulder pain. Methods: Seventy-nine automotive seat frame assembly workers completed a questionnaire about the prevalence and severity of shoulder pain and were videotaped performing assembly tasks. 3DMatch, a posture-matching software program, was used to calculate the peak and cumulative shoulder moments and forces by matching postures seen in the video with predetermined ranges of posture to be used in the biomechanical model. Results: Of the 45.6% who reported shoulder pain, there was a mild correlation of pain severity with posterior shear of the shoulder. There were no significant differences in peak loads between Pain and No Pain groups; however, the No Pain group experienced significantly more cumulative caudal shear. Conclusion: Although there was no difference in percent time spent in different flexed postures between pain groups, those working some jobs may be at an increased risk of developing MSDs based on the amount of time spent in flexed postures, as well as the peak flexion moment acting on the shoulder. Keywords: Shoulder pain, cumulative load, peak load, posture, workplace 1. Introduction Over the past few decades, much of the research on workplace musculoskeletal disorders has focused on risk factors for developing low back pain. Researchers have, with some success, identified work posture [18, 19], task repetition [5,19], and load magnitudes for ∗ Address for correspondence: Wayne J. Albert, 2 Peter Kelly Drive, Fredericton, NB, Canada E3B 5A3. Tel.: +1 506 447 3254; Fax: +1 506 453 3511; E-mail: [email protected] lifting [9,10,15] as variables to control in attempts to decrease the high incidence of low back pain. Shoulder pain and injury, though, have received much less attention. In a recent study of on-site physiotherapy visits by employees at an automotive plant, the shoulder was the most commonly treated area of the body and required the highest number of physiotherapy visits as compared with other parts of the body . In certain industries, such as manufacturing (i.e., vehicles), nursing, and mining, musculoskeletal disorder (MSD) rates 1051-9815/10/$27.50 2010 – IOS Press and the authors. All rights reserved 40 F.A. Seaman et al. / Shoulder loads and postures in automotive assembly are up to four times higher than those in the general population . The United States Bureau of Labor Statistics reported 141,800 injuries and illnesses in the vehicle manufacturing industry alone in 2006, 12.4% of the total annual average employment. Fifty-three percent of those resulted in days away from work, job transfer, or job restriction . The Workplace Safety & Insurance Board (Ontario, Canada) reported 22,614 (25.2%) lost time claims in the manufacturing industry in 2005; the highest of any other industry . Several studies have identified an association with increased physical demand and postural demands of the shoulder in relation to MSD development [18,19]. The repetitive nature of many upper extremity jobs has also been reported as a major risk factor, however, to date, there is little information quantifying the load demands associated with the repetitive manufacturing tasks and the associated cumulative loading on the shoulder. Furthermore, the available workplace assessment tools [11, 13,16] which include the shoulder are limited to postural evaluations and are unable to provide biomechanical loading information. Cumulative loading has become an accepted risk factor in low back pain development [10,15] and is now being used to assess the demands of the shoulder , however, only a few biomechanical models have been developed to calculate loading of the shoulder . The inconsistency in methodologies to determine shoulder demands or models has made it difficult to compare results, and a lack of injury data has made it impossible to determine a dose-response relationship. The main purposes of this study were to: 1) document peak and cumulative shoulder loads; 2) determine whether there are differences in peak and cumulative shoulder loads between sexes and pain groups; 3) determine whether there are differences in shoulder posture between genders and pain groups; 4) assess the relationship between cumulative shoulder loading and reported shoulder pain severity; and 5) determine potential risk of injury by comparing peak shoulder loads with published strength data. 2. Methods Data collection took place in an automotive seat frame assembly plant over a two-week period, and involved video data and questionnaire approaches. This information was used to document the peak and cumulative loads experienced by employees, as well as the prevalence of reported shoulder pain. Approval to conduct this research study was obtained from the University of New Brunswick’s Research Ethics Board. Informed consent was obtained from all participants. 2.1. Participants Employees from an automotive seat frame assembly plant were recruited through a meeting with all employees. Employees who were on reduced workload for any reason were excluded. Of the 110 participants who agreed to be videotaped and complete the questionnaire, the company released two participants before data collection began and two withdrew from the study, leaving 106 participants. Of these, 27 could not be used in the analysis because of incomplete data sets (one person did not complete the questionnaire and 26 did not have complete video data sets), leaving a final count of 79 participants, 34 males and 45 females (mean age of 37.1 years ± 11, mean height of 168.7 cm ± 10.3, mean mass of 75.7 kg ± 16.5). 2.2. Job and task selection The mass of the seat frame was approximately 8 kg, depending on the model being built. The jobs that contributed to the seat frame assembly were separated into the following groups: Epsilon, GeFit, Welders, Rewelders, and Train. The Epsilon line attaches the runners to the bottom of the seat frame and bolts it to a base. The GeFit line inserts small parts into the dials used to manually recline the seat. Welders insert several pieces into an autoweld machine to build the seat frame. Rewelders check all of the welds on the seat frame and touch up areas that need reinforcement. Train Drivers deliver parts to all assembly lines and cells in the plant, loading and unloading bins from a trolley with three levels of shelving. Of the 79 participants who took part in the study with complete questionnaires and video data, ten jobs that did not fit into the other 5 groupings (Epsilon (n = 17), GeFit (n = 17), Welders (n = 17), Rewelders (n = 5), Train Drivers (n = 5)) had only one or two workers who regularly performed the tasks. For statistical purposes and to maintain participant anonymity, these jobs were combined into a Miscellaneous group (n = 18) and will not be presented in the job-specific results. Although there were few workers in the Train and Rewelder job groups, the company wanted information about these specific jobs, and thus, were analyzed as separate groups. F.A. Seaman et al. / Shoulder loads and postures in automotive assembly 2.3. Documenting shoulder loads Each task was assessed by observing workers at each station as they performed regular assembly line duties. Prior to collecting video data, a period of time was set aside to allow the employee to become comfortable with the presence of the investigator, and with being videotaped. Two Panasonic video cameras (Secaucus, New Jersey, USA) were used to collect video at 30 Hz at an appropriate angle to the sagittal plane where the participant’s actions could be seen clearly . Participants were videotaped for the amount of time it took to collect at least ten cycles of the task, in order to choose three clear cycles to be used for analysis . A digital Chatillon force gauge (AMETEK Inc., Paoli, PA, USA) was used to measure all loads lifted, lowered, pushed and pulled by the participant. If the force gauge could not be attached to the piece being worked on, the participant was asked to exert the same amount of force as they would normally use to complete the task. This data was recorded and later entered into a biomechanical model as the three-dimensional hand forces. 2.4. Documenting shoulder pain Each participant was asked to fill out a questionnaire about the onset and severity of any current shoulder pain, history of shoulder pain, effect of shoulder pain on employment and activities of daily living, and whether treatment was sought for shoulder pain. Shoulder pain was defined as an ache, pain or discomfort in the shoulder region, whether or not it extended to the neck or down the arms. The pain could have been described as an aching, stiffness, burning, numbness or tingling sensation. A representative body diagram was included to help the worker identify the area with which they had experienced pain. A Visual Analogue Scale (VAS) was used to assess their perceived severity of pain, with the scale running from 0–10 cm (no pain to greatest amount of pain possible). Participants marked an ‘X’ on the line and their VAS score was then determined based on its position. 2.5. Data processing Each video file was imported into video capture software on a computer using Pinnacle Studio version 7.15.1 (Pinnacle Systems, Inc., CA, USA). Jobs were broken down into tasks and sub-tasks, when needed. Three clear cycles of each task were chosen and used for 41 analysis. The three cycles chosen for posture matching were selected based on video quality and ability of the investigator to clearly see the worker’s positions. If more than three cycles were of good quality, three cycles were chosen randomly for posture matching. Task cycles were clipped using the video capture software and converted into AVI format. Each video clip was then decimated to 3 Hz . 2.6. Posture matching Each decimated file was imported into the 3DMatch (University of Waterloo, Waterloo, ON, Canada) posture matching software used to assess cumulative and peak loads, and postures of the shoulder . 3DMatch is a custom program used for assessing the participant’s three-dimensional posture frame by frame and matching observed postures to the categories on the computer screen. Posture categories (bins) were developed based on joint range of motion and data from other postural approaches that have been used [17,24]. The median values from the posture bins selected are used as inputs in the rigid link quasi-static model to determine cumulative and peak loading variables (reaction moments for abduction/adduction, internal/external rotation, flexion/extension, and reaction forces for anterior/posterior shear, cranial/caudal shear, and compression/tension) [2,8]. 3DMatch also outputs information regarding time and percent of work cycle spent in flexed and abducted postures. Each calculation is done separately for the right and left shoulder. The height and mass of each participant was entered into 3DMatch in order to develop customized models for each participant. The hand forces were also entered into the model, specifying the direction of force (push, pull, or lift). After each video clip was posture-matched and run through the model, the three cycles for each participant’s task were averaged to create a mean value. If there were any sub-tasks, they were added together to account for each time they occurred during each cycle of the task. Each employee received a 30-minute lunch break and two 15-minute breaks which were accounted for, in addition to the seven hours of work time. Cumulative loading during a full shift was then estimated by extrapolating the average value of the three cycles. The number of cycle repetitions performed by workers during each shift was obtained from the company. This was used to determine the total time the participants performed assembly tasks. The remainder of the time was assessed with the participant in a neutral position, such as waiting for parts or walking around . 42 F.A. Seaman et al. / Shoulder loads and postures in automotive assembly Table 1 Prevalence (n) of reported shoulder pain by sex Sex Right shoulder Left shoulder Both Total only only shoulders Male (n = 34) 1 4 8 13 Female (n = 45) 7 6 10 23 Total (n = 79) 8 10 18 36 2.7. Data analysis All statistical analyses were performed using SPSS 14.0 for Windows. Statistical significance was set at an alpha level of 0.05. Descriptive statistics and a one-way ANOVA were conducted to assess differences in posture, cumulative and peak loads between sexes and pain groups (those who reported having shoulder pain and those who did not). The Kruskal-Wallis test was used as the nonparametric alternative when the Levene’s test for homogeneity was significant. A separate analysis was performed solely on the group of participants who reported shoulder pain. A Pearson Product Moment correlation was used to check for any correlations between the severity of reported shoulder pain and cumulative load variables. To provide a sense of potential risk, the shoulder peak loading was compared to published maximum strength data . For comparative purposes, maximum flexion at 30 degrees for women in the group of 30–39 years of age was used as it best corresponds to the average age of the group in the current study. Moderate potential risk of injury was determined as a peak flexion moment between 75% and 100% of maximum strength. High potential risk of injury was determined as greater than 100% of maximum strength. attributed to their work, four participants did not believe their shoulder pain was attributed to their work, and two participants were unsure. The most common activities reported as making their shoulder pain worse were: lifting (n = 26), pulling (n = 20), pushing (n = 17), twisting from the shoulder (n = 15), raising the arm in front (n = 14), overhead motions (n = 14), and raising the arm to the side (n = 13). There were no significant differences in the prevalence of reported shoulder pain between sexes (p = 0.996), therefore the sexes were combined for further analyses between Pain and No Pain groups. 3.2. Posture On average, the participants spent nearly half of their shift in mild shoulder flexion; 43.3% and 46.8% for the right and left shoulder, respectively. They also spent the majority of the shift in neutral abduction with 96.8% and 97.3% for the right and left shoulder, respectively. Males spent significantly more time in neutral right shoulder flexion (p = 0.007), and females spent significantly more time in mild right shoulder flexion (p = 0.003). There were no significant differences in percent time spent in neutral, mild, and severe flexed postures between the pain groups. 3.3. Peak load There were no significant differences in mean peak moments and forces between the Pain and No Pain groups (Fig. 1). 3.4. Cumulative loads 3. Results 3.1. Questionnaire results Of the 79 participants, 45.6% reported shoulder pain, 38.2% of males and 51.1% of females, the majority of whom reported experiencing shoulder pain in both shoulders. Table 1 presents the information on prevalence of reported shoulder pain. The following information refers only to responses from participants who reported experiencing shoulder pain (n = 36). The median VAS score for right shoulder pain was 3.5 cm (range from 0.5 to 7.6 cm), and 2.7 cm (range from 0.8 to 8.4 cm) for the left shoulder. Thirty-one participants believed their shoulder pain was Figure 2 compares the mean cumulative moments and forces acting on the shoulder between the Pain and No Pain groups. The No Pain group experienced significantly more cumulative caudal shear than the Pain group for both the right (p = 0.009) and left shoulder (p = 0.009). 3.5. Analysis of pain group There were no significant correlations between the right shoulder VAS scores and cumulative load variables. For the left shoulder, a mild correlation was found between the VAS scores and cumulative posterior shear (p = 0.023), and caudal shear (p = 0.036). F.A. Seaman et al. / Shoulder loads and postures in automotive assembly 43 Table 2 Risk of shoulder injury or disorder based on peak and cumulative loads experienced for each job. Bolded numbers signify high potential risk of injury and moderate risk of injury are shaded Peak loads Caudal Flexion shear (N) moment (Nm) 71.6 28.2 Epsilon (n = 17) GeFit (n = 17) Train (n = 5) Welders (n = 17) Rewelders (n = 5) 40 Pain 998.7 122.1 116.1 47.1 1600.9 297.2 52.1 16.3 1162.6 52.2 104.3 35.6 1224.2 162.1 b) No Pain 45 Pain 40 No Pain 35 20 n io n al ns Ex R Fl ot te at ex io io n n io at ot rn te 100 90 80 70 60 50 40 30 20 10 0 Force (N) Pain ns io n Te om pr es si on lS ra ni a C C he ar r he a al S au d C Po st er io rS he ar he ar No Pain te rio rS An ns io n Te om pr es si on C r he a ra ni al S C al S au d C er io rS he a he a r r r Po st Ex In Ex No Pain he a ct rn te rn te te In Pain rS io n io du ct ns Ab al Ex te Fl R R al rn io n ex io at at ot du Ad io n n io io ct io ct du Ab d) 100 90 80 70 60 50 40 30 20 10 0 An te rio R 0 n 5 0 n 10 5 n 15 10 du 15 25 Ad 20 30 al Moment (Nm) 25 Force (N) c) 14.8 30 n Moment (Nm) 35 48.9 ot a) Cumulative loads Caudal Flexion shear (kN·s) moment (kNm·s) 861.4 99.3 Fig. 1. Mean peak moments of the a) left shoulder and b) right shoulder, and mean peak forces for the c) left shoulder and d) right shoulder for Pain and No Pain groups. * denotes significance. 3.6. Risk of developing musculoskeletal disorders of the shoulder Table 2 illustrates the shoulder peak loading variables that may signify potential risk based on normative strength values from previous research . Table 3 shows the odds ratios for risk of developing shoulder MSDs, as determined by Punnett et al. , specific to the postures used by the participants in the current study. 4. Discussion Forty-five percent of the workers who took part in the study reported shoulder pain, which is similar to 44 F.A. Seaman et al. / Shoulder loads and postures in automotive assembly Table 3 Percent time in shoulder flexion and abduction of the right shoulder with corresponding odds ratios (in brackets) for risk of shoulder disorders based on work by Punnet et al.  Flexion Mild 53.6 45.2 37.1 19.4 39.2 43.8 Neutral 44.8 54.7 55.4 80.3 60.4 55.3 Epsilon GeFit Train Welders Rewelders Miscellaneous Severe* 0.5 (1.7) 0.1 (1.7) 6.1 (1.7)** 0.2 (1.7) 0.0 1.3 (1.7) Abduction Mild Severe 4.5 0.1 (1.7) 1.4 0.0 6 0.5 (1.7) 1.4 0.4 (1.7) 8.9 0.0 2.4 0.8 (1.7) Neutral 94.0 98.7 91.8 98.2 90.6 97.2 *The odds ratio for severe flexion of both shoulders at least once per task cycle is 4.0. **If the Train group had reached 10% of shift in severe flexion, the O.R. increases to 2.8 for the right shoulder and 5.1 for the left shoulder. Pain b) 250 No Pain 200 Moment (kNm*s) 150 100 50 100 50 An t er io rS io n rn al Ex te ns io ex io at ot R Fl n n io at ot R te io n he ar on Te ns i si pr es C om he ar lS C ra ni a al Sh e au d on 0 ar 200 0 C Ex 400 200 he ar al 600 Te ns 400 No Pain 800 da lS he ar C ra ni al Sh ea C r om pr es si on 600 Pain 1000 C au No Pain er io rS ct rn Force (kN*s) 800 Po st du te In 1200 Pain rS he ar io io ct du 1400 rS he ar 1000 An te rio n n n ns Ab al Ex te ex Fl R io n io n ot at io rn te Ex In te rn al Ad R ot du at ct io n n io n io ct du Ab d) n 0 1200 Force (kN*s) No Pain 150 0 c) Pain 200 Ad Moment (kNm*s) 250 Po st er io a) Fig. 2. Mean cumulative moments of the a) left shoulder and b) right shoulder, and mean cumulative forces for the c) left shoulder and d) right shoulder for Pain and No Pain groups. * denotes significance. the incidence rate of McClellan et al.’s  automotive manufacturing research. The majority of the reported pain was experienced in both shoulders. This result is not surprising in that 31 of the 36 who reported shoulder pain attributed the pain to their work and many of the jobs analyzed required symmetrical movements (such as lifting bins) or were equally demanding for both sides (e.g., retrieving the part with the left hand, working on it with both hands, and disposing of it with the right hand along the assembly line). All workers reporting only left shoulder pain were left-handed. Handedness was split evenly amongst those reporting only right shoulder pain. Although there was no significant difference between the shoulder pain groups and loading or postural outcome variables, previous research has provided evidence for the increase risk of MSDs when non-neutral postures are required for a significant percentage of the cycle time . As well, excessive exposure to peak and cumulative loads, especially for risk in low back pain development has been advocated [10,15]. Hughes et al.  suggested the maximum strength at 30 F.A. Seaman et al. / Shoulder loads and postures in automotive assembly degrees of flexion to be 37 Nm for women aged 30– 39 years, 75% of which would be 27.8 Nm. Using this criteria, job tasks with peak flexion moments falling within the range of 27.8–37.9 Nm were classified as moderate risk of shoulder injury development, and job tasks with peak flexion moments above 37.9 Nm were classified as high risk. Based on this method, Epsilon and Rewelders have moderate risk of injury. Although only males worked as Train Drivers (which explains the higher peak flexion moments), if a female were to perform the Train Driver tasks they would have a high potential risk of injury. The maximum strength at 30 degrees of flexion for males is 55 Nm, 75% of which is 41.3 Nm. Therefore, males are also at a moderate potential risk of injury when performing Train Driver tasks. Since there are no normative tables for caudal shear or threshold values for cumulative shoulder loading values from which to compare, an estimation of potential risk is not currently possible. 4.1. Posture Many studies have shown that increased time spent by workers in non-neutral postures increases the risk of developing MSDs and reporting shoulder pain . Comparing the postural demands of the auto manufacturing workers to the established odds ratios suggest a moderate risk of MSD development (OR = 1.7). On average, the Train group spent 6.1% of their shift in severe shoulder flexion, suggesting an OR of 1.7. Since the train drivers are lifting bins to and from the trolley, the risk of shoulder MSD development increases by a factor of 4 because both shoulders are put in severe flexion when loading and unloading from the top shelf. If they spend more than 10% of their work cycle in severe shoulder flexion it can increase their risk of developing MSDs by up to six times . Rewelders spent the most time in mild abduction (8.9%) since they need to manipulate the seat frame to check the welds. There are few studies with which to compare these results, however in a similar study in an assembly plant of automotive mouldings, McClellan et al.  reported that the workers were in mild shoulder flexion for 30% of their work shift. Although those workers spent less cycle time in flexed postures they spent more time in abducted postures than workers in the current study. In an assessment of the trunk, neck, shoulder, and elbow work postures of orthodontists, Newell and Kumar  suggested that increased time spent in static and deviated postures may lead to musculoskeletal disorders. In the case of the jobs in the current study, 45 workers assumed mild shoulder flexion for most of the shift. However, the jobs were fairly dynamic whereby the participants would reach for a part and then return to a neutral position. 4.2. Peak loads Unlike the NIOSH action limit for spine compression while lifting, there are no threshold values available for the shoulder beyond assessments of isometric shoulder strength. Based on the complexity of the shoulder, there are several modes of loading and the corresponding resulting peak loading variables, none of which have yet been deemed as the most detrimental to the shoulder joint and surrounding tissues. This study adds to the limited current knowledge on peak shoulder loading and provides a database for shoulder loading in a manufacturing environment. The reported shoulder loads show large standard deviations. When all jobs are combined for reporting purposes, the large standard deviations are a reflection of the diversity in the various jobs analyzed. For example, there is a large discrepancy in peak loads experienced for the Welders and the Train group. It was thought that by grouping by job, the variance would be smaller but individual technique such as posture used to complete the job, or choices made to complete the task faster (e.g. carrying more bins or parts) may contribute to large variations within jobs. Interestingly, the large variance was unaffected whether the participant level was large or small in a given group. This suggests that individual techniques used at each job should be examined more closely in future investigations. Although it is critical to document peak moments that workers experience, it is equally important to know their association to individual strength in order to move towards a threshold limit. We did not collect maximum strength values from our participants, but used normative strength tables from previously published research. Hughes et al.  published normative strength values for males and females in different age groups. Participants were tested in 30, 60 and 90 degrees of shoulder flexion and abduction with both their dominant and non-dominant arms. The published strength values for the age range of 30–39 years for 30 degrees of flexion and 30 degrees of abduction, were used to estimate the percentage of maximum strength used by the autoworkers in this study (all jobs combined). Males used approximately 42.2% of their maximum shoulder strength for the right shoulder, and 37.5% for their left shoulder. For abduction, males used approximately 31.5% 46 F.A. Seaman et al. / Shoulder loads and postures in automotive assembly of their maximum strength for the right shoulder, and 21.4% for the left shoulder. Females, on the other hand, used approximately 58.4% of their maximum shoulder strength for the right shoulder, and 46.0% for the left shoulder. For abduction, females used approximately 42.7% and 32.2% of their right and left maximum shoulder strength, respectively. Interestingly, the Train group (which was all male) used approximately 85.6% and 89.2% of their maximum shoulder flexion strength for the right and left shoulder respectively, and 78.2% and 48.4% of their maximum abduction strength for the right and left shoulder, respectively. Although not part of the current study, it would be interesting to know what percentage of time employees work above 30% of the maximum strength values, which would speak to endurance. 4.3. Cumulative loads As was seen with peak loads, there were large variances in the cumulative loads experienced by the participants. With all jobs combined, the large standard deviations could be related to the different demands of the jobs. When separated by job, however, the standard deviations are still large. This may be a result of differences in work technique. It is possible that the number of parts each participant was able to produce will be different. With GeFit, for example, it is estimated that one worker makes 2300 parts per shift. This number was then used to extrapolate the cumulative load values of one cycle to a full work shift. The remaining time was deemed to be unplanned rest . It is possible that the large standard deviations within job groups are a result of the extrapolation method. Since some participants had shorter cycle times, they subsequently had longer unplanned rest when extrapolating, and thus lower cumulative load values. In reality, the participant probably does not have that much rest time, but continues making parts and actually has a higher than average production rate, which would also lead to higher cumulative loads. Currently the only published limit with respect to cumulative loading of relative use for the shoulder is the ISO 11228-1 standard  set at 10,000 kg handled per day. This standard has been set for manual material handling activities and represents a limit based on the load being handled and not the net biomechanical loads experienced by the low back or shoulder structures. Using the maximum NIOSH lifting constant of 23 kg in an ideal lifting situation, this translates to less than one lift per minute in an 8-hour day. This limit is also based on the low back and not specifically for shoulder injury prevention. If the ISO limit were to be considered, the jobs assessed in this automotive plant would not be deemed risky as the total load handled per day averaged around 2000 kg. Unfortunately, the ISO limit neglects to consider static posture, non-neutral postures and cycle times in its assessment of cumulative load. These are all aspects that have been shown by previous researchers to have an association with cumulative loading [9,10,13,15]. Interestingly, the No Pain group experienced greater cumulative caudal shear than the Pain group. Intuitively, if cumulative loading is suggested as being associated with shoulder pain, then the higher caudal shear should be experienced by the Pain group. However, it is possible that the participants who experienced shoulder pain have altered their technique used to complete their work tasks in order to reduce their pain. By doing so, they have reduced the caudal shear that is acting on the shoulder. Since the data was collected post-injury, we do not have the ability to assess technique changes, only technique differences between groups. Longitudinal studies would be required to establish changes in specific technique as well as helping to establish a shoulder injury model. Finally, with the lack of homogeneity among the job tasks we were not able to determine specific variables that may have led to the differences in caudal shear. With a more homogeneous group, we may have been able to uncover additional information. It is possible that shoulder injury is the “response” variable that must be looked at instead of perceived pain. Pain is extremely subjective and the determination of a dose-response relationship may be impossible if it is based on the participants’ report of pain and pain severity. A model of shoulder injury is needed as one of the next steps, which should include the mechanism of injury. Through the development of an injury model, we may find that cumulative load is not a mechanism of shoulder injury at all or that it is predominately muscular in nature and thus the cumulative moments are the only variable of importance. 4.4. Limitations As with any study, there are limitations. For one, there was a fairly high attrition rate of 26%, mainly due to incomplete video data sets. Some of the assembly lines went down during the collection period resulting in the loss of video collection of participants F.A. Seaman et al. / Shoulder loads and postures in automotive assembly at each station on the line. Despite this, there was still a significant sample size, but some small job groupings. Only two investigators posture-matched the video data in order to reduce error. Each investigator posturematched the same participant for all jobs, but did not posture-match all videos for each job. As a result, any variance from the actual posture being used in the video would not necessarily be consistent between the investigators, however there is very little error in posture matching when performed by a trained observer . When collecting the video data, it was not always possible to collect at 45 degrees from the sagittal plane for all participants. Depending on the assembly line setup, the line of view was often obstructed by machines; therefore investigators were required to find an alternate position monitoring the participant’s movements. This was not deemed to be an issue since the use of 3DMatch for posture matching was found to have strong inter-observer reliability regardless of camera angle . The hand-held force gauge allowed the use of several different attachments to measure lift, push and pull forces with less than 0.25% error. Although it is possible that the technique of the participant was altered by holding the force gauge while simulating the task, it is not believed that it significantly impacted the technique of the participant based on the specificity of attachments used. neutral postures. They also suggest that there may be a protective adaptation in work performance based on the higher caudal shear in the No Pain group. A longitudinal study is required to further examine this relationship. It is therefore concluded that pain is not a robust measure within the workplace due to the subjective nature of its reporting. The lack of a clear relationship between the biomechanical variables and pain suggest that other variables (e.g., psychosocial, physical) may be necessary to explain the variability in the “dose” and that pain may not be the best indicator of the “response”. Future research needs to focus on the mechanism of shoulder injury to better establish the variables of this relationship and the database of shoulder loading needs to be expanded to include other industrial tasks. Acknowledgments The authors wish to acknowledge the Natural Sciences and Engineering Research Council of Canada and AUTO21 Network of Centres of Excellence for financial support, as well as the employees and management at Faurecia Automotive Seating Canada Ltd. References  5. Conclusion The 3DMatch postural assessment tool was used to compare shoulder postures, peak and cumulative loading collecting during automotive parts assembly. The documenting and databasing of actual shoulder loading in an industrial setting is the first step in establishing a dose-response relationship between cumulative loading and shoulder pain. The database of peak and cumulative exposures documented in this study has provided critical information for the “dose” component of the equation. Although it was the initial intent to obtain greater insight into the “response” component of the equation, the results of this study failed to demonstrate a clear association. However, the results did provide significant insight for future consideration in understanding this relationship. The results suggest that workers performing some light automotive assembly tasks are still at risk of developing shoulder disorders based on the magnitudes of peak loading and the percent of time spent in non- 47       D.M. Andrews and J.P. Callaghan, Determining the minimum sampling rate needed to accurately quantify cumulative spine loading from digitized video, Appl Ergonomics 34 (2003), 589–595. J.P. Callaghan, J.A. Jackson, W.J. Albert, D.M. Andrews and J.R. Potvin, The design and preliminary validation of 3DMatch – A posture matching tool for estimating three dimensional cumulative loading on the low back, Association of Canadian Ergonomists National Conference, 2003, London, ON, Canada. N.M. Dunk, K.J. Keown, D.M. Andrews and J.P. Callaghan, Task variability and extrapolated cumulative low back loads, Occup Ergonomics 5 (2005), 149–159. S.L. Fischer, W.J. Albert, A.J. McClellan and J.P. Callaghan, Methodological considerations in the calculation of cumulative exposure: An assessment of joint model and time standardization approaches on error, Ergonomics 50 (2007), 1365– 1376. G-A. Hansson, I. Balogh, K. Ohlsson, B. Palsson, L. Rylander and S. Skerfving, Impact of physical exposure on neck and upper limb disorders in female workers, Appl Ergonomics 31 (2000), 301–310. R.E. Hughes, M.E. Johnson, S.W. O’Driscoll and K.-N. An. Age-related changes in normal isometric shoulder strength, Am J Sports Med 27 (1999), 651–657. International Standards Organization. ISO 11228-1:2003(E) Ergonomics – Manual handling – Part 1: Lifting and carrying. Switzerland. 48           F.A. Seaman et al. / Shoulder loads and postures in automotive assembly J.A. Jackson, R.J. Reed, J.P. Callaghan, W.J. Albert and D.M. Andrews, Usability of 3DMatch – an evaluation of the intraobserver reliability of posture matching to calculate cumulative low back loading. Association of Canadian Ergonomists National Conference, 2003, London, ON, Canada. M. Jager, C. Jordan, A. Luttmann and W. Laurig, Evaluation and assessment of lumbar load during total shifts for occupational manual materials handling jobs within the Dortmund Lumbar Load Study – (DOLLY), Int J Ind Ergonomics 25 (2000), 553–571. S. Kumar, Cumulative load as a risk factor for back pain, Spine 15 (1990), 1311–1316. L. McAtamney and E.N. Corlett. RULA: A survey method for the investigation of work-related upper limb disorders, Appl Ergonomics 24 (1993), 91–99. A. McClellan, W.J. Albert, S. Fischer and J.P. Callaghan, Estimating cumulative loads on the shoulder during automotive assembly tasks, Occ Ergonomics 8 (2009), 81–90. J.S. Moore and A. Garg, The strain index: a proposed method to analyze jobs for risk of distal upper extremity disorders, Am Ind Hyg Assoc J 56 (1995), 443–458. T.M. Newell and S. Kumar, Comparison of instantaneous and cumulative loads on the low back and neck in orthodontists, Clin Biomech 20 (2005), 130–137. R. Norman, R. Wells, P. Neumann, J. Frank, H. Shannon, M. Kerr et al., A comparison of peak vs cumulative physical work exposure risk factors for the reporting of low back pain in the automotive industry, Clin Biomech 13 (1998), 561–573. S. Pinzke and L. Kopp, Marker-less systems for tracking working postures – results from two experiments, Appl Ergonomics 32 (2001), 461–471. L. Punnett, L.J. Fine, W.M. Keyserling, G.D. Herrin and D.B.         Chaffin, Back disorders and nonneutral trunk postures of automobile assembly workers, Scand J Work Environ, Health 17 (1991), 337–346. L. Punnett, L. Fine, W.M. Keyserling, G. Herrin and D. Chaffin. Shoulder disorders and postural stress in automobile assembly work, Scand J Work, Environ, Health 26 (2000), 283– 291. L. Punnett and D.H. Wegman, Work-related musculoskeletal disorders: the epidemiological evidence and the debate, J Electromyography Kinesiol 14 (2004), 13–23. J. Sadi, J. MacDermid, B. Chesworth and T. Birmingham, A 13-year cohort of musculoskeletal disorders treated in an autoplant, on-site physiotherapy clinic, J Occup Rehab 17 (2007), 610–622. C.A. Sutherland, W.J. Albert, A.T. Wrigley and J.P. Callaghan, The effect of camera viewing angle on posture assessment repeatability and cumulative spinal loading, Ergonomics 50 (2007), 877–889. C.A. Sutherland, W.J. Albert, A.T. Wrigley and J.P. Callaghan, A validation of a posture matching approach for the determination of 3D cumulative back loads, Appl Ergonomics 39 (2008), 199–208. United States Bureau of Labor Statistics, Department of Labor. Number of cases – detailed industry level. Industry Illness and Injury Data 2006. United States of America. P.M. van Wyk, P.L. Weir, D.M. Andrews, K. Fiedler and J.P. Callaghan, Determining the optimal bin size for a video-based posture assessment method, Applied Human Factors and Ergonomics Conference, 2008, Las Vegas, NV, USA. Workplace Safety & Insurance Board. Lost time claims by standard industrial classification (SIC). Statistical Supplement 2005; 21. Ontario, Canada.