Why cumulative loading calculated using non-weighted integration may not be suitable for assessing physical stress of the lower back: an empirical investigation of strain during lifting and lowering tasks

Abstract

When work-related physical stress is assessed using non-weighted integration, it is assumed that different loading conditions have a sufficiently comparable effect on the human body as long as the area under the loading curve is the same. Growing evidence cast doubt on whether this simple calculation can adequately estimate physical work-related strain. This study investigates in vivo, focussing on the lower back, whether the non-weighted method adequately reflects work-related physical strain of the lower back. Strain data resulting from lifting/lowering tasks performed in a laboratory study with an identical area under the loading curve but different load intensities were compared. Results showed that the non-weighted method does not sufficiently reflect the resulting muscular, cardiovascular and perceived strain but underestimates the influence of higher load intensity even in the range of medium physical exposure. Further research is needed regarding the determination of weighting factors and limit values.

Practitioner Summary Given the dynamic nature of most physical work activities, the assessment of time-varying loading of the lower back is of particular interest in practice. Results show that the widely used non-weighted calculation method does not accurately reflect the resulting physical strain but underestimates the influence of higher load intensity.

Abbreviations: MSD: musculoskeletal disorders; WMSD: work-related musculoskeletal disorders; KIM-LHC: Key Indicator Method Lifting, Holding, Carrying; RES: right erector spinae longissimus; LES: left erector spinae longissimus; HR: heart rate; RPE: rating of perceived exertion; EMG: surface electromyography; ECG: electrocardiography; SENIAM: Surface ElectroMyoGraphy for the Non-Invasive Assessment of Muscles; MVC: maximum voluntary contraction; ANOVA: analysis of variance; Std. error: standard error

HIGHLIGHTS

• Results of this empirical investigation suggest that the widely used non-weighted calculation method is not fully suitable for calculating cumulative loading of the lower back.

• Even in the range of medium physical exposure the non-weighted calculation method does not accurately reflect the resulting strain on the human body but tends to underestimate the influence of higher load intensity due to higher external weight.

• Despite the same cumulative loading value obtained when using the non-weighted method, the resulting physical strain values are generally about 20–25% higher.

• The results may be used to further develop ergonomic assessment methods in order to avoid a misclassification of loading conditions and to prevent the risk of overexertion.

Keywords:

• Cumulative loading
• manual material handling
• non-weighted integration
• lower back

Introduction

Musculoskeletal disorders (MSD) are one of the most common causes of chronic pain and activity limitations and lead to high health care costs worldwide (Dagenais, Caro, and Haldeman 2008; Geurts et al. 2018; Olafsson et al. 2018). In this context, the lumbar vertebra region is of particular interest as low back pain accounts for a large proportion of work-related MSD (WMSD) (Maher, Underwood, and Buchbinder 2017). With regard to risk factors for the development of WMSD of the lower back, there is consistent evidence of a causal relationship between forceful lifting tasks and WMSD (da Costa and Vieira 2010; Manchikanti et al. 2014; Putz-Anderson et al. 1997). For the ergonomic analysis of workplaces, the assessment of physical stress of the lower back is of central importance.

Given the dynamic nature of physical work activities in occupational practice (Iridiastadi and Nussbaum 2006), the consideration of the temporal variability of the loading is a fundamental requirement for assessment methods. In this context, however, approaches based only on the assessment of average or peak loading have proven to be inadequate, as they carry the risk of systematic under- or overestimation, respectively (Garg and Kapellusch 2016). Based on the viscoelastic nature of human tissue, Kumar (1990) argued that it would be necessary to assess the time history of loading. Accordingly, by taking into account the temporal aspect of loading, the relationship between work-related stress and physiological reactions could be better described than by focussing exclusively on peak exposures. In retrospect, Waters, Yeung, Genaidy, Callaghan, Barriera-Viruet, and Deddens (2006) see this reasoning as the starting point of the concept of cumulative loading following the general principle of integrating loading over time. Also, in view of recent advances in wearable sensor technology, which enables the collection of occupational exposure information with high frequency (Lee et al. 2017; Valero et al. 2016), cumulative loading seems an advantageous approach to using large amounts of exposure data for the evaluation of workplaces.

Based on Armstrong et al. (1993) and Checkoway, Pearce, and Kriebel (1989), Winkel and Mathiassen (1994) established a formal definition for cumulative loading resulting from physical exposure, according to which cumulative exposure is calculated as exposure level times duration, which corresponds to the SI unit Newton second and contradicts the use of exponential weighting factors (Waters, Yeung, Genaidy, Callaghan, Barriera-Viruet, Abdallah, et al. 2006). Multiplicative weighting factors are also not included. Instead, this non-weighted approach corresponds to the calculation of the area under the loading curve by multiplying the calculated load intensity by the respective load duration and summing these products. As a result, the non-weighted approach was widely used in the literature to assess work-related physical stress of the lower back (Andrews and Callaghan 2003; Callaghan, Salewytsch, and Andrews 2001; Fischer et al. 2007; Frazer et al. 2003; Gregory et al. 2009; Holmes, Hodder, and Keir 2010; Kumar and Narayan 2005; Lad et al. 2018; Parkinson, Bezaire, and Callaghan 2011; Sullivan, Bryden, and Callaghan 2002; Sutherland et al. 2008).

Still, the non-weighted approach is based on the assumption that different loading conditions have a sufficiently comparable effect on the human body as long as the area under the loading curve is the same. However, evidence derived from in vitro studies on the failure strength of human lumbar vertebra challenges this assumption and instead suggests a disproportionately higher overexertion risk of high forces (Brinckmann, Biggemann, and Hilweg 1988; Gallagher and Schall 2017; Jäger et al. 2000). This indicates that the non-weighted approach may not necessarily sufficiently reflect the impact on the human body posed by most working tasks. Huangfu et al. (2019) recently investigated the effectiveness of non-weighted calculation in vivo by examining muscular exposure data resulting from physical tasks with an identical area under the loading curve but a different combination of load intensity and duration. Overall, the results also indicated that a non-weighted approach carries the risk of systematically underestimating the effects of high load intensity. Nevertheless, muscular strain data between tasks of submaximal load intensity were not significantly different, suggesting that in some cases, the non-weighted approach may sufficiently reflect the effects on certain aspects of the human body. Given the submaximal nature of most modern work demands (Mathiassen 2006; Sonne et al. 2015), this is an interesting finding that requires further analysis. Still, Huangfu et al. (2019) examined muscles of the elbow flexor and therefore the transferability to the particularly endangered lower back seems unclear.

The purpose of this study is therefore to investigate the non-weighted calculation method for cumulative loading of the lower back with regard to its suitability to sufficiently reflect the resulting strain during lifting and lowering tasks. All tasks correspond to a medium exposure. The strain data analysed are muscular data with regard to muscular activity, heart rate characteristics, which reflect cardiovascular demand, and ratings of perceived exertion to capture the psychophysical aspect of physical work. In this context, the term strain is used in the sense of the stress-strain concept (Rohmert 1986) according to which the objective factors of a working situation that are equal for every person, such as the weight of loads to be lifted, are referred to as stress and the resulting response, which describes the demand on a person’s capacity, is referred to as strain. This investigation is of particular interest, since, to the authors’ knowledge, a systematic in vivo evaluation of the suitability of the non-weighted calculation approach with regard to the lower back including the resulting muscular, cardiovascular and perceived strain has yet to be carried out. For this purpose, strain data when performing lifting/lowering tasks featuring an identical area under the loading curve but a different combination of spinal load intensity and duration are compared. If strain data are overall significantly different, this would suggest that the non-weighted calculation does not sufficiently reflect the resulting strain. The spinal load intensity was biomechanically calculated as compression force of the intervertebral disc at L5/S1.

Based on the findings of Huangfu et al. (2019), Jäger et al. (2000), Gallagher and Schall (2017) and Brinckmann, Biggemann, and Hilweg (1988) regarding the general disproportionate effect of load intensity on the human body, we hypothesised that non-weighted calculation is not fully suitable for assessing work-related physical stress of the lower back even for the range of medium exposure intensity. Moreover, this study evaluates the relationship between both the spinal load intensity and the area under the loading curve on the indicators of muscular, cardiovascular and perceived strain and identifies further research needs for the development of a validated calculation method for cumulative loading of the lower back.

Method

Participants

Twelve participants (eight female and four male) participated in the study after signing the informed consent. The sample was not gender balanced since significant gender-specific differences are not expected when using relative strain values such as the normalised muscle activity (Kienbacher et al. 2014) or the heart rate reserve (Daugherty et al. 2011). Participation was limited to participants without recent (12 months) history of cardiovascular and musculoskeletal disorders. Prior to the study, the mean (standard deviation) age (23.7 (2.7) years), height (172.0 (7.0) cm) and weight (64.9 (10.2) kg) were recorded.

Experimental design

As shown in Table 1, a 2 (external weight) × 2 (area under the loading curve) within-subject design was employed. External weight was manipulated on two levels (1 kg or 6 kg), as was area under the loading curve (350 kNs and 700 kNs). The four resulting conditions were Low/Small, High/Small, Low/Large and High/Large. Every participant performed all four conditions. The first word indicates the level of external weight (low or high) and the second word indicates the level of area under the loading curve (small or large).

Table 1. The experimental design was a 2 × 2 within-subject design.

Condition	Combination	External weight	Area under the loading curve
Low/Small	Low external weight and small area under the loading curve	1 kg	350 kNs
High/Small	High external weight and small area under the loading curve	6 kg	350 kNs
Low/Large	Low external weight and large area under the loading curve	1 kg	700 kNs
High/Large	High external weight and large area under the loading curve	6 kg	700 kNs

The first word indicates the level of external weight and the second word indicates the level of area under the loading curve.

All test conditions corresponded to a medium exposure for which ergonomic design measures are reasonable, measured by the risk range assessed with the Key Indicator Method Lifting, Holding, Carrying (KIM-LHC) (Steinberg 2012). In each condition, participants performed a lifting/lowering task with the external weight held in both hands. The participants’ body posture was recorded at a rate of 30 Hz using a Kinect sensor (Kinect V2, Microsoft, WA, USA). Based on posture, body weight and external weight, the compression force of the intervertebral disc at L5/S1 was calculated for each frame using a biomechanical model based on the model of Jäger and Luttmann (1989) validated for material handling in the sagittal plane. The calculated spinal compression force represented the loading curve. The area under the loading curve was calculated in real-time by multiplying the calculated load intensity by the duration of each frame and summing these products. No weighting factors were used. Once the defined value of 350 kNs or 700 kNs was reached, the lifting/lowering task was stopped and a 10-min resting period started.

Figure 1 shows a schematic representation of the four conditions. When the area under the loading curve remains the same, but the weight level changes from small to high, this results in a higher spinal loading (dashed lines) and thus in a shorter task duration and fewer movement repetitions. When the weight remains the same, but the area under the loading curve is doubled, the task duration is also doubled due to the periodic nature of the loading curve.

Figure 1. Schematic illustration of the four conditions. The area under the loading curve is identical in each diagram. High external weight results in a higher spinal load intensity and therefore a shorter task duration as time until the defined area under the loading curve is reached. For conditions High/Small and High/Large the spinal load intensity is higher and therefore the task duration as time until the defined area under the loading curve is reached, is shorter.

Dependent variables

Bilateral muscle activity of the erector spinae longissimus (RES/LES), heart rate (HR) and ratings of perceived exertion (RPE) were collected. Since surface electromyography (EMG) is a suitable estimate for physical stress imposed by dynamic loads (Dolan et al. 1999), the muscle activity was measured using EMG during the lifting/lowering task and the resting period. RES/LES was selected as a representative of the back muscles whose function is the extension of the spine (Hermens 1999) and which is particularly strained during repetitive lifting lifting/lowering tasks (Dolan and Adams 1998). To monitor cardiovascular demand, HR was measured using electrocardiography (ECG) during the lifting/lowering task and the resting period. The Borg RPE scale (Borg 1998) was used to gather RPE with respect to the whole body after completing each task.

Experimental procedure

As shown in Figure 2, participants performed a two-dimensional, symmetric movement in the sagittal plane with an external weight held in both hands. Since empirical evidence suggests a significant influence of trunk inclination and shoulder angle on the muscle activity of RES/LES (Hellig et al. 2020), the movement was standardised accordingly. The trunk inclination was standardised to a range of 0° to 40° with the angle measured between the vertical line and the trunk longitudinal axis as shown in Figure 2. A stretched rope marked the lowest point of the movement. Touching the rope with the forehead signalled reaching of the lowest point. The arms were always perpendicular to the ground. The working pace was set to one lifting or lowering every three seconds and audibly signalled using a digital metronome. This working pace corresponds to a rate typical for occupational practice (Marras et al. 2006). Participants completed a short practice session in order to familiarise with the task and pace.

Figure 2. Symmetric lifting/lowering task in the sagittal plane with marked trunk inclination angle. The arms were always held perpendicular to the ground and the external weight was always held in both hands. Participants performed a lifting and lowering movement in the sagittal plane with a weight held in both hands. The trunk inclination was standardised to a range of 0° to 40°.

The order of the tasks was counterbalanced using a 4 × 4 Latin square design, and to minimise carryover effects, there was a 10-min resting period after every task. The participants spent each resting period sitting straight on a chair without backrest to avoid disturbances of the EMG signal with the arms placed in the lap and the feet on the floor.

The skin was prepared in accordance with SENIAM before attaching EMG and ECG electrodes (Hermens 1999). EMG electrodes were placed over the belly of the muscle under consideration of anatomical landmarks. ECG electrodes were placed according to the 3-lead system as described by Dupre, Vincent, and Iaizzo (2005). To stabilise the electrical condition, a minimum of 5 min passed between electrode placement and data collection. After warming up and visual inspection of the signal, maximum voluntary contraction (MVC) of RES/LES was collected. The MVC test was repeated after a pausing period of 60 seconds. As reference value for HR, HR_rest was determined as the mean value of a 1-min measurement in a sitting position after a resting period following the previous condition or the practice session. Figure 3 shows the overview of the measuring times of the different variables.

Figure 3. Overview of the measuring times of the different variables. Each task was followed by a resting period. Muscle activity and heart rate were measured continuously and RPE was obtained after completing the lifting/lowering task. As reference value, HR_rest was determined before each task after a resting period and in the same, sitting position. Muscle activity and heart rate were measured continuously, RPE was obtained after completing the lifting/lowering task and HR_rest was determined before each task after a resting period.

Data collection and processing

As the calculated compression force varies intraindividually because of the body weight, average minimum and maximum load intensity as well as resulting average task duration were calculated. The average maximum and minimum load intensity were calculated from individual average values for both levels of external weight. Average task duration was calculated as average of the time until the cumulative loading was reached for each condition. EMG and ECG data were recorded wirelessly by a Direct Transmission System (Desktop DTS Receiver, Noraxon, Scottsdale, AZ, USA). Signals were collected and processed using the software MyoResearch 3.12 (Noraxon, Scottsdale, AZ, USA). Ag/AgCl self-adhesive and pregelled 8-shaped dual EMG-electrodes (diameter of adhesive areas: 1 cm; inter-electrode distance: 2 cm) and single ECG-electrodes (diameter: 4.3 cm) were used to measure muscle activity of the RES/LES and HR, respectively.

EMG signals were hardware band-pass filtered (10–500 Hz) and sampled at 1500 Hz. For the normalised mean muscle activity, the data were rectified and smoothed using root-mean square (RMS) algorithm with a time window of 100 ms. RMS values were normalised to the MVC reference contraction. ECG signals were sampled at 1500 Hz using the DTS BioMonitor (Noraxon, Scottsdale, AZ, USA).

HR was analysed by calculating the percentage heart rate reserve (HR reserve) using the formula by Karvonen, Kentala, and Mustala (1957) as (HR_task − HR_rest)/(HR_max − HR_rest), with HR_task = mean HR of a lifting/lowering task and HR_max = 220 – age (Fox and Naughton 1972). Higher HR reserve values reflect an increased cardiovascular demand and a higher relative intensity of the lifting/lowering task. HR was further analysed with regard to the speed of recovery during the resting period. During recovery, the physiological functions related to recovery such as HR decrease exponentially, with the speed being proportional to the concentration of fatigue substances at that moment (Rohmert and Rutenfranz 1983). The rate of HR change was calculated as (HR₁ − HR₂)/1 min, where HR₁ = mean HR of the last second of a lifting/lowering task and HR₂ = mean HR of one second after the first minute of the resting period. A higher rate of HR change indirectly reflects higher concentration of fatigue substances.

Data analysis

Statistical analyses were conducted using IBM SPSS Statistics 24. A two-way within-subject design analysis of variance (ANOVA) was used for analysis. The independent variables were the four conditions (2 external weight × 2 area under the loading curve), and the dependent variables were the five measured indicators of strain, bilateral muscle activity of the erector spinae longissimus (RES/LES), heart rate (HR) with regard to HR reserve and rate of HR change as well as ratings of perceived exertion (RPE). Where the Shapiro-Wilk test indicated a statistically significant deviation from normality, non-parametric Friedman’s two-way ANOVA by ranks were applied instead. In case of a significant overall analysis, post hoc pairwise comparisons with a Bonferroni correction to adjust for an inflated family-wise error rate were used. If normal distribution could not be assumed, Wilcoxon signed-rank test with a Bonferroni correction to adjust for an inflated family-wise error was used for pairwise comparison. To assess the suitability of the non-weighted calculation method, strain data resulting from lifting/lowering tasks with the same area under the loading curve but a different external weight were compared. To assess the influence of doubling the area under the loading curve, strain data resulting from lifting/lowering tasks with the same external weight were compared. Significance was accepted at the α-level of p < .05. Spearman’s rank correlation coefficients were also calculated to analyse the direction of association between external weight and area under the loading curve and the five indicators of strain, bilateral muscle activity of the erector spinae longissimus (RES/LES), heart rate (HR) with regard to HR reserve and rate of HR change as well as ratings of perceived exertion (RPE).

Results

All participants were able to complete the four lifting/lowering tasks in the given pace and posture. Figure 4 shows the course of the calculated loading curve for conditions Low/Small and High/Small averaged over all participants. The minima and maxima correspond to the reversal points of the lifting/lowering movement as indicated in the figure. The continuously higher spinal loading due to the higher external weight results in a shorter time until the defined area under the loading curve is reached and therefore fewer movement repetitions. Due to the periodic nature of the loading curve, a doubling of the area under the loading curve leads to a doubling of the task duration for conditions (not depicted). The mean (standard deviation) duration of the four conditions was as follows: 223.4 (41.0) s for Low/Small, 194.8 (37.4) s for High/Small, 449.6 (82.5) s for Low/Large and 384.0 (63.9) s for High/Large.

Figure 4. Averaged course of the calculated loading curve for Low/Small and High/Small. The minima and maxima correspond to the movement. For an area of 700 kNs the task duration, as time until an area is reached, doubles. The maximum compression force for high and low external weight was 2805 N and 2466 N, respectively. For the small area, this results in a mean duration of 223 s for Low/Small and 195 s for High/Small.

Muscle activity

Figure 5 shows that the normalised mean muscle activity of RES/LES is similar during the resting period after a task (left). The mean muscle activity during the lifting/lowering tasks (Figure 5 on the right) shows differences especially between conditions with the same area under the loading curve but a different external weight (Low/Small and High/Small or Low/Large and High/Large).

Figure 5. Normalised mean muscle activity (± Standard error) of the resting periods (left) and lifting/lowering tasks (right). The muscle activity is at a similar level during resting period but significantly higher for conditions with high external weight despite the same area during the lifting/lowering tasks.

As the Shapiro-Wilk test indicated non-normality for muscle LES in conditions Low/Small (p = .033) and High/Large (p = .04), Friedman’s ANOVA was used to further analyse the normalised mean muscle activity. Friedman’s ANOVA indicated significant differences of the muscle activity between the different lifting/lowering tasks. Statistics are presented in Table 2. Pairwise comparisons in Table 2 show that the muscular strain differed significantly between conditions with different external weight despite the same area under the loading curve (Low/Small and High/Small or Low/Large and High/Large). As can be seen from the descriptive values in Table 3, the resulting mean muscle activity was on average 20% higher for the condition with higher external weight within the same area. Pairwise comparisons in Table 2 also show that a doubling of the area under the loading curve does not lead to a significant difference between conditions with the same external weight (Low/Small and Low/Large or High/Small and High/Large).

Table 2. Results of Friedman test for normalised mean muscle activity and Wilcoxon test for pairwise comparisons.

Variable	Total N	Test statistic	df	p Value
RES	12	29.700	3	<.001
LES	12	22.563	3	<.001
Variable	(I)	(J)	Test statistic	Error	Std. test statistic	Adj. p Value
RES	Low/Small	High/Small	−1.750	.527	−3.320	.005
	Low/Large	High/Large	−2.250	.527	−4.269	<.001
	Low/Small	Low/Large	.500	.527	.949	1.000
	High/Small	High/Large	<.001	.527	<.001	1.000
LES	Low/Small	High/Small	−1.958	.527	−3.716	.001
	Low/Large	High/Large	−1.542	.527	−2.925	.021
	Low/Small	Low/Large	−.167	.527	−.316	1.000
	High/Small	High/Large	.250	.527	.474	1.000

Adjusted p value: Bonferroni correction for multiple tests was used for adjustment.

Table 3. Mean and Standard error for normalised mean muscle activity.

Variable	Low/Small		High/Small		Low/Large		High/Large
	Mean	Std. error	Mean	Std. error	Mean	Std. error	Mean	Std. error
RES	16.460	1.910	19.558	2.224	16.106	1.912	19.417	1.988
LES	19.041	2.294	22.892	2.779	18.930	2.356	22.458	2.515

Heart rate (HR)

Figure 6 shows the mean HR reserve and rate of HR change with differences evident between conditions with the same area under the loading curve and different external weight.

Figure 6. Mean HR reserve (± Standard error) and rate of HR change (± Standard error). The mean HR reserve and rate of HR change is clearly higher for conditions with high external weight despite the same area during the lifting/lowering tasks.

Two-way repeated measures ANOVA revealed a significant effect of the external weight but no significant effects for area under the loading curve or their interaction. Statistics are presented in Table 4. As can be seen from the descriptive values in Table 5, the cardiovascular demand for conditions with high external weight was on average higher than for conditions with low external weight despite the same area under the loading curve. The HR reserve was about 20% higher for the condition with higher external weight within the same area. For the rate of HR change the percentage change was even higher with 46 % (High/Small to Low/Small) and 68 % (High/Large to Low/Large) despite relating to the same area under the loading curve.

Table 4. Results of two-way repeated measures ANOVA for effects of external weight, area under the loading curve and their interaction.

Variable	Predictor	Sum of squares	df	Mean squares	F	p Value	Partial η^2
HR reserve	External weight	49.946	1.000	49.946	5.084	.046	.316
	Error (External weight)	108.068	11	9.824
	Area under the loading curve	.209	1.000	.209	.015	.906	.001
	Error (Area under the loading curve)	156.332	11	14.212
	Weight*Area	.285	1.000	0.285	0.018	0.894	0.002
	Error (Weight*Area)	169.700	11	15.427
Rate of HR change	External weight	578,883	1.000	578.883	8.162	.016	.426
	Error (External weight)	780.173	11	70.925
	Area under the loading curve	.365	1.000	.365	.005	.947	.000
	Error (Area under the loading curve)	884.490	11	80.408
	Weight*Area	14.760	1.000	14.760	.501	.494	.044
	Error (Weight*Area)	324.146	11	29.468

Table 5. Mean and standard error for HRR and HRC.

Variable	Low/Small		High/Small		Low/Large		High/Large
	Mean	Std. error	Mean	Std. error	Mean	Std. error	Mean	Std. error
HR reserve	9.514	1.330	11.708	1.285	9.536	1.385	11.422	1.362
Rate of HR change	12.754	2.517	18.590	2.732	11.819	2.332	19.874	2.386

Ratings of perceived exertion (RPE)

As the Shapiro-Wilk test indicated non-normality for the RPE in condition Low/Small (p = .033), Friedman’s ANOVA was used to further analyse the RPE. Ratings from 8 to 16 were awarded on a scale of 6–20. Figure 7 shows the mean RPE. Friedman’s ANOVA indicated that the mean RPE changes significantly depending on the experimental condition. Statistics are presented in Table 6.

Figure 7. Mean RPE (± Standard error). The RPE is significantly higher for conditions with high external weight despite the same area during the lifting/lowering tasks.

Table 6. Results of Friedman analysis for RPE and Wilcoxon test for pairwise comparisons.

Variable	Total N	Test statistic	df	p Value
RPE	12	26.644	3	<.001
Variable	(I)	(J)	Test statistic	Error	Std. test statistic	Adj. p Value
RPE	Low/Small	High/Small	−1.458	.527	−2.767	.034
	Low/Large	High/Large	−1.875	.527	−3.558	.002
	Low/Small	Low/Large	−.333	.527	.527	1.000
	High/Small	High/Large	−.750	.527	−1.423	.928

Adjusted p value: Bonferroni correction for multiple tests was used for adjustment.

Pairwise comparisons in Table 6 show that the perceived strain differed significantly between conditions with different external weight despite the same area under the loading curve (Low/Small and High/Small or Low/Large and High/Large). As can be seen from the descriptive values in Table 7, the mean RPE was about 20% higher for the condition with higher external weight within the same area. Pairwise comparisons in Table 6 also show that a doubling of the area under the loading curve does not lead to a significant difference between conditions with the same external weight.

Table 7. Mean and standard error for RPE.

Variable	Low/Small		High/Small		Low/Large		High/Large
	Mean	Std. error	Mean	Std. error	Mean	Std. error	Mean	Std. error
RPE	10.000	.603	11.750	.730	10.333	.569	12.917	.633

Spearman’s correlation coefficients

Spearman’s correlation coefficients indicate a positive association between the external weight and the indicators of strain as well as overall no correlation between the area under the loading curve and the indicators of strain. Statistics are given in Table 8.

Table 8. Spearman’s correlation coefficients for the factors examined.

	Correlation coefficient
	RES	LES	HR reserve	Rate of HR change	RPE
External weight	.233	.218	.223	.386**	.449**
Area under the loading curve	−.018	−.016	−.014	0.10	.156

**Correlation is significant at the .01 level.

Discussion

In line with our expectations, the non-weighted calculation is not fully suitable for assessing work-related physical stress of the lower back even for the range of medium exposure intensity. For the normalised mean muscle activity, pairwise comparisons consistently show significantly higher values for conditions with higher external weight despite the same area under the loading curve. Results also prove the suitability of the resting position, as the mean muscle activity during rest is below 10 %MVC (Figure 5, left) and this value was previously defined by Qin et al. (2014) as the upper threshold for rest. During the lifting/lowering tasks, the mean muscle activity of LES was generally higher than the mean muscle activity of RES. This observation is consistent with other studies on the activation behaviour of the erector spinae (Feldwieser et al. 2012; Hagberg and Hagberg 1989; Zimmermann, Cook, and Goel 1993) and could be traced back to side-related differences in the proportion and recruitment of muscle fibres (Hagberg and Hagberg 1989). As such differences are not the focus of this study, they were not analysed further. Both HR reserve and rate of HR change show a significant dependency on external weight thus indicating a higher cardiovascular demand and concentration of fatigue substances. For RPE, higher external weight leads to higher ratings despite the same area under the loading curve.

As hypothesised, these results collectively suggest that the non-weighted method is not fully suitable for assessing work-related physical stress of the lower back, as it tends to underestimate the influence of higher spinal load intensity due to higher external weight even for medium exposure tasks. Results show that the resulting strain values are generally about 20–25% higher despite the same area under the loading curve. For the rate of HR change, which indicates concentration of fatigue substances in the organism, the differences are even 46% and 68%. The big difference between the indicators is striking. Therefore, before weighting factors for an alternative calculation method can be determined empirically, further research is needed on the question of which indicator represents the bottleneck for specific types of physical work activities, and should be used to calculate cumulative loading. Spearman’s correlation coefficients show the greatest correlation between load intensity and both rate of HR change (.386) and RPE (.460), although the RPE may to some extent be influenced by the use of handheld weights, since Le et al. (2012) have shown that psychophysical parameters when handling loads depend on tactile sensations of the hands. Due to the periodic nature of the loading conditions shown in Figure 4, a doubling of the area under the loading curve leads to a doubling of the duration of the lifting/lowering task. Spearman’s correlation coefficients in Table 8 overall indicate no tendency for the indicators of strain to increase when the duration of the lifting/lowering taks is doubled. Further statistical analysis also shows no significant differences when doubling the duration for conditions with the same external weight. However, some limitations to this study have to be considered. First, it should be noted that the lifting/lowering tasks examined here were activities of short duration and future research is advisable regarding long activities. This study also used a relatively small sample of participants. Still, the results clearly confirm that the non-weighted method is not fully suitable for assessing work-related physical stress of the lower back. Based on posture, body weight and external weight, the compression force of the intervertebral disc at L5/S1 was biomechanically calculated. The posture was recorded using a Kinect sensor. Diego-Mas and Alcaide-Marzal (2014) concluded that the Kinect sensor records back and arm positions reliably when in a frontal position and with minimal occlusion. Since the movements examined were standardised accordingly, an accurate recording of the body posture can be assumed. Whereas body height and weight were considered in the biomechanical calculation, dynamic effects during the movements of body parts and the external weight were neglected. However, while the absolute load values likely would have increased (Howarth, Beach, and Callaghan 2010), dynamic effects are not relevant here, as the focus is on the relative comparison of strain indicators and the speed of movement was kept constant for all participants. Further investigation of the influence of the lifting frequency in the context of the aggregation of physical exposure over time seems reasonable.

Moreover, muscular, cardiovascular and perceived strain were measured. In this context, parameters of muscle activity in particular are recommended as an indicator of strain when designing jobs (Gallagher and Schall 2017; Nussbaum 2001). Possible strain of spinal musculoskeletal tissue, however, was not determined. As presented by Gallagher and Schall (2017), strain of musculoskeletal tissue is likely to play an important role with respect to work-related MSD, as growing evidence suggests that MSD may be a result of damage propagation of musculoskeletal tissue. It, therefore, seems appropriate to include failure behaviour of musculoskeletal tissue in a yet to be determined alternative calculation method, so that in addition to muscular, cardiovascular and perceived strain, the actual injury risk can be quantified. However, as stated by Gallagher and Schall (2017), the maximum strength of musculoskeletal tissue varies depending on the loading condition. Consequently, in order to include injury risk of musculoskeletal tissue when assessing time-varying loading, research is required regarding the maximum strength and damage propagation of musculoskeletal tissue depending on the loading condition.

Further research is also needed regarding limit values against which a calculated value can be assessed. According to Li and Buckle (1999), the decision on the necessity of an ergonomic intervention is one of the main functions of ergonomic assessment methods. For this purpose, limit values for both maximum cumulative exposure during a working day and a life-time cumulative exposure limit are needed. When determining a life-time exposure limit, in addition to tissue damage, tissue repair must be considered (Gallagher and Schall 2017; Waters, Yeung, Genaidy, Callaghan, Barriera-Viruet, Abdallah, et al. 2006).

Conclusion

The purpose of this study was to investigate the non-weighted calculation method for cumulative loading of the lower back with regard to its suitability to sufficiently reflect the resulting muscular, cardiovascular and perceived strain. As hypothesised, the results indicate that the widely used non-weighted calculation approach does not fully reflect the resulting muscular, cardiovascular and perceived strain but seemingly underestimates the influence of higher load intensity due to higher external weight even for medium physical stress.

Further research is required with regard to the question of which indicator represents the actual bottleneck for specific types of physical work activities. Based on this, weighting factors for the load intensity relative to the load duration could be determined in order to define an alternative calculation method that better reflects the impact of a working task on the human body. Research is also required regarding the investigation of longer-lasting tasks and limit values for cumulative loading. Furthermore, a better knowledge of the failure behaviour of musculoskeletal tissue depending on the loading condition is needed, as this may enable to include the actual risk of WMSD in the calculation of cumulative loading.

Acknowledgments

This article is part of the projects ‘INDIZ’ (grant number 16ITA206) and ‘workHEALTH’ (grant number 01EC1905B), which are funded by the German Federal Ministry of Education and Research (BMBF). The authors would like to thank for the support they have received.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This article is part of the projects ‘INDIZ’ [grant number 16ITA206] and ‘workHEALTH’ [grant number 01EC1905B], which are funded by the German Federal Ministry of Education and Research (BMBF).