Variability of lumbar spinal alignment among power- and weightlifters during the deadlift and barbell back squat

ABSTRACT The aims of the study were to evaluate the relative and absolute variability of upper (T11-L2) and lower (L2-S2) lumbar spinal alignment in power- and weightlifters during the deadlift and back squat exercises, and to compare this alignment between the two lifting groups. Twenty-four competitive powerlifters (n = 14) and weightlifters (n = 10) performed three repetitions of the deadlift and the back squat exercises using a load equivalent to 70% of their respective one-repetition maximum. The main outcome measures were the three-dimensional lumbar spinal alignment for start position, minimum and maximum angle of their spinal alignment, and range of motion measured using inertial measurement units. Relative intra-trial reliability was calculated using the two-way random model intraclass correlation coefficient (ICC) and absolute reliability with minimal detectable change (MDC). The ICC ranged between 0.69 and 0.99 and the MDC between 1°-8° for the deadlift. Corresponding figures for the squat were 0.78–0.99 and 1°-6°. In all participants during both exercises, spinal adjustments were made in both thoracolumbar and lumbopelvic areas in all three dimensions. In conclusion, when performing three repetitions of the deadlift and the squat, lumbar spinal alignment of the lifters did not change much between repetitions and did not differ significantly between power- and weightlifters.


Introduction
Deviations from a neutral spinal alignment during performance of the barbell back squat, henceforth referred to as the squat, and deadlift strength training exercises could be a risk for future low back pain (Sjoberg, Aasa, Rosengren, & Berglund, 2018). Since the interest of strength sports is growing worldwide, and injuries in the low back are common in powerlifters (Stromback, Aasa, Gilenstam, & Berglund, 2018) and are becoming more common in weightlifters (Burekhovich et al., 2018), physiotherapists often treat patients whose pain is associated with either performance of the squat or the deadlift. In the process of rehabilitation, the physiotherapists analyse the movement patterns of lifters during performance of the pain-provoking exercises to examine whether side bending, rotating or flexing of the lumbar spine during the lift is associated with the lifters' pain experience. If an increased relative flexibility (Sahrmann, Azevedo, & Dillen, 2017) in either of these directions is found, the functional impairments causing it are targeted in the rehabilitation. Excessive movements in these planes are also considered contraindicated for a number of other reasons, not least the correlation between spinal alignment and degenerative changes. For example, it has been suggested that axial twisting in combination with repetitive flexion-extension motion might predispose joint diseases such as vertebral stress fractures (Leone, Cianfoni, Cerase, Magarelli, & Bonomo, 2011) as well as bulging of the lumbar discs and herniation (Marshall & McGill, 2010).
When evaluating movement patterns for the squat and deadlift, it is important to remember the inherent variability evident within the human movement system (Hamill, Palmer, & Van Emmerik, 2012). In fact, the movement system has the ability to spontaneously reorganise movement coordinative strategies in a variety of ways to adapt to external and internal factors (functional variability) (Davids, Glazier, Araujo, & Bartlett, 2003). For example, the ability of the lifters to always initiate the deadlift without rounding their lower back could be affected by internal factors such as fatigue (Santamaria & Webster, 2010). Repeatability is the opposite to variability. Repeatability practices were introduced by Bland and Altman (1999). Relative intra-trial repeatability could be used to evaluate the consistency of the spinal alignment within a training session when performing repetitive repetitions of the squat or deadlift, and hence gain a better understanding of whether lifters perform the exercises in a similar manner across repetitions. To our knowledge, only one previous study has investigated the relative reliability of repeated measures of spinal alignment during squat performance (McKean, Dunn, & Burkett, 2010). In that study, calculation of the intrarater test-retest relative reliability was reported with good (>0.7) to excellent (>0.9) intraclass correlation coefficients (ICC) indicating low intraindividual variability of the lumbar and pelvic movements between repetitions when the squat was performed with 50% of body weight. Further, in addition to the relative reliability, the absolute reliability can be used to calculate the minimal detectable change (MDC) in order to reveal whether measured variations are genuine changes or not (Bland & Altman, 1996). No previous study has examined the absolute intra-trial reliability or MDC of the spinal alignment during deadlift or squat performance.
A growing body of scientific literature has investigated the utility of inertial measurement units (IMUs) for monitoring of spinal alignment during resistance training (Gleadhill, Lee, & James, 2016) and to evaluate exercise technique (Taylor, Almeida, Hodgins, & Kanade, 2012). IMUs are portable tri-axial motion sensors (containing gyroscope, accelerometer and magnetometer) that are attached on the skin and can be wirelessly connected to a computer. They have been shown to be as effective as markerbased systems at measuring joint angles in sport (Bonnet, Mazza, Fraisse, & Cappozzo, 2013) and work-related (Stenlund et al., 2014) tasks, and have been shown to distinguish between acceptable and aberrant squat techniques with excellent accuracy and to identify exact technique deviations (O'Reilly, Whelan, Ward, Delahunt, & Caulfield, 2017).
The deadlift and the squat are essential exercises among both power-and weightlifters (Wretenberg, Feng, & Arborelius, 1996). However, since they are also two of three competitive lifts in the sport of powerlifting (Keogh, Hume, & Pearson, 2006;Siewe et al., 2011;Swinton, Lloyd, Agouris, & Stewart, 2009), but not in weightlifting, powerlifting and weightlifting athletes might use different lifting styles and thereby move differently in their lumbar spines. Today, it is unknown whether they show different spinal alignment during the lifts.
The present study aimed to 1) evaluate the relative and absolute variability of the spinal alignment during both the squat and deadlift exercises, and 2) determine whether power-and weightlifters show different spinal alignment in their lumbar spine.
We hypothesised that 1) the movement patterns of the lifters would not change between three repetitions when performing the lifts at approximately 70% 1RM and that 2) competitive power-and weightlifters would show similar spinal alignment during the lifts.

Study design
This cross-sectional study measured the tree-dimensional movements in the upper (thoracolumbar, T11-L2) and lower (lumbopelvic, L2-S2) lumbar spine with IMUs when performing the squat and deadlift exercises. Data collection was performed in a gym. Spinal alignment was measured during the performance of one set of three repetitions of 70% of self-estimated one-repetition maximum (1RM) squats and deadlifts.

Participants
Fourteen powerlifters (men = 10, women = 4) and 10 weightlifters (men = 4, women = 6) were recruited from local clubs in Umeå, Sweden (for characteristics, see Table 1). Inclusion criteria were power-and weightlifters ≥150 cm in height with at least two years of lifting experience, without injuries that could affect performance, and with the intent of competing. To ensure that eligibility criteria were met, all participants completed a questionnaire. They also signed an informed consent form prior to participation. At the day of data collection, one of the women powerlifters chose not to perform the squat. Written informed consent was obtained from all lifters prior to participation and the study was approved by the Regional Ethical Review Board of Umeå, Sweden (Dnr 2014-285-3M).

Procedures
First, the participants completed a self-administered warm-up with the intention to be prepared for heavy deadlifts. The warm-up typically consisted of sub-maximal deadlifts with increasing loads. Thereafter, three calibrated IMUs were affixed to their backs using double-sided tape and self adherent wraps at the following anatomical landmarks: Processus spinosus Th11 and L2, and Sacrum (S2) (Figure 1). The placements of the units were palpated with the lifters standing erect by the same experienced person. Before commencing the respective data collections of the deadlift and squat exercises, the lifters further performed one set of the relevant exercise to ensure that the IMUs did not hinder performance.
For data collection, the lifters performed one set of three repetitions at 70% of his/her self-estimated 1RM for each exercise. The lifters stood holding the barbell, which was placed on the floor in front of them, with straight arms and an optional grip and with flexed knees and hips (Start position deadlift). In accordance with the rules of the International Powerlifting Federation (2019), the barbell was then lifted by extension of the knees and hips until the lifter was standing erect with their shoulders back (Stop position). Once the lifter was motionless in the Stop position a down signal was given by the test leader and the lifter then lowered the barbell to the ground before releasing their grip. The lifters were instructed to stand erect momentarily before beginning with the next repetition. Since only conventional style deadlifts were allowed, so that the measurements would be uniform, all participants used a conventional deadlift style where the barbell is held with the arms laterally to the legs. For the squat, participants were standing erect (Start position) with the barbell on their shoulders and instructed to descend by flexing at the hip, knee and ankle joints until the crease of the hip was lower than the top of the knee in accordance with the rules of the International Powerlifting Federation (2019). From the bottom position, the participants ascended to the start position by extending the same joints. The lifters were instructed to stay in this position until a start signal was given by the test leader to begin the next repetition. Notably, apart from the practical instructions, no more instructions were given. The basic premise of this form of evaluation is to measure the individual movement and joint loading patterns of each participant (Nielsen et al., 2017). No equipment (knee wraps, belts) other than wrist wraps was allowed. The lifters were allowed to use chalk and, if preferred, they could use alternated hand grip for the deadlift.

Instruments and measurements
We wanted our findings to be representative for the real-life environments (Dingenen & Gokeler, 2017), hereby acknowledging the importance of the environmental and task constraints within the dynamic system theory (Davids et al., 2003;Holt, Wagenaar, & Saltzman, 2010). We used a portable movement analysis system in a gym to assess spinal alignment in the thoracolumbar and lumbopelvic areas during the lifts. The system included three tri-axial IMUs (MoLab TM POSE, AnyMo AB, Umeå, Sweden) that were wirelessly connected using WiFi to a computer equipped with the software MoLab TM measure (AnyMo AB, Umeå, Sweden). Each sensor was 60 × 45 x 10 mm (length x width x height), included a three-dimensional gyroscope, accelerometer and magnetometer and weighed 14 g. The sampling frequency was 128 Hz with a 16-bit resolution and an antialiasing low pass filter set at 64 Hz. The full-scale range was ±1000°/s for the gyroscopes, ±8 g for the accelerometers and ±4800 µT for the magnetometers. The measurement precision and accuracy of the MoLab TM POSE system for measurements in the spine has been validated against a gold standard optical system (Ertzgaard, Ohberg, Gerdle, & Grip, 2016;Öhberg, Lundström, & Grip, 2013). Outcome measures were based on the IMUs detection of three-dimensional spinal alignment and real-time orientation (Öhberg et al., 2013). The IMUs sent information regarding their orientation relative to each other. We recorded both the thoracolumbar (measured as the angle between the IMU on processus spinosus Th11 and the IMU on processus spinosus L2) and lumbopelvic (measured as the angle between the IMU on processus spinosus L2 and the IMU on processus spinosus S2) angles since the human body functions as an integrated series of highly interacting multiple segments across multiple planes within a kinetic chain.
Five measures for the deadlift and four measures for the squat were selected to quantify the spinal alignment in the thoracolumbar and lumbopelvic areas, respectively: (1) Start position, (2) Stop position (deadlift only) (3) Min angle (the minimum angle in degrees [°] during each exercise), (4) Max angle (the maximum angle in degrees during each exercise) and (5) range of motion (ROM) (difference in degrees between the min and max angles during each exercise).

Data handling and statistical analysis
Orientation data (i.e., segment angles) from the IMUs were processed in MoLab TM analysis (AnyMo AB, Umeå, Sweden). The Euler sequence used for the segment angles were X (rotations in the sagittal plane), Y (rotations in the frontal plane), and Z (rotations in the transverse plane). A more detailed description of the used algorithms can be found in Öhberg et al. (2013).
Statistical analyses were performed using the Statistical Package for the Social Sciences (SPSS) version 23 (IBM Corp., Armonk, NY, USA). Intrarater test-retest relative reliability and absolute reliability were used as outcome variables. Intrarater test-retest relative reliability was estimated by ICC and ICC values <0.75 represented poor to moderate reliability, 0.75-0.90 good reliability, and 0.91-1 adequate reliability for clinical measurement (Portney & Watkins, 2014). The ICC was calculated with a 95% confidence interval (CI) and a two-way random model was used since tester and participants were considered random effects. The absolute reliability was calculated as described by Bland and Altman (1999) using all three measurements. Accordingly, a one-way analysis of variance was carried out to obtain the within-group residual mean square. The square root of the withingroup residual mean square is the within-subject standard deviation (s w ). The minimal detectable change (MDC) is based on s w and is calculated using the equation √2 * 1.96 * s w . This provides information about the MDC with a 95% CI.
A factorial-repeated measures analysis of variance (ANOVA) was conducted to compare the influence of the independent variable (Group 1 = powerlifters and Group 2 = weightlifters) and the effects of the depended variables. For the deadlift, segment angle 'SegAng' at four different time points (Point 1 = Angle in Start position, Point 2 = Angle in Stop position, Point 3 = Minimum angle at any time point, Point 4 = Maximum angle at any time point) and total ROM for the sagittal, horizontal and frontal planes, respectively, were chosen as dependent variables. For the squat, segment angel at three different time points (Point 1 = Angle in Start position, Point 2 = Minimum angle at any time point, Point 3 = Maximum angle at any time point) and total ROM for the sagittal, horizontal and frontal planes, respectively, were chosen as dependent variables. The analyses were performed for both the upper and lower lumbar spine areas. For each time point, we used the mean values for the three repetitions. Sphericity was calculated using Mauchly's test of Sphericity. If sphericity was not assumed, a correction was made using the Greenhouse-Geisser estimation. If significant within-participants effects were found, post-hoc pairwise comparisons were calculated. Effect sizes of within-participants effects were calculated with partial eta squared.
Significance level was set at 0.05 and Bonferroni corrections were performed for multiple comparisons.

Results
The relative intra-trial reliability analysis for the deadlift (Table 2) showed adequate ICC values for 16 of the 24 measures and good ICC values for seven measures. For the squat (Table 3), the corresponding figures were 18 of 24 and six of 24, respectively. Table 4 shows the values of MDC for the measures included in the description of the spinal alignment. For the deadlift, the smallest amount of difference in individual scores that represents a true change for ROM was 8°in the upper and 4°in the lower lumbar spinal areas. For the squat, the corresponding figures were 4°in the upper and 5°in the lower lumbar spine.
The three-dimension angles of the upper and lower lumbar spine during the deadlift for the Start position, Min angle and Max angle, and ROM are presented in Tables 5 and 6
in the lower lumbar spine during the squat (total range of motion was on average 18°), the change has to be greater than 5°to be a true value. Beforehand, we hypothesised that although competitive power-and weightlifters use different lifting styles (powerlifters generally perform the low-bar back-squat that typically require them to increase their upper body forward lean whereas weightlifters preferably perform a high-bar squat and strive to keep their upper body upright), would show similar spinal alignment. This was confirmed in the analyses. It was shown that in all participants during both exercises, spinal adjustments were made in both thoracolumbar and lumbopelvic areas in all three dimensions. This finding should be contrasted to the findings of Sjöberg et al. (2018) where an expert panel of eight powerlifting experts (researchers, coaches, lifters) identified side bending and twisting of the lower back during the deadlift and squat exercises as important risk factors for the development of injuries. Further risk factors included a flexed lower back during initiation of the deadlift and a flexed lower spine in the bottom position of the squat. This is also in line with instructions in gym settings, and when the deadlift exercise is included in rehabilitation (Aasa, Berglund, Michaelson, & Aasa, 2015;Michaelson, Holmberg, Aasa, & Aasa, 2016), where it is commonly indicated that the spine should be kept in 'its neutral spinal alignment' during the lifts. A neutral spinal alignment is, however, not necessarily a singular, static position, but more of a zone or 'a region of intervertebral motion around the neutral posture' as described by Panjabi (1992aPanjabi ( , 1992b. In this zone, the load is equally distributed on the tissues, whereas in the outer range of a motion the load will be unequally distributed between the loaded tissues, which might eventually lead to microtrauma in areas that are excessively loaded (McGill, 2001;Sahrmann et al., 2017). There Table 8. The three-dimensional angles (SegAng) in degrees [°] of the lower lumbar spine (lumbopelvic region) during the squat for the Start position, Minimum (Min) angle, Maximum (Max) angle and range of motion (ROM) as well as results of the two-way factorial-repeated measures ANOVA (withinparticipants effect) in all lifters (n = 24) and separately for the powerlifters (n = 13) and weightlifters (n = 10). seem, however, not to be described in scientific literature what angles of movements in the sagittal, frontal and horizontal planes that could be considered harmful.
In the present study, the highest ROM during both the deadlift (22°) and the squat (18°) was found for flexion-extension movements in the lower lumbar spine, but spinal adjustments were made in all three dimensions also in the upper lumbar spine. The mean ROM of flexionextension motions indicates that most powerlifters and the weightlifters keep their spines within their neutral zones. The reason why we conclude so is that it has been described that the lumbar lordosis of the unloaded lumbar spine of a standing person is about 25°and the normal ROM for flexion from this point is about 50° (Ng, Richardson, Kippers, & Parnianpour, 2002) (or in other words from upright standing the spine can flex about 80° ( Dvorak, Vajda, Grob, & Panjabi, 1995)). The participants' starting position as measured with the IMUs was at about 8°in the upper and 12°in the lower lumbar spine, and their total ROM was about 8°and 18°for the upper and lower lumbar spine, respectively. For the deadlift, the mean total ROM was 22°and the stop position 16°. The standard deviation figures indicate, however, that some of the lifters might be close to the outer range of flexion in the start position.
Notably, according to the dynamic system theory (Davids et al., 2003), the ability to maintain the spine in its neutral zone depends on both intrinsic and extrinsic factors and may therefore vary between individuals. Examples of intrinsic factors are the individual lifters' pattern of relative flexibility, for example, between the lumbar spine and hip joints (Sahrmann et al., 2017), neuromuscular activation patterns, muscle strength and tissue tolerance. Regarding relative flexibility, the body is considered as a linked system of interdependent segments achieving the desired movement in an efficient manner (Karandikar & Vargas, 2011) and each segment in this system influences the motions of its adjacent segments (Sahrmann et al., 2017). Therefore, if a lifter is relatively stiffer in the hips than in the lumbar spine, the lumbar spine will more readily move into flexion compared to another lifter with less stiff hips or more stiff lumbar spine (Sahrmann et al., 2017). It is important though to differentiate whether rounding of the back is due to stiffness in the hips or whether the muscles surrounding the lower back are relatively too weak (and are therefore not able to maintain neutral spinal alignment). Also, it must be noted that there are successful deadlifters that seem to use back-dominant lifting techniques with visually rounded backs to shorten the external moment arm (Cholewicki & McGill, 1992). This shorter external moment arm allows the lifters to handle heavier weights. The importance of moving away from the zone of neutral posture may vary between individuals due to many factors (Davids et al., 2003). Regarding elite lifters who lift with visually rounded backs and have no low back injuries, there might also be a source of selection bias ('healthy worker effect' (Shah, 2009)), i.e., only lifters whose tissues withstand heavy flexion loading are still competing at this level. They probably have a tremendous tissue tolerance since they have been practising for a long time. Namely, the human movement system has a great ability to adapt to tissue loading to maintain tissue homoeostasis and function (Hodges & Smeets, 2015); the more the tissues are loaded the stronger they become. Sometimes it is therefore argued that it might not be important to maintain the lumbar spine in its neutral zone since 'repetitive flexion-extension loading' strengthens the tissues (Lehman, 2018). However, all tissues have a breaking point (Marras, Davis, Ferguson, Lucas, & Gupta, 2001) (although tissue breakdown might not always be associated with low back pain (Brinjikji et al., 2015)), but the issue is whether all tissues can adapt and be strengthened indefinitely or if there is a maximum, or a maximum ROM, at which tissues cannot be further strengthened. To this date, no study has investigated the ultimate potential for tissue strengthening during power and weightlifting.

Limitations
When calculated for powerlifters and weightlifters separately, the ICC values decreased for both groups. This could be explained by the smaller sample sizes when the groups were analysed separately.
We wanted to investigate the variability between repetitions. However, the exact number of repetitions needed to have an appropriate outcome measure is not straightforward and dependent on the activity, the subject and the variable under investigation (Preatoni et al., 2013). We chose three repetitions to ensure that the total number of lifts would stay within the lifters capacity and to avoid introducing fatigue which would have undoubtedly changed the kinematics during the set (Hooper et al., 2014).

Conclusions
Despite the general consensus that a neutral spinal alignment should be maintained during execution of the deadlift and squat exercises, our study found that when experienced powerlifters and weightlifters perform three repetitions of each exercise at approximately 70% 1RM, they adjust their lumbar spinal alignments in all three planes. However, the three-dimensional spinal alignment adjustments show low variability and do not seem to reach outer ranges of lumbar spinal flexion or extension for the squat. Lumbarspinal alignment of the lifters did not differ significantly between power-and weightlifters.