Following participation in our earlier research 14151617, five high schools volunteered for this project. This provided a study population of approximately 7500 students (around 1500 students per school).

Sample selection was by random numbers applied to class lists, seeking five males and females from each year level in each school. Student names were excluded from the class lists by teachers prior to selection if they were known to have recent fractures or sprains anywhere in the body, musculoskeletal or neurological diseases, or were unable to stand upright on two feet for testing. Selected students, in conjunction with their parents, consented in writing. Further screening was undertaken prior to testing by the research team, to eliminate consenting students who had scoliosis or kyphosis.

Sample size was based on findings from our previous investigations into backpack load and head-on-neck posture 11 (measured as the craniovertebral angle (CVA)14). This is reputedly a robust posture measure for adults 9111213 although it has been less frequently reported for adolescents 1519. Power analysis 18 incorporating significance of 0.05, power of 0.80, a standard deviation of four degrees, and an expected change of CVA of at least five degrees under an experimental condition, suggested a minimum sample of 225 to detect significant change between baseline and experimental conditions, within subgroups of gender and year level. For ease of trial management, we tested 250 students (50 students per school reflecting 5 girls and 5 boys from each of the five high school year levels.

We tested nine experimental conditions (combinations of percentages of body weight (three, five, ten per cent), carried in one backpack positioned with its centre at upper (T7), middle (T12) and lower spinal (L3) positions (backpack test positions illustrated in Figure 1).

To position the backpack in the three test positions, shoulder straps were adjusted according to subject height to lift the backpack into the appropriate position on the spine. The experimental conditions were applied using a block (Latin square) design (illustrated in Table 1), which provided efficiencies for treatment administration and calculation of condition order effects 18. We also took three measurements of unloaded (baseline) postures for comparison measured prior to commencement of testing, after intervention four, and at completion of the testing block. In this way, subjects acted as their own controls, where baseline (control) posture could be compared with posture under the different experimental treatments.

Validated study protocols were derived from our earlier research 14, and were compiled into an instruction manual which is available form the authors at no cost 26. We used the one backpack for all tests. It had features that were typical in the majority of school backpacks observed in our cross-sectional study 14, comprising a soft backpack with no internal framing or back support, two padded adjustable shoulder straps, no internal compartments, one size only and with no waist, chest or load (side) compression straps. Its capacity was 35 litres. Experimental loads were applied as percentage of subject body weight to control for the effect of body weight within our sample. Subjects were weighed on the one set of regularly calibrated electronic scales (Mettler TE120) (accurate to within 0.001 kg, up to 120 kgs). Subject-specific loads were assembled from a range of weights whose dimensions replicated typical educational material. Loads were fixed within the backpack at 100 mms of the spine, to standardise load displacement in the backpack.

Measures of sagittal standing posture are commonly used estimates of the response of the human body to its environment 8911132025, and are accurately measured by photograph 1112132025. Still photographs provide point-in-time posture measures, which can be quantified as the position of landmarks on the body against standard references. One Canon SLR camera (EOS 500), loaded with Fuji ASA 100 film, was used to take still photographs of subjects' sagittal postural response to experimental conditions. We had previously found that this camera and film type minimised distortion in image capture 26. The camera was mounted on a tripod (Hama Profil 76) set at 3.1 metres and at 90 degrees to the right of the subject's midline to accommodate the tallest subject. The three planar positions of camera on tripod were monitored throughout testing by spirit levels 26.

Subjects wore bathers or tight 'bike' shorts and a sleeveless T-shirt top. Lateral anatomical landmarks were marked by adhesive paper dots contrasting to skin colour. They comprised the tragus of the ear, mid-acromion of the shoulder, lateral superior iliac crest, greater trochanter, head of fibula and lateral malleolus of the ankle, and a projecting reflective marker was placed posteriorly on the spinous process of C7 891011121320.

Subjects were allocated sequentially to the Latin square order of testing by an independent research assistant. A second research assistant applied all the experimental treatments. One photographer took a sagittal photograph for each of the 12 conditions (nine experimental treatments and three baseline measures). Subjects stood in a standard position for each photograph 26. The research assistant removed the backpack after each experimental treatment and subjects were encouraged to relax and move about. This person also checked the sagittal position of each subject to correct for segmental rotation, prior to the next photograph being taken. Photographs were taken three seconds after positioning the backpack on the subject.

Photographs were developed as negatives which were scanned as electronic computer files using a 35 mm film scanner (Nikon LS 2000). We used standard positions of the negative within the scanner 26, scanning speeds and image resolution.

Vertical and horizontal coordinates were calculated of the centre of each anatomical landmark on each photograph by the one measurer (digitiser) using digitising software 27 and using standard protocols for digitising 26. All coordinate values were standardised to the ankle (representing x = 0, y = 0), allowing comparison of posture change between and within subjects.

Our pilot studies indicated that by strict use of the protocols, the error of measurement could be estimated as two pixels at each anatomical point in the photograph.

Subjects could not be blinded in this study to the test conditions, but they were blinded to the hypothesised outcomes. We attempted to blind the photographer and digitiser as much as possible to experimental backpack weights, order of administration of weight and hypothesed outcomes.

There is no one standard or accepted method for interpreting sagittal posture change 8910111219202122232425. This paper reports investigation of change in the horizontal (x) displacement of each anatomical point, relative to baseline, and also to each experimental treatment. Repeated measures ANOVA models were applied to test differences between the measurements of unloaded posture (2 degrees of freedom), seeking a difference of less than error in measurement (2.5 pixels) to indicate consistency in postural positioning in this condition. Multiple ANOVA models were applied to test the effect of experimental conditions relative to baseline measures, with independent variables of:

♦ weight contrasts (3 degrees of freedom) (unloaded position with measures made under 3%, 5% and 10% loads)

♦ backpack position contrasts (2 degrees of freedom) (the backpack positioned at T7, T12 and L3)

♦ the interaction between loaded weight and position (4 degrees of freedom)

♦ the effect of gender, year level and order of testing.

Recruitment was more difficult from students in Years 9 and 10 than in the other year levels. Measurement of posture using photographs may have been a disincentive to these adolescents whose growing bodies may be a source of embarrassment 23. This may explain refusals by many 13–15 year old girls and boys, despite parental consent, and assurances of privacy and confidentiality. Although it is possible that our Year 9 and 10 subjects (14–15 years old) had better body image than non-participants, the lack of age or gender differences across our sample suggests that the postural responses of these students were no different than those of the subjects in the other year levels.

Potential differences in adolescents' posture due to gender-specific rates of pubertal growth 23 underpinned our within-subject approach to determining experimental conditions, as opposed to testing standard loads or heights of backpack placement on the trunk. The lack of year level or gender influences on postural responses suggested that our design adequately controlled for these effects. This finding indicates that generic backpack load recommendations for adolescents could be developed in future studies, provided the recommendations are based on loads and spinal positioning that are relative to the individual's weight and height.

We did not attempt to identify subjects with 'poor posture' either at baseline or under experimental treatments. We sought only to quantify the within-subject differences related to the experimental conditions. Despite the common acceptance that 'poor' sagittal adult standing posture is not closely aligned with the gravitational line 8910, there is no standard threshold of deviation as a marker of 'poor posture'. Thus it may be that subjects with 'poor posture' had a different response to experimental conditions than subjects with posture that was more closely aligned with a vertical reference. We are currently undertaking this investigation.

Positioning the backpack high on the spine produced largest postural response at all anatomical points. This finding contradicts the 'rule-of-thumb' that higher load positioning is better. These findings also support our observations of common practice by adolescents regarding lower carriage of backpacks 14. Lower backpack positions on the adolescent spine plausibly approximate load to the individual's centre of gravity 67, which we suggest requires less postural adjustment to maintain the body's position in space 89. Compared with other landmarks, the lack of a corroborating effect of the largest horizontal effect of the T7 backpack position at the shoulder may relate to the multidirectional construction of the shoulder joint, and local mechanical effects from approximation of this joint this backpack backpack position.

Adolescents' response to increasing posterior load, irrespective of backpack position, was to place all anatomical points progressively more anterior to the ankle. Shoulder displacement appeared to determine movement of all points above. Our findings support those from the literature, in which any posterior load produces a different sagittal position from unloaded 671925. This is plausible because of the need to adjust the body's centre of gravity to accommodate a posterior load. What we did not find was a difference in postural response of 10% body weight, compared with less weight. Thus the hypothesis that loads should be limited to 10% body weight could not be supported. It appears from this study that there is a dose-response relationship between backpack load and horizontal placement of sagittal plane anatomical points. It is implausible however, that this dose-response would continue with increasing loads at all anatomical points above the ankle, because of the body's requisite adjustment of the centre of gravity to accommodate the load 9. Therefore, further testing is required of postural responses to greater loads, relative to body weight. Ethical permission for such testing on adolescents would require assurances that no harm would occur to them from spinal 'overloading'. Despite common practices by adolescents demonstrated by our cross-sectional study 14, where 10% body weight was the average load carried, rather than the greatest load carried, such assurances would require more knowledge than is currently available regarding spinal tissue responses to posterior loading.

We chose sagittal anatomical points to allow us to visualise the postural response of each body segment. These points had been reported previously 891011121314 and allowed us to take a comprehensive view of whole body and inter-segmental responses to posterior loading. However, the position of the backpack precluded measurement of anatomical points on the thorax or low back. Measurements from these areas would provide further information on the body's postural response to repeated positioning of a posterior load, particularly the inter-segmental response to posterior loads. Research is required to develop processes to measure landmarks on the anatomical areas whilst they are covered by a backpack.

Backpacks featuring more sophisticated load carrying systems and/ or load reduction features may perform differently from our 'typical' school backpack, potentially enabling students to carry more weight with different postural adjustments. Thus, it may be inappropriate to extrapolate our findings to all school backpacks. Thus other styles of backpack used by adolescents should be tested in the same manner as this investigation for their effect on posture under different load arrangements (for instance, the effect of waist and chest straps, side compression straps, and compartments within the backpack to partition the load).