This was a three-part radiographic study. The first part was used to identify the problem with unstandardized posture for radiographs, the second was to identify variations between multiple radiographs, and the third part was to study the effects with varied malpositioning. We utilized C57BL/6J mice for this study bought from the University of Hong Kong animal laboratory. For sample size of animal tests, we aim to utilize at least 12 mice per experiment, following a protocol on a previous interventional mouse model.(18) Based on the numbers of mice in the three experiments as seen below, our sample size was adequate for such study. As this is a radiographic study of mice positioning, all mice were included in the study. There were no exclusion or randomization as all mice were tested. All mice were anesthetized by intraperitoneal injection of 5 mg/ml phenobarbital sodium (Dorminal 20%, Alfamedic) dissolved in saline at the dose of 50 mg/kg. After losing reflection to firm pinch on the paw, they were immediately transferred to warm pad for imaging. All radiographs were completed within an hour to guarantee the mice in the best stationary state. After the experiments, they were put back to the warm pad until sober. Weight and feeding behaviors were monitored for several subsequent days. All anesthetic and radiographic studies were performed by the same investigator. All the animal experiments were approved by the Committee on the Use of Live Animals in Teaching and Research (CULATR) (Ref# 3720-15). Animals were not sacrificed at the end of study and was passed to the university laboratory for other studies. No animals experienced adverse events.
Cobb angle measurements
All radiographic Cobb angle measurements were performed in AUTOCAD (version 0.48.M.570; © 2017 Autodesk, Inc.). Tiff image files were imported by Raster image reference. The measurement method used was the same as in humans with an angle formed by the superior endplate of the most tilted cranial vertebra from the apex and the inferior endplate of the most tilted caudal vertebra from the apex (Fig. 1). For the mouse, the upper and lower endplates of each vertebra can also be seen clearly for the measurement. Measurements were carried out with the reader blinded to the mouse number and allocation, and each image was measured twice with an interval of one month. All measurements were within 5° and the final reading for analysis was the average of the two measurements.
Preliminary experiment
For the first stage of the radiographic study, we collected 77 wild-type C57BL/6J mice with no documented genetic susceptibility to spinal deformities to perform radiographs. There were 39 females and 38 males with age (mean ± standard deviation/SD) of 166 ± 68 days. This was a general litter available to our faculty. A digital diagnostic X-ray machine (KeenRay Top-U) was used to obtain plain radiographs at 40 kilovoltage (kv) and 40 milliamperes (mAs). After anesthesia, mice were kept in the prone position until they were transferred to the examining table right beneath the radiation source. During transfer, we used the index finger and the thumb to clamp the scruff in addition to pinching the back fur between the middle finger/fourth finger and the thenar eminence (Fig. 2). This formed three holding points to allow horizontal movement while maintaining the natural status of the spine. Because the mouse was biologically left-right symmetrical, after it was put on the table, we adjusted the head and limbs by gentle movements to make sure that the snout, interorbital space, cervical vertebrae, spine and caudal vertebrae were kept in a line, and the limbs were placed on the board naturally while symmetrical to the trunk. This posture was defined as the “correct posture” (Fig. 3). No outer force was applied to the spine. A position and reposition test was done to confirm repeatability of the positioning maneuver before proceeding with the x-rays.
Direction of curves was not taken into account for this part of the study and so all angles were considered positive. Descriptive Statistics for mean, SD and standard error (SE) of mean calculations were used to describe the data. All statistical analyses were performed using SPSS IBM Statistics 20 (IBM, Armonk, New York, USA). A p value of < 0.05 was considered statistically significant.
Weekly X-ray Experiments
For the second study, we explored the variation of curve magnitude measurements with multiple radiographs. We collected 24 wild-type C57BL/6J mice with no documented genetic susceptibility to spinal deformities for 4 consecutive weekly X-rays. There were 5 females and 19 males. The first radiograph was taken at the age of 4 weeks. We followed the standard rule and used the correct posture described above for radiographs. Relative standard deviation (RSD) was introduced for normalization, and SDs for RSD of the 24 subjects were calculated and used to describe the overall variation in multiple measurements. All calculations were reported by the descriptive statistics.
Position Study
For this third part, we designed 14 different postures to test the impact of each malpositioning on the spinal curvature. These postures were based on the three main movable parts of mouse namely the limbs, head and tail. The postures are as follows:
1) 4 postures related to the curling of limbs (Curled left forelimb, Curled right forelimb, Curled left hindlimb and Curled right hindlimb).
2) 4 postures related to head tilt (Head tilt left 30°, Head tilt left 60°, Head tilt right 30° and Head tilt right 60°). The angle was formed by the longitudinal axis of the trunk and the axis of the head.
3) 2 postures related to pelvic tilt (Pelvis tilt left 30° and Pelvis tilt right 30°). The angle was formed by the longitudinal axis of the trunk and the top of the pelvis.
4) 2 postures related to tail tilting (Tail tilt left 30° and Tail tilt right 30°).
5) 2 postures related to head rotation (Head rotates clockwise and Head rotates anti-clockwise) in the direction of the observer.
For this part of the study, 82 wild-type C57BL/6J mice with no documented genetic susceptibility to spinal deformities were collected. There were 37 females and 45 males. The mean age was 54.5 ± 1.7 days. Each mouse was applied to each of the 14 postures as well as the “correct posture”. In order to quantify the tilt angles, a round plastic pad was made which was 20 cm in diameter and placed over the imaging platform. It was labeled every 30° with branch-like auxiliary lines for accurate malpositioning of the mouse (Fig. 3). The ULTRAFOCUS (Faxitron Bioptics LLC) was used to obtain plain radiographs at the parameter of 25 kv, 3.00 s and 0.4 mA. After anesthesia, mice were kept in the prone position until they were transferred to the tray with the method described before.
A standardized order of the radiographs (Fig. 4) was utilized for each mouse. The subject was first positioned right in the middle of the plastic pad for the “correct posture” image. After each of the subsequent malpositioning images, the mouse was repositioned into the “correct posture” for further malpositioning. The left forelimb was first pushed towards the body to the limit to yield the “Curled left forelimb” image. Then we performed the similar push maneuver for the right forelimb towards the body to the limit for the “Curled right forelimb” image. The left hindlimb was then pushed towards the root of the tail to the limit for the “Curled left hindlimb” image. Following this, the right hindlimb was then pushed towards the root of the tail to the limit for the “Curled right hindlimb” image.
The mouse was moved by three-finger pinch to adjust its location to be fit for quantification of head tilt. The skull was gently pinched and turned to 30° to the left as measured on the pad scale to produce the “Head tilt left 30°” image. Then the skull was further turned to 60° for the “Head tilt left 60°” image. The skull was then turned 30° to the right for the “Head tilt right 30°” image. The skull was further turned to 60° to the right for the “Head tilt right 60°” image.
For the next step, the mouse was moved by three-finger pinch to adjust its location to be fit for quantification of pelvis tilt. The pelvis was turned 30° to the left by simultaneously rotating the hindlimbs clockwise in terms of the pad scale. This yielded the “Pelvis tilts left 30°” image. The next image entailed turning the pelvis 30° to the right by simultaneously rotating the hindlimbs anti-clockwise. This position yielded the “Pelvis tilt right 30°”.
After moving the mouse by three-finger pinch to adjust its location to be fit for quantification of tail tilting, we gently pushed the tail 30° to the left for the “Tail tilt left 30°” image. The tail was also pushed 30° to the right for the “Tail tilt right 30°” image.
Finally, for the head rotation, we placed the mouse in the middle of the pad again. The skull was pinched and rotated clockwise to the limit for the “Head rotates clockwise” image. The skull was similarly rotated anti-clockwise to the limit for the “Head rotates anti-clockwise”.
Statistical analysis
The posture change in mouse would result in the change of curve direction. In this part, we regarded the 82 subjects as 82 dependent variables and assigned the curve with apex towards left as negative, and the curve with apex towards right as positive. There were usually three kinds of tests used for the situation that equal variances were not assumed(19). Dunnett’s T3/Dunnett’s C, Games and Howell and Tamhane’s T2 have the same t ratio and the same df values, but they differ in the P value calculation. T2 is more conservative than T3 for large sample sizes while C is more conservative for smaller samples (< 50 per group). The Games-Howell test is suitable for larger sample sizes and is more powerful than C, T2 and T3. Here, One-way ANOVA with Games-Howell tests was used for multiple comparisons with significance level of 0.05.
In order to compare the effect of postures, Glass’s delta was introduced. It is an alternative to Cohen’s d when equal SD is not assumed. It only uses the SD of the control group. Cohen’s d is calculated by the formula: Cohen's d = |M2 - M1| ⁄ SDpooled where M1 represents the mean of control and M2 represents the mean of experimental group. SDpooled is calculated by the formula: SDpooled = √ ((SD12 + SD22) ⁄ 2). SD1 and SD2 are the SD of control group and experimental group respectively. A small effect is below 0.2, a small to medium effect is between 0.2–0.45, a medium effect is between 0.45–0.65, a medium to large effect is between 0.65–0.8 and > 0.8 is large effect.