Custom spine staples were implanted into the midthoracic region of the spine (Fig. 1) in seven live, skeletally immature female domestic pigs weighing 240 to 334 N (24 to 34 kg) with an age of three to four months. In the female pig, puberty occurs at five to six months and adolescence is considered to be between five and eighteen months. The procedures were approved by an Institutional Animal Care and Use Committee and complied fully with the Animal Welfare Act. The animals were maintained in adjacent, individual cages for the eight-week postoperative period, and they were confined except during cage cleaning. Six staples per pig were implanted into the left side of adjacent vertebrae from T6-T7 to T11-T12 with use of thoracoscopic procedures. After the animals were killed, the thoracic spines were harvested, the screws and staples were carefully removed, and the spines were stored frozen. Radiographic results have been previously described1.
The purpose of the staples was to control displacement on one side of the intervertebral joint and thereby induce a compressive stress gradient across the joint. The staples were designed by two of the investigators (E.J.W. and D.I.B.-A.) and were fabricated of stainless steel (Techform Advanced Casting Technology, Portland, Oregon). A guidewire (a Kirschner wire with a 1.1-mm diameter) (MicroAire, Charlottesville, Virginia) was used to center the cannulated staples over the disc. Self-tapping stainless steel bone screws with a diameter of 2.7 mm and a length of 22 mm (Synthes, Monument, Colorado) were used to prevent the staples from backing out; two screws were used per staple.
The spines were thawed and were cut along the midcoronal plane with a band saw. Digital photographic images of the entire thoracic cross-section were acquired from a distance of 0.9 m. The spines were then fixed in 10% neutral buffered formalin (Richard-Allan Scientific, Kalamazoo, Michigan). After complete fixation, the spines were decalcified (Decalcifier II; Surgipath Medical, Richmond, Illinois) in more than twenty times the volume of the specimen. Two motion segments were resected and were split at the midsagittal plane. T4-T5 was chosen nominally as the unstapled control level, and T8-T9 was chosen as the treated level. Halved motion segments were embedded in paraffin and were sectioned parallel to the longitudinal axis of the spine at a thickness of 4 µm. Mounted sections were stained with hematoxylin and eosin (Fig. 2-A). Growth plate variables were measured at one physis per level. The cephalad or the caudad growth plate was chosen on the basis of which one exhibited better overall section quality. Digital images of the growth plates and a calibration scale were acquired with use of light transmission microscopy (nominal, ×100) and a digital camera. The sections were assembled into a composite microscopic image of each growth plate across the entire coronal plane. Growth plate variables were measured within four sampling locations located at 20%, 40%, 60%, and 80% of the distance from the staple to the contralateral cortex. At each of the four locations, variables were measured along a 1.5-mm linear distance, which defined the controlled sample space. The total vertebral width at the growth plate was approximately 26 mm, so that about one-quarter of the growth plate was sampled per level.
Qualitatively, differences between sampling locations were apparent, particularly between unstapled and stapled levels on the stapled side (Fig. 2-B). To quantify the differences, the height of the hypertrophic zone of the growth plate and the height and width of the cells within the zone were measured (Photoshop 7.0; Adobe Systems, San Jose, CA) (Fig. 3). The main variables of interest were the relative differences from side to side and from level to level within each spine. The hypertrophic zone was defined as the area in which chondrocytes were larger and lighter in comparison with those of the proliferative zone. Zone heights were measured along the longitudinal axis of the local growth plate boundaries in the direction of the cell columns. Only cells with clear boundaries were measured. The height of the hypertrophic zone was defined as the distance from the top edge of the most cranial cell to the bottom edge of the most cephalad cell. About fifty zone heights were measured for each sampling location, or 200 per level, at approximately equally spaced intervals.
Cell heights were aligned with the local zone height direction (Fig. 3). The width of each cell was defined as the largest left-to-right distance orthogonal to cell height. Only cells within the hypertrophic zone and with distinct boundaries were measured. Every eligible cell was included, regardless of size, with no attempt to discern bisected cells, for a total of approximately 200 cells per location, or 800 per vertebra.
Disc heights were measured from the digital images of the gross cross sections with use of commercial software (Photoshop 7.0). Every disc was measured at nine levels, from T4-T5 to T12-T13 inclusive, at three locations in the plane: 25%, 50%, and 75% of the distance from the staple to the contralateral cortex (Fig. 1, insets).
Statistical Methods
Mixed linear modeling for repeated measures34,35 was performed with use of three factors: animal (Specimens 1 through 5), thoracic level (stapled or unstapled), and coronal plane location (20%, 40%, 60%, 80%). The null hypotheses were that the outcome variables were not different between the stapled and unstapled thoracic levels and across the coronal plane locations. Separate tests were conducted for the dependent variables of zone height, cell height, cell width, and disc height. When significant differences were found for the interaction (p < 0.05) in the overall model, the Bonferroni correction was used. For growth plate variables, the significance level to test for differences between locations was 0.0125 (0.05/4), because only the differences between stapled and unstapled levels at corresponding locations were of interest, e.g., the difference between the two 20% locations. When left-to-right-side differences in disc height were compared, the significance level was 0.025 (0.05/2).
Next, regression models were separately fitted, again with use of a mixed modeling process, to stapled and unstapled levels in order to test whether each slope was different from zero. The dependent variables were zone height, cell height, and cell width, and the independent variable was coronal plane location. The intercept was the value of the dependent variable at the left edge of the vertebra, which would correspond to a coronal plane location of 0%. The slope was the rate of change of the dependent variable as a function of normalized distance from the left edge of the vertebra (coronal plane location, 0%) to the right edge (coronal plane location, 100%). The p values for slopes indicate the probability that the slope was equal to zero, i.e., the probability that the dependent variable was constant across the coronal plane. The significance level to test if slopes at the stapled and unstapled levels were different from zero was 0.025 (0.05/2).
Source of Funding
Funds from DePuy AcroMed (now DePuy Spine) were provided to conduct the surgical portion of the study at Ethicon Endo-Surgery, Inc., and were used for animal purchase and per diems, surgical supplies, and veterinary technical support. No funds were transferred to any author or to the authors' institution. Funds for student salary for one author (A.M.) were provided by the Women in Science and Engineering Program of the University of Cincinnati.
Five animals completed the eight-week protocol and were included in the analysis. Of the other two animals, one died two days postoperatively with hemorrhage of the costal vasculature and the other exhibited asymmetric movement of the hind limbs after surgery and failed to gain weight for six weeks. Gastric ulcers were found at the time of the autopsy. These two animals were eliminated from further analysis. The remaining animals gained an average of 142% of their weight as measured at the time of surgery. The thoracic levels used for growth plate measurements (see Appendix) varied from the protocol by as many as two levels in two animals because of variations in staple placement and the number of thoracic levels removed at the time of harvest. Hypertrophic zone height (Fig. 4), cell height (Fig. 5-A), and cell width (Fig. 5-B) were all lowest on the stapled side of the stapled level. All three growth plate variables showed significant differences in the overall model (p < 0.02). The disc heights (Fig. 6) of stapled vertebrae were significantly less than those at unstapled levels (p < 0.0001) across the entire coronal plane.
The mean zone height, across the four coronal plane locations, of the unstapled control level was 98.2 µm, whereas that of the stapled level was 77.1 µm. The mean zone height under the staple was 69% of the control zone height and 74% of the zone height on the contralateral side. At the stapled level, the zone height on the side contralateral to the staple was 92% of the mean control zone height (Fig. 4). Zone heights at corresponding coronal plane locations were significantly different from each other (Table I). Zone height was directly related to distance from the staple at the treated level (Table II), whereas zone height at the control level did not vary across the plane.
The average number of cells measured per animal was 1558 ± 201, with no significant difference between stapled and unstapled levels (p > 0.1). The mean cell heights of the control and stapled levels were 17.8 and 15.5 µm, respectively. The mean cell height, across the four coronal plane locations, under the staple was 85% of the control value and 93% of the value on the contralateral side, whereas the value on the contralateral side of the stapled level was 91% of the control value (Fig. 5-A). Cell height varied between stapled and control levels. Cell heights at corresponding coronal plane locations were significantly different from each other (Table I). The slopes at unstapled and stapled levels were not significantly different from zero (Table II).
The mean cell width, across the four coronal plane locations, was 17.9 µm at the control level and 16.4 at the stapled level. The mean cell width under the staple was 91% of the control value and 94% of the value on the contralateral side. The value on the contralateral side of the stapled level was 97% of the control value (Fig. 5-B). Cell widths at corresponding coronal plane locations were significantly different from each other, with the exception of the right side (80%) locations (Table I). The slopes at unstapled and stapled levels were not significantly different from zero (Table II).
Mean disc heights, across the four coronal plane locations, were lower at the stapled levels (2.1 mm) than at the control levels (3.2 mm) (p < 0.0001). The disc height under the staple was 58% of the control value, whereas the height on the contralateral side was 61% of the control value (Fig. 6). At unstapled levels, the side-to-side difference in the height of the disc anulus (0.42 mm) was significant (p = 0.008), with the height of the anulus on the contralateral, unstapled side being smaller than that on the ipsilateral side.
Spinal hemiepiphysiodesis with use of a staple-like implant induced structural changes in the hypertrophic zone of the growth plate and in the intervertebral disc. Growth plate hypertrophic zone height was related to the presence of the staple and the distance from the staple. On the stapled, concave side of the thoracic spine, hypertrophic zone height and cell height and width were significantly smaller36. Structural changes were consistent with increased compression of both the disc and the growth plates. These results support the concept of a compression gradient being at least a part of the mechanism of graduated growth inhibition, as is widely presumed to be the case for epiphyseal staples in the treatment of lower limb deformities. Growth plate structural changes were greatest under the staple and diminished across the coronal plane to nearly control levels. Cell width was decreased in addition to cell height, so chondrocytes did not expand laterally to maintain a constant volume or to grow perpendicular to the applied stress direction.
Limitations in the experimental design included a relatively small sample size, one evaluation time, grouped means at each of four sampling locations, a quadruped model, and the use of unstapled levels in treated animals as controls. Control physis measures may vary by level so that differences between the control level and the more caudad treated level may be due, in part, to a gradient by thoracic level. It is not known how much T4-T5 normally differs from T8-T9. The possibility of compensatory curves at the unstapled levels was a limitation of using unstapled levels as controls. Associated changes in vascular and other structures were not differentiated in the present study, and pressure effects have been hypothesized to act through changes to the blood supply, with similar histological effects37.
The effects of piercing the disc with a guidewire during the surgical procedures and the effects of vascular or other biological changes38,39 may not be isolated from the effects of compression in the present study, and no surgical sham was performed. Disc puncture in isolation has been used as a model of mild disc degeneration40. Biological and biomechanical effects due to disc compression41-54 and due to puncture of the anulus40,55,56 have been reported. In the current study, a guidewire was convenient but was not required, and one was not used in subsequent experiments. The controlled mechanical variable, intervertebral joint displacement on the stapled side, was not measured. Joint motion was assumed to be zero at the staple base given that the base was essentially infinitely stiff compared with the disc. Displacements of the intervertebral joint would be expected to be small within the blades and to increase with distance from the staple.
Technical limitations included relatively simple tissue processing and structural measurements. Variations in the number of ribs, the location of the staples, and section quality resulted in some differences in terms of which levels were used for growth plate measurements. The instrumentation was relatively easy to remove, and gross tissue sections did not contain evidence that the removal of staples and screws affected the morphology near any sampling location. The histomorphometry methods did not yield absolute zone and cell sizes. The process of fixing and decalcifying paraffin-embedded bone samples is well known to cause tissue shrinkage and morphologic changes28,57, and no attempt was made to measure only cells cut at midplane. The location of the sectioning plane relative to the midline of the staple varied both between spines and between levels. The sections cut closest to the midline of the staple exhibited the greatest differences in growth plate variables compared with controls. All sections were used in the analysis, not only those that were cut most closely to the staple midline. More consistent section plane locations may have yielded more significant differences across the coronal plane. Sectioning artifacts included fissures in the growth plate. Artifacts were judged to be relatively mild and were managed with the use of redundant slides and the ability to choose either the cephalad or caudad growth plate. Because the zone heights on the stapled side and level were smaller, the total area and volume of the stapled locations were decreased. Therefore, the length of each location was chosen as the controlled space to provide a spatial reference basis for sampling that was independent of treatment and location. Growth plate height measurements were performed twice, with the second study reproducing the patterns of the first with smaller variances.
Disc heights were measured unloaded and after cutting and so were not necessarily comparable with uncut or in vivo heights. Again, paired differences between level and side were the goal. A calibration photograph showed no measurable change in calibration factor along the spine section. Disc heights were measured twice by different observers, with no changes in any statistical conclusion.
The results of the present study may be compared with those of studies of tissues routinely processed after autopsy in humans58. Corresponding measurements from patients with scoliosis may help to quantify the potential for growth inhibition in humans. A preliminary report of hypertrophic zone and cell heights in humans58 indicated that they may be smaller than those of animal models. Other limitations in the ability to translate the results to possible human use include both global and local anatomical and biomechanical differences between species. However, comparisons between the biomechanical properties59 of quadruped and biped spines, and similarities across species in the dose-response curve of the growth plate to compression22, support at least in part the relevance of the study to human vertebrae. Normal physiological compressive stresses in the porcine spine are likely approximately half of those in the human spine60.
While tissue processing, definitions of variables and controls, and other experimental differences will affect quantitative results, zone and cell height reductions due to treatment may be compared with relative values from studies on other animal models17-20,22. In one study, staples were placed across the proximal tibial growth plate of four-week-old rats for fourteen days18. Under the staple, growth decreased to 33% of the control value, whereas on the side furthest from the staple, growth was 85% of the control value. In another study, growth rate differences in the spine that were produced by sustained mechanical loading of rat tail vertebrae were associated with reductions in the expected height increases of hypertrophic chondrocytes in the direction of growth20. After four weeks with an applied compressive load of 60% of body weight (1.2 N; nominal stress, 0.1 MPa), the average zone height and cell height in compressed tail vertebrae were 87% and 85%, respectively, of control values. With respect to disc height, another study of rat tails44 demonstrated that after eight weeks with applied compressive loads of up to 100% of body weight, or approximately 0.15 MPa above physiological stress, disc heights were reduced by nearly 50%. In a preliminary report of in vivo compressive stresses in the porcine disc, the static mean stress was increased 0.1 to 0.2 MPa following staple implantation61.
Results from a study of pigs would presumably be most directly applicable to humans if the growth rates were similar. However, human spinal growth rates are quite slow. A comparison of growth rates between the two species would be needed to interpret and translate the results. The increase in length of the thoracic spine (T2 to T12) in domestic pigs from the initial age to the final age in the present study was approximately 7 cm, which corresponds to an average growth rate of 50 µm per physis per day. The average remaining thoracic spine growth is approximately 7 cm in girls with a skeletal age of six years62, and the average growth rate in children and adolescents is on the order of 1 µm per physis per day22,63,64. Therefore, while the growth rate of the pig is much faster than that of the human, the duration of growth is much shorter, so the total growth of the spine is similar. In clinical use, the assumption is that if treatment occurs sufficiently early and is in place long enough, the treatment effect is likely to be similar.
In conclusion, a method of spinal hemiepiphysiodesis that was previously reported to cause curvatures in an inverse analog treatment model altered vertebral growth plate structure in a porcine animal model consistent with transmission of compressive stress to the physes. The results of the present study support the concepts that scoliosis progression involves a compressive stress gradient and asymmetrically decreasing vertebral growth in a positive feedback control loop65,66 and that an early developing scoliotic curve may be slowed by differential inhibition of the growth plate.
Note: Financial support was provided by DePuy AcroMed (now DePuy Spine), for the surgical portion of the study, and by the Cincinnati Children's Hospital Research Foundation. Laboratory support was provided by Ethicon Endo-Surgery, Inc.; special thanks to Ronald J. Kolata, DVM, and engineer Anthony Sexton. The design and fabrication of the collet-based endoscopic screwdriver is attributed to Arthur Case and Daniel Radigan (April 2001) of 7 Letters, Cincinnati, Ohio. Technical assistance by David Sheyn, Jonathan Henkel, Alison Grimaldi, and Keith Stringer, MD, and statistical consultation by Judy Bean, PhD, are gratefully acknowledged.