The halo orthosis has been widely recognized as the best available device for external immobilization of the cervical spine6,16. It has great utility for the treatment of post-traumatic, neoplastic, congenital, and postoperative conditions that require secure fixation4,5,10,12,13,20,21,23,25. The most important clinical complications related to its use have been associated with loosening of the pins8-10. Other complications have included infection at the pin sites, brain abscesses due to perforation of the skull by the pins, failure of fixation of the ring to the skull, failure of immobilization of the cervical spine, and dislocation of the injured portion of the cervical spine3,8,9,11,19.
It is widely believed that loosening of the pins is directly related to the insertional torque used at the time of the application of the halo device17,19. Recent research has suggested that the use of 0.90 newton-meter (eight inch-pounds) of torque rather than the traditional 0.68 newton-meter (six inch-pounds) can markedly reduce the prevalence of these complications9,11,14,15,20,22,24. Those studies suggested that increased torque provides more appropriate loading at the pin-skull interface, diminishes motion between the bone and the pin and between the pin and the skin, and subsequently decreases the prevalence of local complications at the pin site.
The goal of appropriate application of torque for insertion of the pin is to achieve optimum loading at the pin-bone interface. The amount of load that the pin applies to the surface of the skull depends on several factors14, including the angle of the pin to the surface of the skull, the thickness and construction of the material of the halo ring, the width and pitch of the thread of the pin, the interaction between the material of the pin and that of the halo ring, the use of lubrication of the thread during insertion of the pin, and torque1,2,7,18,20,24,26. In usual practice, the only variables at the time of the operation are torque and the use of lubrication.
Most manufacturers of halo orthoses supply their own devices for measuring or limiting torque during the insertion of pins. Some manufacturers provide the clinician with several choices. Three classes of devices are available: fully adjustable, spring-loaded screwdrivers equipped with a dial that indicates torque; destructible, single-use, torque-limited, shear-off devices; and fixed or partially adjustable screwdrivers with a spring-loaded ratchet that provides an audible or palpable signal once the desired torque has been achieved.
Despite the abundance and diversity of torque-setting devices, the accuracy (the ability to apply the intended torque) and reliability (the ability to achieve the same torque on numerous occasions) of these devices have not been well documented. Information regarding accuracy and reliability is needed to aid the surgeon in choosing the best device for the insertion of pins. This information also is relevant to the interpretation of previous research concerning the ideal torque needed to avoid complications during the insertion of pins. The purposes of the present study were to investigate the accuracy and reliability of commonly used commercially distributed devices and to determine if any particular devices are superior in their ability to apply the intended torque.
*With the exception of the donation of the insertional devices by several manufacturers, no benefits in any form have been received or will be received from a commercial party related directly or indirectly to the subject of this article. Funds were received in total or partial support of the research or clinical study presented in this article. The funding source was the Minnesota Spine Center Research Foundation.
†Minnesota Spine Center, 606 24th Avenue South, Suite 602, Minneapolis, Minnesota 55454.
‡College of Physicians and Surgeons, Columbia University, 161 Fort Washington Avenue, New York, N.Y. 10032.
Five major United States manufacturers of halo devices (Ace Medical, Los Angeles, California; Bremer Medical, Jacksonville, Florida; Jerome Medical, Mount Laurel, New Jersey; Levtech, Dallas, Texas; and PMT, Chanhassen, Minnesota) voluntarily provided torque screwdrivers or shear-off devices for use in the present study (Figs. 1, 2, 3 through 4). Each manufacturer provided three units of the most popular reusable devices that orthopaedic surgeons would be likely to use in practice; multiple units were needed in order to assess the variability among devices. Three of the manufacturers also supplied a minimum of thirty-five units of disposable, single-use, self-destroying devices for use in the study. We tested every type of device when a manufacturer offered several choices. In addition to the new devices supplied by the manufacturers, we tested two other torque-setting devices: a spring-loaded screwdriver with a maximum adjustable torque of 0.68 newton-meter that had received long-term repeated use in our spinal unit and a basic screwdriver without an indicator that was used with a manual two-fingered technique. The manual technique was tested to determine the variability of torque without the aid of an indicator because the first anecdotal descriptions of the use of torque for the insertion of halo pins were in studies in which a manual two-fingered method had been used1,5. Over-all, twelve different devices were tested. Eight devices were tested at either 0.68 or 0.90 newton-meter (six or eight inch-pounds) of torque, and the other four were tested at both levels (Table I).
One of us (M. D. S.) tested all of the disposable shear-off devices in one session. The rest of us (L. J. J., J. H. P., and B. A. R.) tested the reusable devices according to the manufacturer's instructions; this allowed us to measure interobserver reliability. All of us are experienced spinal surgeons and are familiar with the application and subsequent management of halo orthoses.
The amount of torque achieved with each device during each trial was measured with a commercially available digital torque-meter (Compudriver; Consolidated Devices, City of Industry, California). The manufacturer verified that this device had a calibrated accuracy of ±1 per cent of the observed value and a reproducibility of ±0.5 per cent of the observed value. The torque-meter was mounted on a stand and fit with adapters to accommodate the different types of halo pins. The pins were rigidly mounted on the torque-meter and could not rotate; as a result, there may have been some loss of tactile feedback. However, because of the variety of pins and drivers, we thought that a rigid mount would provide the least variability among devices.
Accuracy was assessed by calculating the percentage and absolute amount by which the achieved torque deviated from the target torque in each trial of each device. We defined an acceptable percentage deviation from the target torque as an observed value that was within ±10 per cent of the manufacturer's stated goal for the device. In other words, when the intended torque was 0.68 newton-meter (six inch-pounds), the device was considered to be accurate if the measured value was within the range of 0.61 to 0.75 newton-meter (5.4 to 6.6 inch-pounds), and when the intended torque was 0.90 newton-meter (eight inch-pounds), the device was considered to be accurate if the measured value was within the range of 0.81 to 0.99 newton-meter (7.2 to 8.8 inch-pounds).
Achieved Torque
Analysis of variance of achieved torque revealed a significant degree of variability as a function of device (low measurement reliability), tester (low interobserver reliability), equivalent device (low intraobserver reliability), and the combined effect of device, tester, and equivalent device.
One-way analysis of variance was used to test the differences in mean achieved torque as a function of device for the three shear-off devices (A, B, and C; Fig. 1), which were tested at 0.90 newton-meter (eight inch-pounds). The main effect of device was significant (p < 0.001). The mean achieved torque was 0.92 newton-meter (8.1 inch-pounds) for device A, 0.79 newton-meter (7.0 inch-pounds) for device B, and 0.96 newton-meter (8.5 inch-pounds) for device C (Table II). All possible pairwise comparisons of these mean achieved torques were significantly different (p < 0.05, Student-Newman-Keuls post hoc comparison technique).
Three-way analysis of variance was used to test the differences in mean achieved torque as a function of device, tester, and equivalent device for devices E, G, H, and I (Figs. 2, 3 and 4) (Table II). When these four devices were tested at 0.68 newton-meter, the main effects of device and equivalent device were significant (p < 0.001 for both), but the main effect of tester was not (p = 0.9). All two-way interactions and the three-way interaction also were significant (p < 0.001). When devices E and I were tested at 0.90 newton-meter, the main effects of device, tester, and equivalent device were significant (p < 0.001 for each). All two-way interactions and the three-way interaction also were significant (p < 0.001).
Two-way analysis of variance was used to test the differences in mean achieved torque as a function of device and tester for the fully adjustable devices D and F (Fig. 2), which were tested at both 0.68 and 0.90 newton-meter, and the partially adjustable device J-new (Fig. 4), which was tested at 0.68 newton-meter (Table II). When devices D, F, and J-new were tested at 0.68 newton-meter, the main effects of device (p < 0.001) and tester (p < 0.01) were significant and the two-way interaction also was significant (p < 0.05). When devices D and F were tested at 0.90 newton-meter, the main effects of device and tester were significant and the two-way interaction also was significant (p < 0.001 for all comparisons).
The mean achieved torque (and standard deviation) across all trials in which the target torque was 0.68 newton-meter was 0.68 ± 0.10 newton-meter (6.0 ± 0.92 inch-pounds), compared with 0.82 ± 0.12 newton-meter (7.3 ± 1.06 inch-pounds) across all trials in which the target torque was 0.90 newton-meter. Substantial overlap between these two over-all distributions of achieved torque was observed: the torque that was achieved in approximately 25 per cent of the trials in which the target torque was 0.90 newton-meter was in the same range as that achieved in 75 per cent of the trials in which the target torque was 0.68 newton-meter (Fig. 5).
These results indicate that the direction of the combined effects of device, tester, and equivalent device were inconsistent, and that the effects of device, tester, and equivalent device were not additive, making generalizations about reasons for the significant variability in achieved torque impossible.
Accuracy
The proportion of trials in which the achieved torque met our a priori standard for accuracy (within ±10 per cent of the intended torque) ranged from less than 1 to 100 per cent for the sixteen combinations of devices and settings tested. Only six testing conditions resulted in 80 per cent accuracy or better: devices A, D, E, and F tested at 0.90 newton-meter of torque, and devices E and J-new tested at 0.68 newton-meter of torque (Table II).
We found that, for a particular device, the proportion of trials in which the achieved torque was within ±10 per cent of the intended torque could be low even when the mean achieved torque across trials was no different from the intended torque (Table II). For example, the mean achieved torque for device D was 0.71 newton-meter (6.27 inch-pounds) when tested at 0.68 newton-meter and 0.91 newton-meter (8.05 inch-pounds) when tested at 0.90 newton-meter. Both of these mean values were within ±10 per cent of the intended torque and might have led to the incorrect impression that the device was equally accurate when tested at both 0.68 and 0.90 newton-meter. However, device D achieved only 68 per cent accuracy when tested at 0.68 newton-meter, compared with 88 per cent accuracy when tested at 0.90 newton-meter. Comparison of the data for device J-new and device J-used also demonstrates that the proportion of trials in which the desired range of accuracy is achieved is a better indicator of the accuracy of a device than the mean torque measured across trials. These were identical partially adjustable devices, but device J-new had been supplied by the manufacturer for the purpose of the present study while device J-used had been used extensively in our spine service over a period of eight years. Both devices were tested at 0.68 newton-meter (six inch-pounds). The mean achieved torques were similar (0.65 and 0.63 newton-meter [5.78 and 5.57 inch-pounds], respectively), but device J-new achieved acceptable accuracy in 100 per cent of the trials while device J-used was accurate in only 58 per cent of the trials.
The mean percentage deviation from the intended torque also failed to distinguish between devices with high and low levels of accuracy (Table II). The amount of deviation from the intended torque varied extensively; the mean percentage deviation ranged from -23 to +17 per cent. Standard deviations of percentage deviation were large, often greater than the mean values, and the ranges of percentage deviation also were large, ranging from 10 to 118 per cent. Nine testing conditions had a range of more than 30 per cent. If the mean percentage deviation from the target torque is taken as an indicator of accuracy when comparing two devices, then a device with the lower mean percentage deviation from the target torque would be considered to be more accurate. When this approach is used, the manual two-fingered technique (mean percentage deviation, 2 per cent) would have been considered more accurate than device F (mean percentage deviation, -8 per cent), despite the fact that device F was accurate in 86 per cent of the trials while the manual technique was accurate in only 35 per cent of the trials (Fig. 6).
Comments on Specific Devices
Only devices A, E, and J-new consistently achieved values that were within ±10 per cent of the target torque across different testers and devices.
For device A, the proportion of trials in which the achieved torque was within the desired range of accuracy was 100 per cent, the mean percentage deviation from the target torque was 2 per cent, and the mean achieved torque (and standard deviation) was 0.92 ± 0.02 newton-meter (8.1 ± 0.22 inch-pounds).
For device E, which was tested at both 0.68 and 0.90 newton-meter, the proportion of trials in which the achieved torque was within the desired range of accuracy was 99 per cent. When the device was tested at 0.68 newton-meter, the mean percentage deviation was -2 per cent and the mean achieved torque (and standard deviation) was 0.67 ± 0.03 newton-meter (5.9 ± 0.24 inch-pounds). When it was tested at 0.90 newton-meter (eight inch-pounds), the values were -3 per cent and 0.88 ± 0.03 newton-meter (7.75 ± 0.24 inch-pounds), respectively.
For device J-new, the proportion of trials in which the achieved torque was within the acceptable range of accuracy was 100 per cent, the mean percentage deviation from the target torque was -4 per cent, and the mean achieved torque (and standard deviation) was 0.65 ± 0.02 newton-meter (5.78 ± 0.17 inch-pounds).
A comparison between device J-new and device J-used showed that the used device was significantly less accurate and that its performance was more variable. The percentage of trials in which the used device achieved a torque that was within ±10 per cent of the intended torque was significantly different than that for the new device (58 per cent compared with 100 per cent) (chi square = 65.69, df = 1, p = 0.0001). The standard deviation of the difference between the measured and target torques was significantly higher for the used device than for the new device (0.67 compared with 0.31) (p < 0.001, Levinets test for equality of variance). For the used device, the proportion of trials in which the achieved torque was within the desired range of accuracy was 54 per cent for one tester, 100 per cent for the second tester, and 21 per cent for the third tester. Interestingly, all achieved torques that were not within the desired range were lower than 0.61 newton-meter (5.4 inch-pounds).
As expected, we found that the effect of tester was highly significant for the manual two-fingered technique. The mean achieved torque was significantly different across testers, with the first and second testers achieving significantly higher mean torque than the third tester (0.94, 1.1, and 0.72 newton-meter, respectively) (p < 0.001). The proportion of trials in which the achieved torque was within ±10 per cent of the target torque was 60 per cent for one tester, 31 per cent for the second tester, and 15 per cent for the third tester, with an over-all proportion of 35 per cent. All of the achieved torques that were not within the desired range were higher than 0.99 newton-meter (8.8 inch-pounds) for the first and second testers, while all were lower than 0.81 newton-meter (7.2 inch-pounds) for the third tester.
We believe that the proportion of trials in which the achieved torque is within a particular range (±10 per cent of the target torque in the present study) is a superior indicator of the level of accuracy of a particular device compared with the mean achieved torque or the mean percentage deviation from the target torque. This is because the mean is biased by the presence of extreme values. The mean achieved torque may be the same as the intended torque for a particular device even if that device does not consistently perform within an acceptable range of accuracy. Two devices could, therefore, be equal in terms of mean torque and yet unequal in terms of clinical performance.
The goals of the present study were to determine the accuracy and reliability of devices used to insert halo pins and to determine if any particular devices were superior in their ability to apply the intended torque. Our data suggest that the insertion of halo pins with most of the currently available torque-setting devices is an inherently inaccurate process. The use of equivalent devices from the same manufacturer did not ensure equivalent insertional torque with similar degrees of error (poor intraobserver reliability). Similarly, the same device used by different surgeons did not reliably apply the same torque (poor interobserver reliability). Our data showed that most types of devices, made by different manufacturers, did not achieve reasonably consistent levels of accuracy or reliability. No particular class of device, such as spring-loaded, shear-off, or fully adjustable devices, consistently achieved acceptable levels of accuracy or reliability. However, we did identify a specific device in each category (devices A, E, and J-new) that performed quite well individually.
We were unable to identify why specific devices yielded unreliable results. One cause of error for the spring-loaded and fully adjustable devices was inaccurate placement of the calibration marks that guide the application of torque (Figs. 7-A and 7-B). Another problem detected during testing was sticking of the metal or plastic surfaces of the non-adjustable spring-loaded devices. The extensive intertester differences that were observed suggest that the specific technique used by a particular surgeon may have a role in the torque produced by a device. For example, surgeons may differ in how they hold the device or in the speed at which they turn it. Our experimental setup required that the pins be rigidly mounted on the torque-meter. Therefore, the possible utilization of axial resistance for tactile feedback normally available to surgeons was lost. This loss was considered to be an acceptable limitation of the experimental setup, but it may have influenced accuracy, especially in the testing of the manual two-fingered technique.
While our clinical experience suggests that increased torque may lead to a decreased rate of complications, our data imply that an optimum torque probably is unknown because independent verification of the actual torque achieved with insertional devices has not been reported in studies of optimum torque11,20,24. Meaningful interpretations of the relationship between insertional torque and complications related to the pin are possible only if the torque achieved at the time of insertion is documented.
Much of the literature regarding complications related to halo pins has focused on insertional torque as the variable to be controlled4,5,8,10. Recently, Whitesides et al. clearly showed that the halo pin load is affected not only by the torque used to insert the pin but also by the design and material composition of the halo ring, the design of the tip of the pin, the design of the thread, the friction coefficient of the materials used, and whether a lubricant is used. The optimum load is unknown. Additional knowledge of the osteology of the skull, especially as it pertains to age and bone density, should assist in the determination of optimum load. We think that such a determination for a particular patient would decrease the complications of loosening, loss of fixation, infections at the pin site, and perforation of the skull with its associated risks of brain injury and abscesses. The optimum load probably would depend on the skull to which the load is applied.
Our results suggest that more standardized procedures for the insertion of halo pins should be developed. An accurate and reliable industry-wide standardized insertional device and a technique of insertion of the pins conceivably could lead to less variability in insertional torque. In addition, standard procedures for lubrication of the pin shafts, based on a better understanding of the frictional properties of the pin-ring interface and, perhaps most importantly, accurate insertional devices, may lead to a better understanding of optimum use of torque during the insertion of halo pins.