Surgical performance when inserting non-locking screws: a systematic review

Screw tightness achieved as a percentage of the maximum

Several experimental studies have investigated the torque achieved by a surgeon (stopping torque) and compared this to the maximum possible torque. The maximum torque being determined at a separate time interval by using a torque meter to further tighten the screw until the maximum torque value is reached when the material is stripped (stripping torque). By defining the maximum tightness as 100% (stopping torque = stripping torque), the ratio of stopping torque to stripping torque enables presentation of the tightness for that insertion as a percentage of the maximum. Many variables, such as the type of screw used and the material they were inserted into, have been assessed, generating a range of different screw tightnesses achieved by surgeons (Fig. 2). The first major work on this topic was published by Cordey et al in 1980. ⁶ Sixty-three orthopaedic and general surgeons manually tapped and inserted one 4.5 mm cortical, stainless steel screw unicortically into one human cadaveric femur, aiming to apply ‘optimal torque for a good fixation’. This procedure was repeated with 35 surgeons inserting the same screws into one human cadaveric tibia. Screws were tightened to 84±13% (mean±standard deviation) and 88±18%, respectively. The authors found that 10 out of 108 screws (9%) were inserted beyond the stripping torque; it was not recorded whether this was detected by the surgeons. In the second part of their study, 36 surgeons were asked to insert three screws into human cadaveric tibiae using three different methods. First, they assessed the effect of different drilling techniques by using either a large air drill for making pilot holes whilst having radiographs available and being able to see the bone, and second in separate holes repeating the first method but with a small air drill instead. Finally, they asked for screws to be inserted with neither radiographs available nor sight of the bone; though methods for blinding surgeons were not stated. None of these experimental setups generated significant differences in screw tightness.

View Full Size

Tightness achieved for each part within each study, where measured. From top to bottom, grouped alphabetically within the following sections: cortical screws in artificial bone (dark blue), cortical screws in human bone (dark orange), cancellous screws in artificial bone (light blue), cancellous screws in human bone (light orange). All bubbles scaled with size representing number of screws used, e.g. Acker et al 2016 – first-year = 40 screws.

The different components of each study, where relevant, are explained as follows: Acker et al and Wilkofsky et al – different years of experience of surgeons; Aziz et al: A – cortical screws in fresh frozen human bone, B – cortical screws in embalmed human bone, C – cortical screws in dried human bone, D – cortical screws in normal density artificial bone, E – cortical screws in osteoporotic density artificial bone, F – cancellous screws in fresh frozen human bone, G – cancellous screws in embalmed human bone, H – cancellous screws in dried human bone; Tsuji et al artificial bone: densities for each part (cortical and cancellous screws respectively) – 0.08 g/ cm³ (A and I), 0.16 g/cm³ (B and J), 0.24 g/cm³ (C and K), 0.32 g/ cm³ (D and L), 0.40 g/ cm³ (E and M), 0.48 g/ cm³ (F and N), 0.64 g/cm³ (G and O), 0.80 g/cm³ (H, cortical only); Tsuji et al, human bone: P – cortical screws, Q – cancellous screws; Cordey et al, 1a – 4.5 mm cortical screws in human femur, 1b – 4.5 mm cortical screws in human tibia; Mears et al, A – 90° past contact of the screw head on the plate; B – 180° past contact of the screw head on the plate; C – two-fingers tight; D – 1.4 Nm; Stoesz et al, high density (0.32 g/cm3), medium density (0.16 g/cm³), low density (0.08 g/cm³).

^#Ratio estimated based on provided data, though not explicitly stated by authors.

Citation: EFORT Open Reviews 5, 1; 10.1302/2058-5241.5.180066

• Download Figure
• Download figure as PowerPoint slide

In 1995 McGuire et al asked 105 orthopaedic surgeons of various experience to insert three titanium and three stainless steel 3.5 mm screws into non-locking plates. ⁷ The instructions to the surgeons were to insert the screws to what they considered ‘two fingers tight using their normal technique’. This instruction being a subjective insertion method thought to reduce applicable torque, as a reduced grip is used due to only two fingers holding the screwdriver handle. The stopping torque was measured; however, no assessment of the stripping torque was performed. They found a significant trend for higher stopping torques with more years of surgical experience, a variable that will be explored later in this review. When inserting stainless steel screws, more torque was applied compared to titanium screws. Whilst the number of surgeons employed, and the number of screws inserted (n = 315 per screw type, three for each of 105 surgeons) is the largest of any study to date, this work was limited by the lack of stripping torque assessment, both whether any of the screws were stripped on insertion and whether or not this was detected.

Dinah et al had one surgeon inserting 200 screws (160 bicortical, 3.5 mm cortical screws and 40 unicortical, 4.0 mm cancellous screws) into human cadaveric fibulae. ¹⁴ They found that, on average, screws were inserted to 71% of the stripping torque. Analysis of their provided data actually shows this value to be 66% given that 83% of the inserted screws were cortical; the stopping/stripping torque ratio was 64% for cortical screws and 77% for cancellous screws. The range reported was 18% to > 100%, with values > 100% being calculated as the stopping torque was greater than the stripping torque that could be generated subsequently, as the material had already been stripped during the initial tightening episode.

Tsuji et al investigated the effect of bone density on tightness, in both human and artificial bone. ¹¹ They measured average tightness for 24 insertions of 3.5 mm cortical screws in artificial bone of eight different densities, 12 insertions of 6.5 mm cancellous screws in Sawbones blocks of seven different densities, three insertions for 3.5 mm cortical screws into each of 16 human cadaveric femurs and two insertions of 6.5 mm cancellous screws into each of 16 cadaveric femoral condyles. Combining all densities, the tightnesses for cortical and cancellous screws in artificial bone were 81% and 77% respectively, and in human cadaveric bone, 67% and 85% respectively. They did not report on the percentage of screws that stripped on insertion either quantitatively or subjectively. However, as average ratios were shown, at times averaging 24 tests, and that some tightness averages were 100%, it is likely that some screws stripped the samples on insertion.

Stoesz et al asked five senior resident and five senior practicing surgeons to each insert eight 4.0 mm cancellous screws into artificial bone of three densities (0.08, 0.16 and 0.32 g/cm³). ¹⁰ Of the 239 screws reported, 131 were successfully inserted without stripping the bone; these had a tightness of 82±16%. The remaining screws (108 out of 239) stripped the polyurethane models. They found only a weak correlation (R ² = 0.54) between surgeons who were able to insert screws close to, but below, the maximum and those who infrequently stripped screws. They also found that as surgeons inserted more screws in each density, stripping rates increased (p = 0.022); however, in another article employing similar methods with eight screws inserted, this effect was not seen. ¹²

Acker et al asked 33 trainees and eight senior surgeons to insert six screws into bone models with a density of 0.48 g/cm³, with the instructions to insert to ‘two-finger tightness’ with their dominant hand. ⁹ This was repeated with their non-dominant hand. Dominant hand data showed no significant difference between screw tightness when combining all surgical trainees (74%) and comparing this with faculty (68%); non-dominant hand data were not reported. The variability between participants grouped by years of experience, however, was large with first-year trainees’ average tightness being 58% and fifth-years’ 103%, i.e. the average for this latter group being beyond the stripping limit of the artificial bone. Additionally, there were large variations in achieved tightness within each group. This is the only study to have investigated the effect of hand dominance, finding a 70% difference in tightness between hands for first-year surgeons and 9% for senior surgeons.

Reitman et al asked one surgeon to insert screws into the anterior aspects of cadaveric human vertebral bodies. ¹³ Initially, one screw was inserted until stripped to establish the maximum torque, followed by a second screw into the same vertebral body to measure the peak perceived torque; this was performed 48 times. They found a tightness of 84%, with only one of these latter screws stripping the bone on insertion (2%).

Two studies used real-time torque feedback via visual displays. ^12,15 Mears et al inserted 10 screws into osteoporotic human humeri at 90° or 180° rotation past the point of first screw head contact, to 1.4 Nm or to two-finger tightness. ¹⁵ Whilst not explicitly defined, 1.4 Nm was likely chosen as this is 70% of the maximum value for these osteoporotic bones and matches the value found for the optimum tightness in the study by Tankard et al. ¹⁸ The torque values that screws were inserted to were recorded, although without direct assessment of the stripping torque. However, based on the assumption that 70% tightness was achieved with the 1.4 Nm tests, tightnesses of 64%, 87% and 70% were generated for the different methods respectively. They found that two out of ten screws were still stripped despite targeting a value of 1.4 Nm, with zero, three and two screws stripping the bone using the other insertion methods respectively. This may be explained by the insertion torque that was targeted, and used as a reference, being beyond the stripping limit of the bone, rather than the technique causing it. However, as no assessment of the maximum torque for the bone samples was performed, this remains unknown. The other study to use visual feedback was performed by Gustafson et al. ¹² They asked five senior surgeons and five attending surgeons to insert eight 4.0 mm cancellous screws into polyurethane bone blocks of 0.1 g/cm³. First, they were asked to insert eight screws to create ‘maximum construct stability’. Screws causing stripping were recorded, being 42% of all insertions for this component of the study. Next, digital torque readings were displayed during insertion for the surgeon to use as feedback. With visual feedback, the rate of bone stripping reduced significantly to 15% (p < 0.001). Visual feedback was then removed, with the stripping rate returning to a significantly (p < 0.007) higher level of 35%, similar to the first part of testing. Awareness of whether the screw holes were stripped was not recorded.

Bone stripping rates

There is very limited research into the rates of bone stripping intraoperatively. A study by Andreassen et al investigated the augmentation of screws if the purchase achieved intraoperatively was determined to be inadequate. ¹⁷ In their selected patients (those over 50 years old with isolated ankle fractures), they found that a synthetic bone void filler was needed for 38% of screws, with 88% of patients having at least one screw that required this. ¹⁷ The remaining data on rates of stripping come from in vitro studies. ^{6,9,10,12–15} The range of mean average stripping rates, when reported, was 0–53% ^10,15 (Table 1), though some individuals within studies stripped more: up to 83%. ¹⁰ Only 19 of the 48 parts of the experiments within studies examining surgical techniques recorded whether screws were stripped, with stripping being confirmed if the torque created after the surgeon had stopped tightening was quantitatively less than the stopping torque. It may be that there was no concern regarding stripping and/or no occurrence of this, explaining why it was not reported. Even when the stripping rates are described, this potentially overlooks the screws inserted beyond the yield torque for the material if recording relies on surgeons’ perception. Dinah et al reported 9% (18/200) of samples being stripped inadvertently; however, they also graded unstripped insertions, finding a further 12% (24/200) were deemed to have been overtightened (90–99% of maximum torque). ¹⁴

Few articles have assessed surgeons’ subjective abilities to detect whether they had stripped the screw. Gustafson et al showed no correlation between occurrence and the perception of stripping (p = 0.768). ¹² With visual feedback, there was increased accuracy in predicting stripping, as one would expect if able to watch a digital readout of the applied torque. However, surprisingly, stripping still occurred in 15% of insertions. No data were provided for the rate of accurate predictions when visual feedback was removed, though it was reported as being significantly (p = 0.008) better than the 6.1% prediction rate at the start of the experiments. Interestingly, when the visual feedback was removed, whilst the improved perception of stripping was partially maintained, there was no significant reduction in the rate of stripping overall, potentially showing a reliance on augmented feedback. This is also the only study that has investigated any retention of a new method or improvement in technique, though over a very short time period, i.e. within the same experimental setting. Stoesz et al found 45% of screws were stripped on insertion, yet only 10 of 109 (9%) of stripped screws were identified correctly. ¹⁰ Identification only occurred when significantly (p = 0.005) past the stripping torque (residual torque being 55% of the maximum, compared to 80% when not detected). Whilst not included in this review as the tightness created was not recorded, one study has attempted to quantify practitioners’ confidence with the screws they inserted. ¹⁹ Siddiqui et al asked one nurse, one junior surgical trainee, one senior surgical trainee and one consultant to insert 4.0 mm partially threaded cancellous screws into chipboard and asked them to assign each insertion (n = 30, 43, 35 and 34 respectively) a score 0–10; with 0 being very weak and 10 being very strong. Each screw then underwent axial pullout. They found correlations between axial pullout force and confidence scoring of r ² = 0.34, 0.26, 0.22 and 0.45 respectively. Unfortunately, the material used, the lack of torsional force assessment or whether samples were stripped on insertion, and the variability in the material properties between different chipboard samples greatly limits the generalizability of these findings to clinical practice.

The effect of surgical experience on screw tightness and stripping rates

Stripping rates appear to be individual to each surgeon, with a wide range of performance. There is both intra- and inter-surgeon variation in insertional torque. ⁹ A stark example of variations between surgeons is seen in Gustafson et al where, when inserting screws without visual feedback, one of their 10 surgeons stripped 16 out of 16 and another 15 out of 16 samples. ¹² Conversely, of their 10 surgeons, two stripped none, and a further two only stripped one sample when there was no visual feedback.

The torsional force applied to screws increases with more surgical experience, ^7,9,16 but this also increases the rate of stripping. ⁹ Wilkofsky et al found that more experienced surgeons applied significantly more torque to screws than either first and second-year (p = 0.003) or third and fourth-year surgical trainees (p = 0.007). ¹⁶ This resulted in a greater tightness of 83±12% compared to 70±19% and 71±20% respectively. Whilst the variation in tightness was less for the most experienced surgeons than other groups, the lateral motion generated whilst creating the higher insertional forces, i.e. non-coaxial insertion or ‘screw wobble’, was approximately 56% larger (p < 0.05). This study also found that the number of screw rotations varied greatly between surgeons, ranging from four to 21 revolutions being needed when inserting a 3.5 mm cortical screw into polyurethane bone. ¹⁶ Apparently contradicting these findings, as previously stated, Acker et al found no significant difference in the applied torque between first (junior) and fifth-year (senior) trainees, nor when compared to senior surgeons. ⁹ Generally, concerns during screw insertion are related to the balancing of the appropriate minimum tightness for the construct to generate sufficient purchase and resistance to failure during locomotion, against overtightening the screw and causing bone stripping. However, more trainee experience appears to lead to an increased chance of this more detrimental, latter situation occurring. ⁹ As no optimum tightness was established for the bone model used in their study (as a function of compression and pullout strength) it is unknown how tight screws should have been inserted, just that stripping the bone should have been avoided. Stoesz et al found no relationship between stripping rates and surgeon experience (p = 0.437), but in a comparison of ten surgeons, there were significant differences in stripping rates between individuals (p < 0.001); ¹⁰ the percentage of samples stripped ranged from 17% to 83%. Seven of the 131 unstripped screws were thought by surgeons to be stripped, however, six of these reports were from one surgeon. These two aspects strongly justify including multiple surgeons in any study investigating techniques in order to reduce the impact from different abilities, or at least ensuring that potential variations between surgeons are taken into account and reported.

Effect of different instructions to surgeons on screw tightness and stripping rates

With the exception of one article that compared four different instructions ¹⁵ and the in vivo study, ¹⁷ the instructions given to surgeons during these studies fall into five categories: subjective feeling of tightness, ^8,11 ‘two-fingers tight’, ^7,9,14,16 optimal for good fixation, ⁶ maximum construct stability ^10,12 or maximum holding force without stripping the bone. ¹³ There are no direct comparisons between different instructions to surgeons to know whether any of these methods make a difference to the techniques employed. ‘Two-fingers tight’ has been reported to be the gold standard for screw insertion, if performed by an experienced orthopaedic surgeon ¹⁵ , and is commonly taught in theatres to trainee surgeons in the USA and Europe. ^9,20 However, the evidence that this technique improves screw insertion is limited, and subsequently, when evaluated, has been shown to be incorrect in that it does not lead to a consistent level of torque being applied. ^9,21 Further to this, previous work, such as Cordey et al, has been reported by others to have used two-finger tightness, ^15,18 despite not defining this in their methods. ⁶ Targeting a specific tightness and comparing surgeons’ ability to repeatably and accurately achieve this target versus other tightnesses has not been investigated, nor has the effect of different instructions on the same physical variables.

Effect of variations in bone density on screw tightness and stripping rates

A common issue with biomechanical research is the model used for testing. It is established that artificial bone models reduce variability, costs and ethical concerns. However, they do not demonstrate many of the biomechanical characteristics of real bone, such as cortical porosity and failure mechanisms, thus limiting generalizability from any research using them. In contrast, in vitro cadaveric human and animal bone models will generate more realistic resistances to screw insertion, but the variability in some models, even between contralateral pairs, ²² means that appropriately powered results can be difficult to generate. Animal models ²³ can address these issues, but ultimately, unless in vivo human bone is used, results may not be fully translatable into clinical practice.

Nicayenzi et al showed no significant difference between human and artificial femora cortices when inserting cortical screws in terms of the stripping torque when normalized to adjust for changes in bone geometries, measured by the bone–screw interface area: normalized stripping torque = stripping torque/(π · screw major diameter · cortical thickness). ²⁴ When comparing maximum torque to the plateau torque during insertion, and comparing these as predictive variables, Reynolds et al found no difference between ovine and human bone maximum torques (p = 0.331). Despite also using synthetic bone blocks, they did not report whether there were significant differences in comparison to these. ²⁵

Aziz et al compared fresh frozen, embalmed and dried adult human humeri alongside normal and low density synthetic bones. ⁸ They found that when one surgeon inserted cortical screws to ‘a subjective feeling of tightness’, there was no difference in the tightness between any of the models used. With cancellous screws, one difference was detected with the tightness being 13% lower (p < 0.05) in fresh-frozen bone than in artificial osteoporotic bone, though the authors reported that all comparisons were underpowered.

Tsuji et al showed, for cortical screws in synthetic bone, as density increased from 0.08 to 0.80 g/cm³, screw tightness decreased (R = –0.63), yet in human cadaveric femora, there was no difference in terms of density changes. ¹¹ For cancellous screws in artificial bone models, they found as density increased from 0.08 to 0.64 g/cm³, screw tightness increased (R = 0.59), yet in human cadaveric femora, the opposite was seen (R = –0.56). This shows the potential variability in the insertion technique of the same individual, given that with different screws in the same material or the same screw in different material, different trends were seen each time. All other studies involving human bone did not have any comparison with artificial models. ^{6,7,13–15,17} Studies just using artificial bones have shown neither an effect on achieved tightness (p = 0.299) nor on stripping rates due to bone density (p = 0.186). ¹⁰