Introduction

The ever increasing interest in 3D visualization and diagnostics concerns all medical fields and specialities. In oral and maxillofacial surgery, cone beam computed tomography (CBCT) in recent years has replaced classical computed tomography (CT) for typical hard-tissue diagnostics such as impacted teeth. Of course, with the increasing availability of this 3D technique, the numbers of CBCT scans rise constantly. However, every CBCT scan that replaces a panoramic radiograph, or is coming on top of that, considerably increases the radiation burden to the patient [1–5]. It would be interesting if the benefit from using 3D diagnostics will indeed outweigh the risk associated with the higher radiation dose, respecting the as low as reasonably achievable (ALARA) principle [6, 7]. Even more so, a quite inevitable side effect of the spread of CBCT machines will be a more frequent use of these techniques for various types of diagnostic questions. Accompanying this more frequent application, claims postulating CBCT as standard technology for certain medical or dental procedures will also occur more frequently. Such a standard procedure may be the removal of mandibular wisdom teeth. Indeed, a growing discussion among experts is noticeable whether a CBCT scan prior to the surgical removal may be indicated. On the other hand, modern medicine no longer believes in expert opinion; rather, it relies on “evidence” gained from studies following a highly reproducible, standardized design with low or moderate bias.

This discussion brought up the idea to initiate a prospective multicentric clinical study to investigate the potential benefit resulting from presurgical 3D visualization and planning on the outcome of the surgery. Since nerve damage is a good measure for complications, our initial approach had been to quantify a potential difference in the appearance of nerve injuries after the removal of lower third molars using 2D panoramic radiography and the 3D CBCT as presurgical diagnostics.

The injury of the inferior alveolar nerve (IAN) and the lingual nerve (LN) are typical risks and most frequent severe complications in the lower third molar (M3) removal; likewise, M3 extraction is the most frequent reason for IAN and LN injury [8–11]. The incidence reported in the literature is ranging from 0.35% [12] to 8% [13] for temporary and from 0% [14, 15] to 1% [16] for permanent IAN injury and from 0% [17] to 11% [18] for temporary and from 0% [15, 19] to 1.1% [20] for permanent LN injury [21–32]. When related to the number of cases (see Table 1), average probabilities of IAN injury of 1.85% (temporary) and 0.53% (permanent) and LN injury of 2.82% (temporary) and 0.46% (permanent) occur. The higher amount of temporary LN injuries is due to the inclusion of studies where the lingual flap technique is used. This technique increases the amount of LN disturbances extremely [9, 33, 34]. When excluding these studies, the figures for a temporary LN disturbance are 1.52% for a temporary and 0.44% for a permanent damage.

Table 1 Overview of neurosensory disturbances of the inferior alveolar nerve (IAN) and the lingual nerve (LN) in percent after lower third molar surgery
Full size table

The LN traverses the region lingually completely through soft tissues and thus is not visible in either radiographic technique [35]. Hence, the IAN only could present a meaningful measure for comparison of both methods. Commonly, the bony canal is likely to be visualized in both imaging techniques.

The anatomic relationship between IAN and M3 is one of the most important information to be obtained during the presurgical evaluation.

In the literature, 2D radiographic risk factors for a neurosensory disturbance of the IAN after surgical removal of M3 are specified, such as darkening of the root, diversion of the inferior alveolar canal, and interruption of the cortical white line, indicating a close relationship between tooth and IAN [36–44].

Clinically, the risk of IAN injury correlates with the exposure of the neurovascular bundle in the operation situs [32, 45–48]. When there is a close relationship between the mandibular canal and the wisdom tooth, the incidence of a nerve injury increases extremely, with a reported incidence ranging between 20% and 60% of these cases [45, 48, 49].

In theory, CBCT should be superior in predicting the exposure, i.e. the contact between nerve and tooth, of the IAN compared to panoramic radiography. A possible absence of the cortical border of the canal can be determined more precisely having the third dimension [46, 50]. Obviously, having available the third dimension allows determining the position of the nerve canal relative to the tooth in all three dimensions. However, Gheaminia et al. [47] did not find a significant difference in the prediction of IAN exposure between CBCT and panoramic radiography. Moreover, several authors in general consider panoramic radiography as a useful tool to predict a contact between the third molar and the IAN [31, 42, 47, 51].

When considering all the arguments listed above, the question arises whether or not the presurgical imaging technique has an influence on the outcome of the study as assessed by an objective measure. Despite the belief of some authors that 3D diagnostic is able to reduce operating time and therefore postoperative complications, such as neurosensory disturbances, swelling, or alveolitis [52, 53], there is only low-level scientific evidence on the actual usefulness of different preoperative imaging techniques for the operative removal of wisdom teeth [49, 54]. In contrary, one study found no significant differences for postoperative complications between 3D and 2D presurgical diagnostics [11]. In intraoral radiography, the parallax technique is able to determine a lingual or buccal position of the IAN in relation to M3 [55, 56], as well as the PA cephalomatric radiography [57].

Apart from that, many studies observed a significant correlation of an IAN injury with the experience respectively skills of the operator [9, 17, 20, 24], the age of the patient [21, 23, 27, 30, 44], and the removal under general anaesthesia for LN injury [23] and for IAN and LN injury [30].

We found no published data revealing an evidence-based effectiveness of using preoperative 3D imaging on the degree of patient morbidity and the overall outcome of third molar surgery. It is unclear if the exact determination of the anatomical position of the IAN in relation to the M3 prior to surgery using 3D diagnostics is required for safe removal with a decrease in neurosensory disturbances.

We discussed how to address this open question in an adequate way. The trial design considerations comprising sample size calculations will be described in this article.

Our hypothesis was that panoramic radiography is not inferior compared to CBCT regarding the postsurgical neurosensory disturbances of the IAN. Testing the superiority of CBCT compared to panoramic radiography, which would also be a feasible way, could only present evidence if superiority is proven. However, if superiority of CBCT is not detected, this would not automatically prove the non-inferiority of panoramic radiography. For this reason, testing non-inferiority of panoramic radiography is the appropriate trial.

Materials and methods

Planned study design

The gold standard to compare treatments is the prospective randomized controlled trial (RCT). In this setting, a comparison of diagnostic imaging techniques with respect to their impact on surgical complications is aimed. The surgeon’s qualification may have a great impact on the risk of neurosensory disturbances, too. One challenge is therefore to distinguish the effects of the imaging technique and the surgical performance. As the surgical performance and other predictors for the risk of complications are expected to be on average similar between the study arms in a randomized controlled trial, an RCT is principally an adequate design.

In order to support external validity of study results, a multicentre approach was chosen [58].

Criteria for inclusion were patient age of 25 years or above, the operation difficulty, which has to be determined with the established Pederson Index, grades 4–10, and patients’ compliance [59–61]. Criteria for exclusion were germs, pregnancy, operation difficulty grade 3 of the Pederson Index, likelihood of an enlacement of the IAN nerve by the root(s), the existence of an external CBCT, and a presurgically existent neurosensory alteration. The surgeon should be experienced in removing lower third molars (a minimum of 150 removals required). The surgical procedure should be standardized by raising a buccal mucoperiostal flap. Bone removal and sectioning of the tooth should be performed when necessary. Possible study endpoints were the operation time (defined from raising the mucoperiostal flap until the finishing of the suture), with an expected average range from 8 to 15 min [62–64], and the percentage of neurosensory disturbances of the IAN. We considered the first postsurgical day to measure the neurosensory disturbance to also include short-term temporary disturbances. Since the LN traverses completely through soft tissues and is not visible in either radiographic technique, only the neurosensory disturbance of the IAN presents a meaningful measure for comparison of both methods and was therefore defined as the primary endpoint of the planned trial. We wanted to use an objective method, the sharp/blunt discrimination with a dental probe, to characterize neurosensory disturbance [65, 66]. As a secondary endpoint, we planned to consider the operation time as a continuous variable and indicator for overall operation difficulty.

Our hypothesis is that 2D imaging is not inferior to the 3D diagnostic with respect to the risk of temporary neurosensory disturbances of the IAN (primary endpoint). We decided to define a relative non-inferiority bound of 10% and so allowing the rate of nerve damage in the panoramic radiography group to be up to 10% higher relative to the CBCT group. Methodological and ethical considerations substantiate the need for a sample size calculation. If an assumed effect cannot be demonstrated within a study due to an insufficient number of patients, study participants would be exposed to unnecessary radiation. On the other hand, if an effect could have been demonstrated even with a lower number of patients, more patients than necessary have been exposed to the inferior treatment, and more time was needed to assess the question of interest [58, 67–69].

Sample size calculation

Basing on the in-depth analysis of data from the literature, we calculated the required sample sizes, i.e., the number of third molar removals, under the following assumptions:

Temporary neurosensory disturbance rates of the IAN are assumed to be 1.85% for both CBCT and panoramic radiography patients. A power of 80% is required and a one-sided type I error of 2.5% is allowed. For the panoramic radiography group, disturbance rates up to 10% higher relative to the CBCT group are considered non-inferior. Sample sizes should be equal in both trial arms.

To illustrate the effects of single parameters on required sample sizes, the relative non-inferiority margin and type I and II error were modified. Besides permanent neurosensory disturbance rates of the IAN were considered.

The same calculations were performed for permanent neurosensory disturbance rates assuming 0.53% for both CBCT and panoramic radiography.

Sample sizes were calculated by means of PASS 2002 (Number Cruncher Statistical Systems, Kaysville, UT, USA).

Results

The required sample size, i.e., the number of third molar removals, for temporary IAN injury is 183,474 under the assumptions considered most adequate (power 80%, type I error 2.5%, relative non-equivalence margin 10%). Modifying these parameters in different ways (power, 90%; type I error, 5%; relative non-equivalence margin, 20%) changes the required numbers of third molar removals considerably yielding a range from 39,584 to 245,724 (Table 2, Fig. 1). For the permanent IAN injury, the corresponding sample size was 649,036 ranging from 140,024 to 869,250 (Table 3, Fig. 2).

Table 2 Sample sizes (the required number of third molar removals) for temporary IAN (inferior alveolar nerve) injury
Full size table
Fig. 1
figure 1

Sample size calculation (the required number of third molar removals) for temporary IAN (inferior alveolar nerve) damage depending on power, specified ratio, and alpha (type I error). The required sample size increases with decreasing alpha and specified ratio as well as increasing power

Full size image
Table 3 Sample sizes (the required number of third molar removals) for permanent IAN (inferior alveolar nerve) injury
Full size table
Fig. 2
figure 2

Sample size calculation (the required number of third molar removals) for permanent IAN (inferior alveolar nerve) damage depending on power, specified ratio, and alpha (type I error). The required sample size increases with decreasing alpha and specified ratio as well as increasing power

Full size image

Discussion

The results of the sample size calculation show that proving non-inferiority of panoramic radiography compared to CBCT with respect to neurosensory disturbance rates would be a definite challenge. Huge sample sizes are mainly a consequence of the rare event rates in both trial arms. Many patients must be allocated to the trial until the first neurosensory disturbance is observed. To show that rare event rates do not diverge in terms of a relative difference definition, very precise estimates are necessary. For example, 10% relative difference implies 5% allowed absolute difference if event rates are assumed to be 50%. If the expected event rate is 1%, 0.1% is the maximum allowed absolute difference.

Even if it was possible to conduct the trial, it would be difficult to get robust estimates for the surgeon effects. Assuming an event rate of 1.85%, a surgeon would have to remove 54 third molars until the first neurosensory disturbance is expected to occur. Therefore, it would be desirable not to allow for too many participating surgeons, which would, on the other hand, limit external validity.

It should be noted that proving superiority of CBCT would require high sample sizes as well in this situation—for the same reasons as described above. The negative outcome of our initial approach also bears some significance for the question that was standing at the beginning of all planning: Is CBCT superior with respect to the (measurable) clinical outcome of lower third molar surgery? Consequently, the aim of this report was to provide some insight into the sample size necessary to answer this particular question and on the conclusions, which may be drawn from these purely theoretical considerations.

The injury of the IAN is a typical but rare complication of third molar surgery, particularly when a permanent disturbance is regarded. We do not believe that the low incidence of IAN and LN injury could be reduced using CBCT. The risk of damage is predetermined by the anatomy. When IAN penetrates through the root, 67% injuries occurred although CBCT was performed [70]. Panoramic radiography occasionally overestimates the risk of IAN injury compared with 3D imaging [48, 71].

M3 removal is recommended to be carried out until the 25th birthday to minimize postsurgical complications [72]; however, the sensitivity to radiation is significantly higher in younger ages [73]. Hence, for this age group, the ALARA principle should be followed with extra caution. No matter how low the dose, it is excessive if it is unlikely to improve the outcome of the treatment provided [6].

The effective doses (ICRP 2007) of the different imaging techniques are reported as follows: CBCT, 13–1,073 μSv [3, 5, 74]; Panoramic radiography, 8.6–24.3 μSv [4, 75, 76]; PA cephalometric radiography, 2.3–5.1 μSv [4, 77]; intraoral radiography, 0.3–5.5 μSv [4, 75]; and head CT, 470–4,000 μSv [73, 74, 76, 78] (Table 4).

Table 4 Effective doses (ICRP 2007) (μSv) for cone beam computed tomography (CBCT), panoramic radiography (PR), PA cephalometric radiography (PACR), intraoral radiography (IR), and head CT (HCT) [3, 4, 5, 73–78]
Full size table

In one study, the vertical position of the root tip relative to the mandibular canal could not be correctly determined in both CBCT and PA cephalometric radiography in combination with panoramic radiography for 1% of the cases [57]. The horizontal relation could not be identified exactly in 9.0% for panoramic radiography and PA cephalomatric radiography versus 2.8% for CBCT [57]. 3D diagnostics were not able to detect the relationship between IAN and M3 in every case of the investigation [48, 50, 57, 70]. Hence, CBCT does not guarantee for an exact 3D location of the IAN in relation to the M3.

The increasing information in CBCT datasets in comparison to panoramic radiographs is undisputed, but it cannot be an indication for its own sake because of the increasing radiation dose, yet without any proof of benefit for the patient. Removal of M3 seems to be a highly effective procedure when done by experienced surgeons, with a very low rate of permanent nerve injuries, even when using panoramic radiography as presurgical imaging technique. A use as routine diagnostic cannot be justified for CBCT because of the absence of evident benefit [7].

In conclusion, from our theoretical calculations based on data published in the literature, we cannot recommend CBCT scans prior to surgical removal of lower wisdom teeth. The major negative outcome, a (temporary or permanent) damage of the alveolar nerve, is such a rare event even when 2D imaging is the basis for the procedure, which the sample sizes required to prove a difference would be exceedingly large. The low rate of complications when experienced surgeons perform the surgery does not warrant the currently approximately tenfold radiation dose imposed by CBCT when compared to panoramic radiography. This is of particular importance, since the majority of patients with lower third molar surgery are under 25 years of age.

A remaining indication for CBCT are the so-called high risk extractions, for example, when the parallax technique and the PA cephalomatric radiography could not determine the location of the mandibular canal properly or an enlacement of the IAN nerve by the root(s) can be expected to be very likely.