In the present pilot study, accuracy of implant positioning following flapless computer-guided surgery has been correlated to clinical and radiological variables that commonly define peri-implant tissue health. The aim was to test the null hypothesis that there would be no statistically significant differences between the level of accuracy obtained during the surgical procedure and the peri-implant health status. Results were heterogeneous, as the null hypothesis has been accepted in some cases, and rejected in others. In particular, the variables that proved to be statistically significantly correlated with inaccuracy were pi-MBR, KM, BOP, and PPD.
In terms of pi-MBR, marginal bone levels measured after 12 months and 5 years from the prosthetic loading were analyzed on orthopantomographs. The data collected from 33 implants still in function at 5 years, showed a mean pi-MBR of 1.1 mm at 12 months and approximately 2 mm at 5 years. These values are in line with the 2017 World Workshop on the Classification of Periodontal and Peri-Implant Diseases and Conditions definitions of implant health  and Albrektsson's 1986 success criteria [12, 13]. When compared with the latter, the mean value of pi-MBR at 12 months after prosthetic loading observed in the present study is not only within the normal range, but is also slightly lower than the mean value reported by Albrektsson et al., which is 1.5 mm . Interestingly, pi-MBR remained rather stable over time, being approximately 2 mm at 5 years, confirming a positive prognosis for the medium-term hard tissue health at implants placed with computer-guided flapless surgery. In this matter, the criteria of implant success described by Albrektsson in 1986 accepted an annual resorption of peri-implant marginal bone of 0.2 mm after the initial remodeling phase . In the present study, the differences in pi-MBR obtained between 12 months and 5 years of function were also analyzed, yielding an overall bone loss of 0.84 mm during 4 years, transposed into a 0.21 annual pi-MBR. Although this value approximates that mentioned above, it remains only hypothetical, as radiological data were not collected on an annual basis. Therefore, any progression or acceleration of the pi-MBR throughout the years cannot be extrapolated from the present data. This constitutes a limitation of the present study. Nonetheless, the results observed herein are consistent with those reported in similar studies. In 2013, Marra and co-workers published a study evaluating pi-MBR after a 3-year follow-up period in 30 fully edentulous patients who had been treated with computer-guided flapless implant surgery . A total of 312 implants were analyzed, and a mean pi-MBR of 1.9 ± 1.3 mm was reported at 3 years. A paper published in 2017 by Lopes et al. evaluated pi-MBR in edentulous patients treated with implant-supported fixed total rehabilitations with All-on-Four concept by means of computer-guided flapless implant surgery with a 5-year follow-up . Overall, 111 patients were included in the study, and the average marginal bone loss calculated after 5 years was 1.27 mm for tilted implants and 1.34 mm for axial implants. Better results were obtained by Tallarico and co-workers who published a study comparing pi-MBR values after a 5-year follow-up from prosthetic loading in edentulous patients rehabilitated with implant-supported prostheses undergoing computer-guided flapless surgery . The mean pi-MBR observed after 5 years was 0.87 mm ± 0.40. According to the results of the present study and those found in pertinent literature, it appears that marginal bone levels obtained after medium-term prosthetic loading following flapless computer-guided workflows are stable and predictable over time.
Apart from the pi-MBR values reported to give an overview on the stability of peri-implant hard tissue, data on the accuracy achieved with computer-guided implant surgery were also pivotal in the present study. Accuracy is intended as the matching of the implant position planned within the software with that actually obtained in the patient's mouth. The linear deviations of the implant head and apex measured in the bucco-lingual/palatal, mesio-distal, and corono-apical directions, together with the angular deviation of the implant long axis were considered. It is known that the level of accuracy strongly depends on the reliability and precision of the workflow and methodology used during the planning phase, from the diagnosis up to the surgical step. Every aspect must be carefully developed in order to reduce the margin of error in all the steps that characterize the planning phase and the operative phase. An adequate precision and accuracy of all these sequential steps is therefore of paramount importance considering that each error is cumulative and is transferred to the subsequent steps. A review published by Bover-Ramos and colleagues analyzed 34 articles providing 3033 implants placed with partially and fully guided surgery in vitro (8 studies), in cadaver (4 studies), and in vivo (22 studies) . The data regarding the accuracy of implants placed in patients with fully guided surgical protocols showed an average implant head deviation of 1.00 mm ± 0.08 mm, an average apical deviation of 1.35 mm ± 0.12 mm and an angular deviation of 3.62° ± 0.29°. No statistically significant differences were also found between the accuracy of implants placed with fully guided surgery in vivo and on cadavers. Marlière and co-workers published a systematic review that included 7 studies realized between 2011 and 2016, in which they evaluated the accuracy of implants placed with computer-guided surgery in patients with total rehabilitations . Angular deviation ranged from 1.85° to 8.4°, implant head deviation fell within a range of 0.71 mm − 2.17 mm, and apical deviation showed an interval of 0.77 mm − 2.86 mm. The systematic review with meta-analysis published by Schneider and colleagues included 8 studies related to implant placement with computer-guided flapless surgery, in which accuracy was also calculated . Considering the in vivo studies only, the mean deviation at the implant head was 1.16 mm, the mean apical deviation was 1.96 mm, and the mean angular deviation was 5.73°. The values found in the present pilot study showed a mean linear deviation of the implant head of 0.57 mm, a mean linear deviation of the implant apex of 0.69 mm, and a mean angular deviation of the long axis of 2.88°. These values were lower when compared to those reported in the systematic reviews mentioned above, meaning that a high level of accuracy was achieved with the workflow described herein. Multiple reasons have been identified: traditionally, intra-oral gutta-percha markers placed inside the radiographic stent were used to integrate the prosthetic plan with the patient's anatomy in virtual planning. However, in the presence of metal prosthetic restorations, the identification of the radiopaque marker given by the gutta-percha is challenging .
In the present study, an extra-oral radiopaque marker was used, consisting of a well-defined geometric device. A total of 30,000 points were scanned and superimposed during the matching procedure, with greater accuracy in overlapping DICOM data and radiographic stent than traditional protocols.
Additionally, in traditional protocols, two CT scans were usually performed, one of the patient and one of the radiological model. In the present study, the radiological template was scanned using an optical scanner, which provides STL data, which are more accurate than DICOM data because they are independent of the Hounsfield unit threshold based on the radiologist-defined gray-level segmentation.
This has allowed the surgeon to determine the exact thickness of the soft tissue, resulting in more accurate virtual planning.
An optimal level of precision was also achieved through fixation of the surgical template to the surgical site.
The surgical protocol performed involved the insertion of vestibular endosseous pins disposed in tripod formation, whose position and depth was guided by special sleeves, previously established not to interfere with the positions of the implants.
This way it was possible to avoid movements and deformations of the surgical template caused by the pressures promoted during the preparation of the implant sites.
Again, in order to achieve a higher level of accuracy in the present protocol, disposable drills were used.
In fact, by increasing the cutting capacity, the risk of possible deviations in the osteotomies, caused by excessive wear of the drills, is reduced. 
The high rate of accuracy found in the present study could explain the medium-term stability of the marginal bone profile evaluated at the 5-year follow-up. On the other hand, the linear deviation of the implant head and apex in the bucco-palatal/lingual direction was significantly correlated to higher values of pi-MBR. A similar trend was also noted for the peri-implant keratinized mucosa. In this case, implants placed more buccally compared to the virtual plan were more prone to develop a contraction of peri-implant hard and soft tissues. A translation of the displacement vector in the vestibular direction could have led to an excessive remodeling of the buccal cortical thickness, with a consequent reduction in the amount of peri-implant keratinized mucosa . In this respect, Monje and colleagues analyzed the critical threshold value of the thickness of the buccal cortical wall to prevent pathological resorption . In particular, an animal study was conducted in which 36 implants were inserted in sites presenting a residual buccal cortical plate < 1.5 mm, while 36 implants were placed in sites that maintained ≥ 1.5 mm of buccal bone. There were no implant failures, but it was interesting to note that implants placed too buccally, with a residual buccal bone < 1.5 mm, showed more recession of the buccal mucosa. Interestingly, a recent study published by Romandini and co-workers confirmed that an implant placed too buccally is associated with an almost three times higher risk of presenting peri-implantitis. This was related to the reduced thickness of the residual buccal bone, which is likely to resorb and provoke mucosal recession, with consequent exposure of the implant surface and increased risk of bacterial colonization . These observations somehow corroborate those reported herein. Although mucosal recession was not mathematically associated to buccal inaccuracy, a contraction of the peri-implant keratinized mucosa was noted. It can be speculated that with a longer follow-up period, more soft tissue contraction may occur, leading to mucosal recession in accordance with the previous studies.
In view of the scientific evidence reported above, the values found in this study regarding the correlation between buccal implant displacement and higher pi-MBR together with a contraction of the keratinized mucosa are interesting. Implants inserted with less accuracy in the bucco/palatal-lingual direction and particularly in a position that is too buccal compared to the initial planning showed greater pi-MBR and less keratinized mucosa. This in turn may lead to an increased risk of developing peri-implant inflammation. Accordingly, the data collected herein supported the fact that inaccuracy of implant positioning in the bucco/palatal-lingual direction had a statistically significant effect on the presence of peri-implant bleeding on probing. In this respect, Perussolo and colleagues investigated the correlation between reduced peri-implant keratinized mucosa (< 2 mm), pi-MBR, plaque, bleeding, and patient discomfort in home oral hygiene procedures . Implants with a reduced amount of keratinized mucosa (< 2 mm) were found to be associated with greater pi-MBR, a higher degree of soft tissue inflammation in terms of plaque and bleeding on probing, and finally, greater difficulty in daily home cleaning by the patient. These results are consistent with those obtained in the present work, as pi-MBR, keratinized mucosa and bleeding on probing were all interconnected and strictly correlated to buccal implant displacement. Although domiciliary compliance was not evaluated in this study, it is likely that buccally placed implants induced a contraction of the keratinized mucosa, leading to more brushing discomfort with consequent plaque accumulation, biofilm-related inflammation, and ultimately bone loss. This favorably complies with a study published by Souza and co-workers who concluded that implant sites with a band of < 2 mm of KM were shown to be more prone to brushing discomfort, plaque accumulation, and peri-implant soft tissue inflammation when compared to implant sites with ≥ 2 mm of KM .
All of this taken together may also explain the significant correlation between inaccuracy and PPD. Indeed, those implants inserted too buccally, that presented with a state of inflammation detected by bleeding on probing, also showed higher values of PPD compared to implants that were placed more accurately. It should be mentioned that, when defining the health of an implant, probing depth values might vary as they are dependent on the peri-implant soft tissue height . Accordingly, a narrative review published by Coli and coworkers aiming at analyzing the correlation between probing depth and peri-implant health, concluded that there is no precise threshold value of PPD that can indicate the presence of disease . In the present work, PPD ranged from 1 to 7 mm, emphasizing the fact that PPD might not be solely correlated to implant health or disease but can also be attributed to other variables not assessed herein. However, the fact that implants placed more buccally were also associated to higher PPD values as a potential consequence of higher inflammation and bone loss is worthy of note.
It should be stressed at this point that the findings reported in this study should not be overgeneralized due to the small sample enrolled. Indeed, although interesting observations emerged from the data, the lack of external validity remains a limitation of the present research. Therefore, further studies including a larger sample of patients will be needed to validate or deny the correlation between accuracy and the variables analyzed in the present study.