Fifty-seven patients (40 females and 17 males, age 20 to 65 years, mean age at implant insertion: 48.2 years) with periodontitis were treated at the School of Dental Medicine, Philipps University, Marburg/Lahn. A 3-month recall schedule followed the systematic periodontal treatment for 4 to 6 years. A detailed description of the periodontal treatment procedure was published previously [12, 13].

Before implant placement, all patients had healthy periodontal tissue with no bleeding on probing (BOP) and probing depth (PD) < 3 mm. The eligibility criteria for participating in the present study were evaluated between 1992 and 2017. The inclusion/exclusion criteria were described in detail previously [12, 13].

Patients were classified as stages I-IV according to the AAP/RFP definition based on periodontal disease severity and progression [14], specifically the interproximal radiographic bone loss. Stage was determined by the worst parameter or site. If the bone loss at the most severe site (mesial or distal) was < 15%, it was designated stage I, 15–33% was designated stage II, and > 33% was stage III/IV. The diagnosis was adjusted for complexity using a pocket depth ≥ 6 mm, the presence of an intrabony defect ≥ 3 mm deep, and the presence of furcation (degree II or III). Grade A, B, or C was added to the diagnosis based on the ratio of the percentage bone loss radiographically to patient age. If the ratio was < 0.25, it was designated grade A, 0.25–1.0 was designated grade B, and > 1.0 was designated grade C. We determined the disease extent as the number of teeth with clinical attachment loss divided by the total number of teeth; generalized disease was defined as > 30% and localized disease as < 30%. Regarding periodontal disease severity, 20 patients were classified as stage I and 37 patients as stage II. Referring to the progression of periodontal disease, 29 patients were classified as grade A and 28 patients as grade B.

Patient treatment was carried out in accordance with the Declaration of Helsinki by the World Medical Association (version VI, 2002). This study was approved by the Ethics Committee Marburg/Lahn (ek_mr_11_07_2017_ mengel).

A total of 162 dental implants were inserted epicrestally: 45 with a smooth surface (Brånemark® Mk II and III, Nobel Biocare, Zurich, Swiss) and 117 with a rough surface (Nobel Replace® Straight Groovy and Nobel Speedy® Replace, Nobel Biocare, Zurich, Swiss). The implants were 8.5 to 15.0 mm in length and had a diameter of 3.75 to 5.0 mm (Table 1). All implant placements were prosthetically driven using pre-fabricated surgical stents. The bone quality and quantity were classified during insertion of the implant [15]. Twenty-one patients received 1 implant, 18 patients received 2 or 3 implants, and 18 patients received 4 or more implants. In 8 patients, the deficient buccal bone walls of 15 implants were augmented during insertion with autologous bone obtained from the surgical site and xenogenic bone substitute material (Geistlich; BioOss Spongiosa, Wolhusen, Swiss). The implants and augmented materials were covered by a xenogenic membrane (Geistlich; BioGide, Wolhusen, Swiss).

Second-stage surgery was performed after 3 or 6 months in the mandible and maxilla, respectively. No additional connective tissue or epithelial grafts were applied. Implant insertion and second-stage surgeries were performed according to the manufacturer’s guidelines by the same periodontist (RM). Four weeks after second-stage surgery, the patients were treated with fixed single crowns (n = 123) or bridges (n = 19). Screw-retained (n = 37) or cemented (n = 125) restorations were used based on the clinical situation and preference of the clinician. All bridges and crowns consisted of a high-gold metal framework veneered with ceramic or full ceramics. Prosthetic treatment was performed at the Dental School of Medicine, Philipps University, Marburg/Lahn.

The first clinical examination was carried out immediately after the final superstructure was inserted and considered to be baseline. Subsequently, patients were treated for 2 to 20 years on a 3 to 6-month recall schedule (Table 2). A detailed description of the treatments in the recall was published previously [12, 13]. All patients were in the recall program for at least 2 years. In addition, 53 of the patients (150 implants) were followed for 5 years, 35 patients with 96 implants for 10 years, 21 patients with 39 implants for 15 years, and 9 patients with 18 implants for 20 years.

A periodontal probe (UNC-15; Hu-Friedy, Chicago, IL, USA) was used for clinical measurements at six sites (buccal, mesiobuccal, distobuccal, lingual, distolingual, mesiolingual) for each tooth and implant. We also investigated the plaque index (PI) [16], gingival index (GI) [17], PD (in mm), and BOP (in %). Clinical examinations were performed by experienced examiners (n = 10) calibrated for intra-examiner (correlation coefficients 0.98 to 0.98) and inter-examiner reproducibility (correlation coefficients 0.96 to 0.97). Routine calibration sessions in which a minimum of 50 sites were measured in duplicate in at least five patients were scheduled every 12 months.

The gingival phenotype of each patient was determined at the central anterior teeth of the maxilla. If these teeth were prosthetically restored or missing, the lateral anterior teeth were used, otherwise we used the canines. All patients had to have at least four natural teeth in the maxillary anterior region at the time the gingival phenotype was determined.

The soft tissue thickness was measured centrally on the buccal side of the tooth using a standard periodontal probe (DB765R, Hu-Friedy, Chicago, IL, USA). The probe was single-ended and color-coded with black markings (1 to 15 mm) and #30 handling. To accomplish this, we observed the periodontal probe by transparency through the gingival tissue after insertion into the gingival sulcus at a depth of 1 mm [18]. We used the visibility of the tip to classify the gingival thickness: thin if the tip was visible, thick if the tip was not visible. All assessments were performed without any magnification in natural light. The oral cavity was not illuminated to avoid light scattering or interference in observing the gingival transparency.

The contour of the gingival margin was determined by the papilla height mesially and distally to the central maxillary anterior teeth. For this purpose, the periodontal probe (DB765R, Hu-Friedy, Chicago, IL, USA) was placed perpendicular to the junction line between the buccal sulcus up to the papilla tip to assess papilla formation [19]. A complete papilla, which filled the proximal space completely, presented with a scalloped gingival margin, whereas incomplete papilla presented with a flat gingival margin.

The width of the attached gingiva was measured mid-buccally at teeth using the periodontal probe (DB765R, Hu-Friedy, Chicago, IL, USA). The evaluation represents the distance (in mm) between the mucogingival junction and free gingival margin. The alveolar mucosa was stretched several times to identify the mucogingival junction. A narrow attached gingiva had a width of ≤ 2 mm and a wide attached gingiva had a width > 2 mm. The gingival phenotype was categorized into three classes based on the thickness of the gingiva, scalloped gingiva, and width of the attached gingiva:

(a) Gingival phenotype 1 (thin-scalloped gingival phenotype) with thin, scalloped gingiva.

and narrow attached gingiva. (Figure 1a and b)

a Implant lateral incisor 5 years of insertion of crown. Gingival phenotype 1 with thin, scalloped gingiva and narrow attached gingiva. b Intra-oral radiograph 5 years of insertion of crown

(b) Gingival phenotype 2 (thick-flat gingival phenotype) with thick, flat gingiva and wide.

attached gingiva. (Figure 2a and b)

a Implant frontal incisor 12 years of insertion of crown. Gingival phenotype 2 with thick, flat gingiva and wide attached gingiva. b intra-oral radiograph 12 years of insertion of crown

(c) Gingival phenotype 3 (thick-scalloped gingival phenotype) with thick, scalloped gingiva and narrow attached gingiva. (Figure 3a and b)

a Implant lateral incisor 8 years of insertion of crown. Gingival phenotype 3 with thick, scalloped gingiva and narrow attached gingiva. b intra-oral radiograph 8 years of insertion of crown

The clinical measurements for gingival phenotype were performed by an experienced and trained examiner (NB) calibrated for intra-examiner reproducibility as described above for clinical examinations.

Standardized intra-oral radiographs were taken of all implants by experienced radiologists using a parallel long-cone technique with Rinn-holders (XCP Instruments, Rinn Corporation Elgin, IL, USA). Customized individual bite registration was not used. Radiographs were obtained at baseline (immediately after superstructure insertion) and then after 1, 3, 5, 15, and 20 years. Initially, each analog radiograph was framed as a slide and digitized at a resolution of 675 dpi using a SnapScan slide scanner (Agfa, Mortsel, Belgium). However, images were taken digitally (Planmeca, Helsinki, Finland) after 2005. All digitized radiographs were evaluated by an independent masked examiner using computer software (Planmeca Romexis, Helsinki, Finland). Before measurement, the intra-examiner reproducibility was calibrated as noted above for other examinations. The radiograph was calibrated by the implant length or the thread of the implant. If the threads were not clearly visible on the radiograph and calibration was not possible, the radiograph was not used. At implants, the distance from the first apical contact between bone and implant to the implant abutment connection was measured mesially and distally (in mm) and related to the implant thread.

All data for all patients were stored in the NBImplant database of the Coordination Center for Clinical Studies (KKS), Philipps University, Marburg/Lahn. Statistical analyses of the clinical and radiographic results were performed using the program R for Windows version 3.4.1. The implants were the evaluation units of the statistical analyses. Normal Q-Q plots and the Shapiro-Wilk test indicated a non-normal data distribution, so nonparametric methods were used for statistical analyses. The implant survival rate was determined by Kaplan-Meier analysis. The survival rates of the three subgroups were compared by the log-rank test. The alpha level of the study was p = 0.05. Clinical parameters were analyzed for all implants together and separately according to gingival phenotype. To compare the means between the gingival phenotypes, a nonparametric Kruskal-Wallis test was used. The Dunn test was used as a multiple comparisons test. Significant levels were corrected by Bonferronie Holm. The Spearmen non-parametric correlation analysis was used to analyze the correlation of clinical parameters with increased crestal bone loss (BLI > 0.1 mm/year) compared with the measured value from the previous year. Qualitative classification of the correlation into effect sizes was achieved using the Cohen analysis.

The diagnosis of peri-implant diseases during the entire study period was determined by clinical and radiological results [12]. We used a peri-implant mucositis definition of PD ≥ 5 mm with BOP or GI ≥ 2 without annual bone loss, whereas peri-implantitis was defined as PD > 5 mm with or without BOP or GI ≥ 2 and annual bone loss > 0.2 mm. This definition of peri-implantitis requires an annual determination of bone loss and results in a more accurate determination of bone loss than with the current AAP/RFP definition [20].