In vivo fractures of endodontically treated posterior teeth restored with amalgam Hansen EK, Asmussen E, Ghristiansen NG. In vivo fractures of endodontically treated posterior teeth restored with amalgam. Endod Dent Traumatol 1990; 6: 49-55. Abstract - The cumulative survival rate (retention of both cusps) and the fracture pattern of 1639 endodontically treated posterior teeth were assessed in a retrospective study. All teeth had an MO/DO or an MOD cavity restored with amalgam without cuspal overlays. The 20-year survival rate of teeth with an MO/ DO cavity was markedly higher than that of teeth with an MOD cavity. The lowest survival rate was found for the upper premolars with an MOD cavity: 28% of these teeth fractured within 3 years after endodontic therapy, 57% were lost after 10 years, and 73% after 20 years. Generally, the cusp most prone to fracture was the lingual one, and lingual fractures caused significantly more damage to the periodontal tissues than did facial or total crown fractures. The severity of periodontal damage increased with posterior location of the tooth. By far the most serious failures, irrespective of the cavity type, were found for the upper second molar, as 10 of 29 fractures led to extraction. It is concluded that amalgam, especially in MOD cavities, is an unacceptable material for restoration of endodontically treated posterior teeth if used without cuspal overlays.
Tooth fracture is a frequent problem in restorative dentistry. Both sound and restored teeth may fracture, but especially those endodontically treated are at risk (1-3). The main problem is that the fracture often extends subgingivally and not infrequently results in loss of periodontal attachment. Subsequent operative procedures for impression may cause futher damage to the periodontal tissues. Also, a subgingival restoration margin, no matter how well adapted, nearly always provokes an inflammatory periodontal reaction (4-8). In a previous, retrospective investigation (3), the frequency of cusp and crown fracture of endodontically treated premolars with MOD cavities was studied. The 3-year survival rate was markedly improved when the cavity had been restored with an enamel-bonded resin filling instead of amalgam. But after 3 years, tooth failure of the resin-restored preinolars happened with nearly the same frequency is that found for premolars restored with amalgam, in that study (3), however, the number of resinestored teeth was rather small (n = 40). The purpose of the present retrospective study
Erik Keith Hansen', Eriit Asmussen\ Niels C. Ciiristiansen^ Institutes of ^Dental Materials and Technology and ^Prosthetic Dentistry, Royal Dental College, Copenhagen, Denmark
Key words: amalgam; cavity preparation; endodontic therapy; tooth fracture. Dr. odont. Erik Keith Hansen, Helsingorsgade 7, DK-3400 Hillerod, Denmark. Accepted for publication September 15, 1989.
was to examine the cumulative survival rates (retention of both cusps) of endodontically treated premolars and molars restored with MO/DO or MOD amalgam without cuspal overlays, and to elucidate the fracture pattern of these amalgam-restored teeth. iWaterial and methods The data were collected from January to October 1988 from 91 dentists working as general practitioners. The results from the previous investigation (3) were updated and included in the present study. All data were derived from clinical examinations and review of the patients' dental records and radiographs. The criteria for including data were: (i) an endodontically treated premolar or first or second molar with an MO, a DO, or an MOD amalgam restoration; (ii) no cuspal amalgam overlays; (iii) no Glass V restoration; (iv) normal functional occlusion; (v) complete dentition in the anterior and premolar region, including cases with dental bridges restoring 49
Hansen et ai. one tooth in the premolar region or 2 in the anterior region; and (vi) at least one molar in each quadrant involved in the occlusion. Endodontically treated teeth without antagonist were not included. The dentists were asked to record the following data on a registration form: tooth number, date of endodontic therapy, cavity type (MO, DO or MOD), and the date of control or last contact. In cases of fracture, the dentists were asked to record the date of tooth failure and whether the facial cusp, the lingual cusp, or the whole crown (total fracture) was lost. Finally, the dentists were asked to record whether the fracture was supragingival, subgingival, at or beyond the alveolar crest, or whether the fracture was so vertical that the tooth had to be extracted. In this article, fractures at or beyond the alveolar crest, but not vertical fractures resulting in extraction, are referred to as subcrestal, even though some of these fractures may have had a fracture level slightly above the alveolar crest. The trial time was defined as the time elapsed between the date of endodontic therapy and the date of fracture; the date of final registration; the date of withdrawal (change of cavity, change of restorative material, extraction of the antagonist); or the date of last contact with the patient. In the analyses, no distinction was made between MO and DO restorations: only between MO/DO and MOD. When MO/DO restorations had been replaced by MOD fillings, the teeth were recorded as withdrawn MO/DO teeth and re-entered as new MOD teeth. The total trial time was set to 20 years with intervals of 1 year. For each interval, the effective number of teeth exposed to risk of fracture was calculated. The cumulative survival rate, i.e. retention of both cusps, for each tooth number with an MO/DO or MOD cavity was then calculated by means of life table analysis (9). The classification based upon tooth number and cavity type gave 16 different combi-
nations of tooth and cavity type: 4 premolars and 4 molars, each with 2 cavity types. Differences between these 16 combinations were analyzed with log-rank tests (9) at the 5% level of significance. Analyses of the fracture pattern were done with Kruskal-Wallis one-way analysis of variance, the Mann-Whitney U-test, Fisher's exact probability test (10) and the Jonckeera-Terpstra test (11); the latter analysis can be described as a Kruskal-Wallis test for trend. The significance level for the fracture pattern was set to 1%; the reason for this will be explained in Results. Most of the analyses were carried out with 2 computerized statistical programs (SAS, version 6.03, SAS Institute, Gary, NG, USA; and MEDSTAT, version 2.1, Astra, Gopenhagen, Denmark). The 95% and 99% confidence limits were calculated using MEDSTAT.
Resuits Data were obtained on 1695 endodontically treated teeth with MO/DO or MOD cavities; all teeth were restored with amalgam. However, 56 teeth were rejected because the cavity type or the date of endodontic therapy or control was not recorded or because of cuspal overlay. A further 55 sets of data were partly rejected because some of the dentists misunderstood the instructions and only recorded fractured teeth, not fractured and non-fractured at random. These 55 teeth were excluded from the Table 2. Cumulative survival rates (%); 95% confidence limits in parentheses Tooth number* Cavity type Group
MO/DO
MOD
44 45 17 47 16 14 46
Table 1. Number of endodontically treated teeth (n) in the survival analyses distributed on cavity type and fracture mode. The number of fractures are those found within 20 years after endodontic therapy
Tooth number*
14 15 16 17 44 45 46 47
Fracture mode Facial
Lingual
Total
Vertical
MO/DO MOD
MO MOD
MO MOD
MO MOD
MO MOD
5 4 7 1 3 2 3 3
10 10 11 2 1 5 6 10
79 64 98 64 43 85 68 75
132 203 194 57 25 123 215 59
34 26 12 4 3 12 10 3
23 72 55 10 4 29 72 17
3 5
7 6 2 2 2 8 2 1
1 1 3
1 3 3 7 1 1 1
* = Viohl's two-digit system. For the sake of clarity, the tooth numbers are only those from the right side.
50
A B C
10 years
20 years
89 (77-100) 90 (78-100) 90 (76-100) 81 (67-96) 80 (68-92) 81 (69-92) 80 (62-97)
85 (60-100) 82 (61-100) 74 (46-100) 73(42-100) 71 (46-96) 67 (45-90) 64 (20-100)
45 17 16 46 47
87 (74-100) 83 (72-93) 81 (73-88) 77 (66-89) 87 (82-92) 80 (74-86) 89 (80-98)
67 (46-88) 62 (44-81) 58 (47-69) 54 (35-74) 66 (57-76) 58 (49-68) 59 (41-78)
53 (18-89) 49 (9-88) 47 (30-63) 36 (13-59) 34(18-51) 34(7-61) 31 (2-60)
14 15
75 (67-83) 70 (63-77)
50 (3&-61) 39 (30-48)
28 (12-44) 27 (13-41)
90 (87-93) 83 (80-86) 72 (67-77)
84 (80-88) 61 (57-65) 43 (37-50)
74 (68-80) 38 (32-44) 27 (20-34)
15
Cavity type
3 years
93 (84-100) 95 (90-100) 95 (89-100) 86 (77-95) 86 (78-94) 86 (78-94) 93 (86-100) 44
n
Cumulative survival rate (%)
* = Viohl's two-digit system. For the sake of clarity, the tooth numbers are only those from the right side.
Fracture ef amaigam-restered endodentically treated teeth
survival analyses but included in the analyses of the fracture pattern. Table 1 shows the 1584 teeth in the survival analyses distributed by cavity type and fracture mode. In this article, no distinction is made between teeth from the right side of the mouth and teeth from the left side. In the tables, only right side tooth numbers will be used.
Survivai rates - retention of both cusps
Table 2 gives the cumulative survival rates of the 16 different combinations of tooth and cavity type ranked after decreasing 20-year survival rate. The endodontically treated teeth could be divided into 3 groups (A, B and G) with different survival rates: the lowest survival rate was that of the 2 upper premolars with MOD cavities (Table 2, Group G); between 67% and 77% survived the first 3 years and the 20-year survival rate was only 20-34% (95% confidence intervals). The highest survival rate was seen for teeth with MO/DO cavities, as 87-93% survived after 3 years and 68-80% after 20 years (Table 2, Group A). The only exception was the upper second premolar. This tooth with an MO/DO cavity had the same low survival rate as lower MOD premolars and upper and lower MOD molars (Table 2, Group B). The difference between Groups B and G was
Statistically significant after 1 year and between Group A and B after 2 years (P< 0.005). The difference between the cumulative survival rate of Group A and those of the 2 other groups increased very rapidly; already after 5 years, this difference was statistically significant at P < 0.001. The cumulative survival rates of the 3 groups are shown in Fig. 1. For the sake of clarity, the 95% confidence limits are omitted in Fig. 1, but are presented in the lower part of Table 2 for the 3-, 10-, and 20-year survival rates. For each of the 8 posterior teeth, a statistical analysis was made of the difference between the 20year survival rate of teeth with an MO/DO cavity and that of the corresponding teeth with an MOD cavity. The analyses showed a markedly better survival rate for MO/DO cavities than for MOD cavities. The most pronounced difference was found for the upper first premolar (/*< 0.001) and the least pronounced one for the lower first premolar (P = 0.035).
Fracture pattern - tooth surface
The following analyses are based upon the fracture modes given in Table 1. The endodontically treated teeth mostly fractured lingually (63%); the frequencies for the other 3 fracture modes were 25% facial, 8% total and 4% vertical failures. But there were 2 significant exceptions: the upper first premolar with MOD cavity, in which nearly half the fractures were found facially, and the upper second molar, which had a very high frequency of vertical fracture. The latter tooth accounted for less than 8% of all the endodontically treated teeth in this study, but 43% of the extractions caused by vertical fracture. The difference between the frequency of vertical fracture for the upper second molar and that of the other posterior teeth was statistically significant (/*< 0.001). Only 4 vertical fractures were found in the lower jaw and solely in MOD cavities.
Table 3. The 99% confidence limits for supra- and subgingival fractures in relation to fracture mode Fracture level 8 12 16 YEARS AFTER ENDODOhfTIC TREATMENT
20
Pig- 1- Cumulative survival rates of endodontically treated posterior teeth restored with amalgam. A. Teeth with MO/DO cavity except upper second premolar. B. Lower premolars with MOD cavity, upper and lower molars with MOD cavity, and upper second premolar with MO/DO cavity. C. Upper premolars with MOD cavity. See Table 2 for 95% confidence limits.
Fracture mode Facial Lingual Total crown
Subgingival
n
Supragingival
Supracrestal
Subcrestal*
135 334 41
47-69% 23-36% 45-83%
23-44% 42-57% 11-48%
4-17% 16-28% 1-24%
* This column does not include vertical fractures.
51
Hansen et ai. Fracture pattern - effect of tooth position in jaw
Fracture pattern - iocation of fracture
The fracture level for 60 teeth was not recorded. Because of the consequent uncertainty, in the following analyses the level of significance was set to 1%. Apart from one tooth (upper first molar), there was no statistically significant difference between the periodontal damage caused by fracture of MO/DO teeth and that caused by fracture of the corresponding MOD teeth; in some case the periodontal damages were slightly more severe for MO/DO teeth than for MOD teeth, but in other cases the reverse was found. The only statistically significant difference was found for the upper first molar, as more subgingival fractures were seen with MO/DO cavities than with MOD cavities (/^ = 0.009). However, we could not find any explanation for that. This statistically significant difference for the upper first molar may be a type 1 error, because only 25 MO/DO fractures were recorded (19 fractures in the survival analysis, 4 in the partly rejected group, and 2 with unknown fracture level). If the latter 2 fractures had been supragingival, the level of probabihty would have been 0.04 instead of 0.009. MO/DO and MOD cavities were therefore pooled (Table 3). The difference between the number of fractures in Table 1 (/z = 532) and in Table 3 (;z = 510) can be explained as follows: 6 fractures, including 1 vertical failure, happened 20 years or more after the endodontic therapy; the 55 teeth, excluded from the survival analyses, were included in the analyses of the fracture levels; vertical fractures are not included in Table 3; and 60 fractures had no recorded fracture level (532 + 6 + 55 - 23 - 60 = 510). There were only minor differences between the fracture levels with facial and total failure (P=0.41, two-tailed) (Table 3). The 2 failure modes were therefore pooled. Table 3, which does not include vertical fractures, shows that JinguaJ failures mostly were subgingival, in contrast to facial and total crown failures. Furthermore, subcrestal fractures were more frequent for lingual failures. The level of probability in both cases was below 0.001. This was the general picture. But as seen in the following analyses, the fracture pattern was more complex.
Thejonckheera-Terpstra test (11) showed a statistically significant trend in both the upper and lower jaw: the more posterior the tooth, the more pronounced the periodontal damage caused by tooth failure (P< 0.001). In order to further examine the previously mentioned pooling of MO/DO and MOD cavities, the Jonckheera-Terpstra test was also carried out with the 2 cavity types separately. The trend was still statistically significant (P
Tooth fracture is a frequent problem in restorative dentistry. Both sound and restored teeth may fracture, but especially those endodontically treated are at risk (1-3). The main problem is that the fracture often extends subgingivally and not infrequently results in loss of periodontal attachment. Subsequent operative procedures for impression may cause futher damage to the periodontal tissues. Also, a subgingival restoration margin, no matter how well adapted, nearly always provokes an inflammatory periodontal reaction (4-8). In a previous, retrospective investigation (3), the frequency of cusp and crown fracture of endodontically treated premolars with MOD cavities was studied. The 3-year survival rate was markedly improved when the cavity had been restored with an enamel-bonded resin filling instead of amalgam. But after 3 years, tooth failure of the resin-restored preinolars happened with nearly the same frequency is that found for premolars restored with amalgam, in that study (3), however, the number of resinestored teeth was rather small (n = 40). The purpose of the present retrospective study
Erik Keith Hansen', Eriit Asmussen\ Niels C. Ciiristiansen^ Institutes of ^Dental Materials and Technology and ^Prosthetic Dentistry, Royal Dental College, Copenhagen, Denmark
Key words: amalgam; cavity preparation; endodontic therapy; tooth fracture. Dr. odont. Erik Keith Hansen, Helsingorsgade 7, DK-3400 Hillerod, Denmark. Accepted for publication September 15, 1989.
was to examine the cumulative survival rates (retention of both cusps) of endodontically treated premolars and molars restored with MO/DO or MOD amalgam without cuspal overlays, and to elucidate the fracture pattern of these amalgam-restored teeth. iWaterial and methods The data were collected from January to October 1988 from 91 dentists working as general practitioners. The results from the previous investigation (3) were updated and included in the present study. All data were derived from clinical examinations and review of the patients' dental records and radiographs. The criteria for including data were: (i) an endodontically treated premolar or first or second molar with an MO, a DO, or an MOD amalgam restoration; (ii) no cuspal amalgam overlays; (iii) no Glass V restoration; (iv) normal functional occlusion; (v) complete dentition in the anterior and premolar region, including cases with dental bridges restoring 49
Hansen et ai. one tooth in the premolar region or 2 in the anterior region; and (vi) at least one molar in each quadrant involved in the occlusion. Endodontically treated teeth without antagonist were not included. The dentists were asked to record the following data on a registration form: tooth number, date of endodontic therapy, cavity type (MO, DO or MOD), and the date of control or last contact. In cases of fracture, the dentists were asked to record the date of tooth failure and whether the facial cusp, the lingual cusp, or the whole crown (total fracture) was lost. Finally, the dentists were asked to record whether the fracture was supragingival, subgingival, at or beyond the alveolar crest, or whether the fracture was so vertical that the tooth had to be extracted. In this article, fractures at or beyond the alveolar crest, but not vertical fractures resulting in extraction, are referred to as subcrestal, even though some of these fractures may have had a fracture level slightly above the alveolar crest. The trial time was defined as the time elapsed between the date of endodontic therapy and the date of fracture; the date of final registration; the date of withdrawal (change of cavity, change of restorative material, extraction of the antagonist); or the date of last contact with the patient. In the analyses, no distinction was made between MO and DO restorations: only between MO/DO and MOD. When MO/DO restorations had been replaced by MOD fillings, the teeth were recorded as withdrawn MO/DO teeth and re-entered as new MOD teeth. The total trial time was set to 20 years with intervals of 1 year. For each interval, the effective number of teeth exposed to risk of fracture was calculated. The cumulative survival rate, i.e. retention of both cusps, for each tooth number with an MO/DO or MOD cavity was then calculated by means of life table analysis (9). The classification based upon tooth number and cavity type gave 16 different combi-
nations of tooth and cavity type: 4 premolars and 4 molars, each with 2 cavity types. Differences between these 16 combinations were analyzed with log-rank tests (9) at the 5% level of significance. Analyses of the fracture pattern were done with Kruskal-Wallis one-way analysis of variance, the Mann-Whitney U-test, Fisher's exact probability test (10) and the Jonckeera-Terpstra test (11); the latter analysis can be described as a Kruskal-Wallis test for trend. The significance level for the fracture pattern was set to 1%; the reason for this will be explained in Results. Most of the analyses were carried out with 2 computerized statistical programs (SAS, version 6.03, SAS Institute, Gary, NG, USA; and MEDSTAT, version 2.1, Astra, Gopenhagen, Denmark). The 95% and 99% confidence limits were calculated using MEDSTAT.
Resuits Data were obtained on 1695 endodontically treated teeth with MO/DO or MOD cavities; all teeth were restored with amalgam. However, 56 teeth were rejected because the cavity type or the date of endodontic therapy or control was not recorded or because of cuspal overlay. A further 55 sets of data were partly rejected because some of the dentists misunderstood the instructions and only recorded fractured teeth, not fractured and non-fractured at random. These 55 teeth were excluded from the Table 2. Cumulative survival rates (%); 95% confidence limits in parentheses Tooth number* Cavity type Group
MO/DO
MOD
44 45 17 47 16 14 46
Table 1. Number of endodontically treated teeth (n) in the survival analyses distributed on cavity type and fracture mode. The number of fractures are those found within 20 years after endodontic therapy
Tooth number*
14 15 16 17 44 45 46 47
Fracture mode Facial
Lingual
Total
Vertical
MO/DO MOD
MO MOD
MO MOD
MO MOD
MO MOD
5 4 7 1 3 2 3 3
10 10 11 2 1 5 6 10
79 64 98 64 43 85 68 75
132 203 194 57 25 123 215 59
34 26 12 4 3 12 10 3
23 72 55 10 4 29 72 17
3 5
7 6 2 2 2 8 2 1
1 1 3
1 3 3 7 1 1 1
* = Viohl's two-digit system. For the sake of clarity, the tooth numbers are only those from the right side.
50
A B C
10 years
20 years
89 (77-100) 90 (78-100) 90 (76-100) 81 (67-96) 80 (68-92) 81 (69-92) 80 (62-97)
85 (60-100) 82 (61-100) 74 (46-100) 73(42-100) 71 (46-96) 67 (45-90) 64 (20-100)
45 17 16 46 47
87 (74-100) 83 (72-93) 81 (73-88) 77 (66-89) 87 (82-92) 80 (74-86) 89 (80-98)
67 (46-88) 62 (44-81) 58 (47-69) 54 (35-74) 66 (57-76) 58 (49-68) 59 (41-78)
53 (18-89) 49 (9-88) 47 (30-63) 36 (13-59) 34(18-51) 34(7-61) 31 (2-60)
14 15
75 (67-83) 70 (63-77)
50 (3&-61) 39 (30-48)
28 (12-44) 27 (13-41)
90 (87-93) 83 (80-86) 72 (67-77)
84 (80-88) 61 (57-65) 43 (37-50)
74 (68-80) 38 (32-44) 27 (20-34)
15
Cavity type
3 years
93 (84-100) 95 (90-100) 95 (89-100) 86 (77-95) 86 (78-94) 86 (78-94) 93 (86-100) 44
n
Cumulative survival rate (%)
* = Viohl's two-digit system. For the sake of clarity, the tooth numbers are only those from the right side.
Fracture ef amaigam-restered endodentically treated teeth
survival analyses but included in the analyses of the fracture pattern. Table 1 shows the 1584 teeth in the survival analyses distributed by cavity type and fracture mode. In this article, no distinction is made between teeth from the right side of the mouth and teeth from the left side. In the tables, only right side tooth numbers will be used.
Survivai rates - retention of both cusps
Table 2 gives the cumulative survival rates of the 16 different combinations of tooth and cavity type ranked after decreasing 20-year survival rate. The endodontically treated teeth could be divided into 3 groups (A, B and G) with different survival rates: the lowest survival rate was that of the 2 upper premolars with MOD cavities (Table 2, Group G); between 67% and 77% survived the first 3 years and the 20-year survival rate was only 20-34% (95% confidence intervals). The highest survival rate was seen for teeth with MO/DO cavities, as 87-93% survived after 3 years and 68-80% after 20 years (Table 2, Group A). The only exception was the upper second premolar. This tooth with an MO/DO cavity had the same low survival rate as lower MOD premolars and upper and lower MOD molars (Table 2, Group B). The difference between Groups B and G was
Statistically significant after 1 year and between Group A and B after 2 years (P< 0.005). The difference between the cumulative survival rate of Group A and those of the 2 other groups increased very rapidly; already after 5 years, this difference was statistically significant at P < 0.001. The cumulative survival rates of the 3 groups are shown in Fig. 1. For the sake of clarity, the 95% confidence limits are omitted in Fig. 1, but are presented in the lower part of Table 2 for the 3-, 10-, and 20-year survival rates. For each of the 8 posterior teeth, a statistical analysis was made of the difference between the 20year survival rate of teeth with an MO/DO cavity and that of the corresponding teeth with an MOD cavity. The analyses showed a markedly better survival rate for MO/DO cavities than for MOD cavities. The most pronounced difference was found for the upper first premolar (/*< 0.001) and the least pronounced one for the lower first premolar (P = 0.035).
Fracture pattern - tooth surface
The following analyses are based upon the fracture modes given in Table 1. The endodontically treated teeth mostly fractured lingually (63%); the frequencies for the other 3 fracture modes were 25% facial, 8% total and 4% vertical failures. But there were 2 significant exceptions: the upper first premolar with MOD cavity, in which nearly half the fractures were found facially, and the upper second molar, which had a very high frequency of vertical fracture. The latter tooth accounted for less than 8% of all the endodontically treated teeth in this study, but 43% of the extractions caused by vertical fracture. The difference between the frequency of vertical fracture for the upper second molar and that of the other posterior teeth was statistically significant (/*< 0.001). Only 4 vertical fractures were found in the lower jaw and solely in MOD cavities.
Table 3. The 99% confidence limits for supra- and subgingival fractures in relation to fracture mode Fracture level 8 12 16 YEARS AFTER ENDODOhfTIC TREATMENT
20
Pig- 1- Cumulative survival rates of endodontically treated posterior teeth restored with amalgam. A. Teeth with MO/DO cavity except upper second premolar. B. Lower premolars with MOD cavity, upper and lower molars with MOD cavity, and upper second premolar with MO/DO cavity. C. Upper premolars with MOD cavity. See Table 2 for 95% confidence limits.
Fracture mode Facial Lingual Total crown
Subgingival
n
Supragingival
Supracrestal
Subcrestal*
135 334 41
47-69% 23-36% 45-83%
23-44% 42-57% 11-48%
4-17% 16-28% 1-24%
* This column does not include vertical fractures.
51
Hansen et ai. Fracture pattern - effect of tooth position in jaw
Fracture pattern - iocation of fracture
The fracture level for 60 teeth was not recorded. Because of the consequent uncertainty, in the following analyses the level of significance was set to 1%. Apart from one tooth (upper first molar), there was no statistically significant difference between the periodontal damage caused by fracture of MO/DO teeth and that caused by fracture of the corresponding MOD teeth; in some case the periodontal damages were slightly more severe for MO/DO teeth than for MOD teeth, but in other cases the reverse was found. The only statistically significant difference was found for the upper first molar, as more subgingival fractures were seen with MO/DO cavities than with MOD cavities (/^ = 0.009). However, we could not find any explanation for that. This statistically significant difference for the upper first molar may be a type 1 error, because only 25 MO/DO fractures were recorded (19 fractures in the survival analysis, 4 in the partly rejected group, and 2 with unknown fracture level). If the latter 2 fractures had been supragingival, the level of probabihty would have been 0.04 instead of 0.009. MO/DO and MOD cavities were therefore pooled (Table 3). The difference between the number of fractures in Table 1 (/z = 532) and in Table 3 (;z = 510) can be explained as follows: 6 fractures, including 1 vertical failure, happened 20 years or more after the endodontic therapy; the 55 teeth, excluded from the survival analyses, were included in the analyses of the fracture levels; vertical fractures are not included in Table 3; and 60 fractures had no recorded fracture level (532 + 6 + 55 - 23 - 60 = 510). There were only minor differences between the fracture levels with facial and total failure (P=0.41, two-tailed) (Table 3). The 2 failure modes were therefore pooled. Table 3, which does not include vertical fractures, shows that JinguaJ failures mostly were subgingival, in contrast to facial and total crown failures. Furthermore, subcrestal fractures were more frequent for lingual failures. The level of probability in both cases was below 0.001. This was the general picture. But as seen in the following analyses, the fracture pattern was more complex.
Thejonckheera-Terpstra test (11) showed a statistically significant trend in both the upper and lower jaw: the more posterior the tooth, the more pronounced the periodontal damage caused by tooth failure (P< 0.001). In order to further examine the previously mentioned pooling of MO/DO and MOD cavities, the Jonckheera-Terpstra test was also carried out with the 2 cavity types separately. The trend was still statistically significant (P
