G. S. P. Cheung & T. K. Chan
Faculty of Dentistry, The University of Hong Kong, Prince Philip Dental Hospital, Hong Kong.

The literature is replete with studies on the outcome of nonsurgical root canal treatment, which has been reported to have a success rate of 53-95% over various periods of observation (Friedman 1998). These success rates were calculated as a percentage of those teeth examined at recall where it was assumed that the cases survived up to the longest follow-up, although the mean observation period was usually shorter. A number of factors have been evaluated using simple statistical tests for their association with the percentage of successful cases. For instance, the presence of a pre-existing periapical radiolucent lesion is generally agreed to result in a lower rate of success (Sjogren et al. 1990, Smith et al. 1993), whereas vital pulpectomy appeared to enjoy a higher success rate compared with root canal treatment of infected pulps (Heling & Tamse1970, Smith et al.1993). These conclusions were based on the calculation of binomial proportions of successful cases, and had not taken into account any confounding of factors being evaluated.

The presence of microorganisms has been implicated to be a major cause of failed root canal treatment (Bystrom et al.1987, Molander et al.1998, Sundqvist et al. 1998, Cheung &Ho 2001). Whereas early failures usually reflect the inadequacy of the treatment in controlling the root canal infection, the late failures may reflect the influence of other factors such as coronal microbial ingress over time (Saunders & Saunders 1994, Drstavik 1996, Hayes & Dummer1997). Although the presence of cultivable bacteria in root canals prior to filling appears to result in a lower success rate (Engstrom et al. 1 964, Sjogren et al. 1997), bacteriological sampling of root canals is not without limitations (Dahle. n & Haapasalo 1998), and hence its use has not been required as routine procedure prior to completion of nonsurgical treatment (Molander et al.1996).To assess the outcome of endodontic treatment, long-term radiographic review is essential to supplement clinical findings. It has been suggested that root canal treatment should be followed for up to 4 years (European Society of Endodontology 1994). Unfortunately, not many studies have incorporated a review period of more than 4 years (reviewed by Friedman 1998). As the rate at which failures occur may not be constant over a certain period of observation that itself may also be a variable for each subject of the study, a different approach to evaluate treatment outcome is necessary. A life-table or survival analysis, which examines the occurrence of critical events such as development of failure throughout the observation period, is preferred (Mitchell & Walls 1991, Weiger et al.1998). Such analysis allows the assessment of the survival function over time and the prediction of longevity of cases related to treatment modality, as well as the effects of any co variables (Anderson & Zarb 2000).
The purpose of this study was to evaluate the long-term survival of primary root canal treatment performed in a dental teaching hospital more than 10 years ago, and to identify factors that might affect the survival.

From the computerized dental records of the Prince Philip Dental Hospital (PPDH), Hong Kong,8156 permanent teeth had received primary root canal treatment betweenJanuary1981and the end of1989.All treatments were carried out according to the then prevailing guidelines by undergraduate and postgraduate students, house officers and faculty members at PPDH. In short, this included rubberdam isolation, a stepback technique with physiological saline as irrigant and cold lateral compaction of gutta-percha. In order to obtain a similar sample size for each tooth in subsequent analyses, a stratified random sampling method was adopted. Using computer software (SPSS 10.0; SPSS Inc., Chicago, IL, USA), a randomcollectionof40caseswas drawn for each tooth number, except for mandibular incisors which were pooled as a single group for sampling because of the small number treated, and because they tend to possess similar root canal anatomy and clinical accessibility. The third molars were excluded. The initial sample consisted of1000teeth in 936 patients, which was the same series as examined by Chan (2001). Fourteen teeth were excluded as having had prior treatment outside PPDH, leaving 986 teeth in the sample.
The written records and radiographs of the selected patients were studied to assess the status of treatment at the last review. For the purposes of this study, any of the following was deemed to be a failure of the primary root canal treatment:

1. Extraction of the tooth, except for established reasons not related to the root canal treatment, as documented in the patient’s record in which case the case would be excluded.
2. Retreatment of the root canal of the tooth either surgically or nonsurgically.
3. Presence of a periapical radiolucent area continuous with the periodontal ligament space, regardless of its size, more than 4 years after obturation.

Teeth meeting the criteria (1) and (2) above were regarded as a definite failure of the primary treatment. In order to determine the time when failure occurred, two time-points for each tooth were established: the date when failure was first diagnosed or ‘first-known-bad’ (Df) and the date of the last recall when the treatment was not considered failed or ‘last-known-good’ (Dg). For time intervals between the date of obturation and recall less than 4 years, incomplete resolution of periapical radiolucency (the third criterion above) was not regarded as failure (European Society of Endodontology 1994).Thus, an interval of 4 years or more was required before Df was regarded as valid for cases failed because of this criterion. All patients were then identified, contacted by telephone or in writing and offered a review appointment.
At the review, patients were examined by an endodontic postgraduate student (TKC). The presence of any clinical signs or symptoms associated with the selected teeth was considered to arise from a failed treatment. The type and integrity of the coronal restoration were also noted. Paralleling-technique periapical radiographs (Ektaspeed Plus, Eastman Kodak, Rochester, NY, USA) were taken with the aid of a positioning device (XCP, Rinn Corporation, Elgin, IL, USA) and developed in an automatic processor (Velopex Intra-XE, Medivance Instruments, London, UK). Using written criteria based on the above, two endodontic postgraduates assessed the radiographs independently over a light-box using a magnifying viewing device (X-produkter, Malmo, Sweden) that excluded extraneous light. Any disagreement was solved by discussion between the two (Halse & Molven 1986). Should any criteria for failure be met at this stage, the date of recall was taken as Df. Otherwise, this date became the Dg. Arrangements were made for patients who attended the recall and for whom further treatment was considered necessary as a result of the clinical examination.
A number of patients failed to attend the recall and thus a postoperative status of more than 4 years was not available for 378 teeth (Table 1). For those who did not attend but the selected teeth had been reviewed on at least one occasion more than 4 years after the primary treatment (n ј128), the radiographic status and date of the longest review were used as these provided some information suitable for survival analysis (Mitchell &Walls1 991).

Table 1. Distribution of teeth that did not have follow-up information available.

Statistical analysis.
Root canal failures usually develop over time and normally would not be noticed until a recall appointment or acute symptoms or a catastrophe event such as tooth fracture. However, the actual date of failure must lie between Dg and Df. The review appointments served to update the status of treated teeth, confirming Dg and establishing Df, or resetting Dg to this new date of examination. Then, taking the date of obturation (D0) as the origin for time measurement, the intervals that the root canal filling had been present to the respective dates were:

Tg = Dg - D0          (1)
Tf = Df - D0          (2)

where Tg and Tf are expressed in months. As survival analysis using the Kaplan-Meier (K-M) estimator requires the entry of a known event date or ‘time to failure’ (Parmar & Machin1995), some means to determine this interval from the values of Tg and Tf was necessary. Whilst the ‘failure’ of many medical conditions or treatment modalities is clearly defined, such clear dates were not readily identified in the present context. It thus became necessary to insert an estimated date of failure for which a probabilistic approach was employed to determine the average relative time of failure between the two index times, Tg and Tf. In a previous study, the halfway point between Tg and Tf was used as crude estimate, but the author pointed out that an appropriate method of making such estimation should be devised (Cheung 2002). Preliminary consideration was made in the present analysis by taking various linear estimates between Tg and Tf for the calculation of the K-M cumulative survival probability, but the resultant plots appeared rather ‘bumpy’ indicating an erratic data trend that would be too complicated to interpret and explain (Fig. 1). Given that for a large number of cases, a smooth low-degree polynomial curve would be expected; these plots calculated from linear time estimates (Fig. 1) were not considered satisfactory. As the underlying data essentially represented a ‘rate of change’ of periapical status, a geometric mean, Tk, was used (Meyer1975) to provide a probabilistic estimate of the ‘time to failure’ for subsequent analysis.

Tk = [(Tg + 1) x (Tf + 1)1/2          (3)

where the offset value of ‘1’ was inserted to bypass the ambiguous condition for Tg = 0 whilst maintaining the absolute interval between Dg and Df (Chan 2001).
The contribution of10covariables (Table 2) to the survival function was further analyzed using the Cox Regression model to examine the possibility of confounding or interdependence of these factors. The selection of significant variables was based on a backward stepwise method (SPSS1999). In determining the effect associated with the medicament used, it was recognized that cases treated in a single visit would automatically preclude the use of a medicament. Thus, the analysis was rerun with all single-visit cases removed specifically to examine the effect of this factor.

Figure 1. Cumulative survival probability and standard error plotted using different linear estimates as the date of failure.



Table 2. Mean survival time of root canal treatment by categorical covariables and the result of Cox Regression analysis.

Of the 986 teeth, 128 were not examined but included in the data set, for they had been reviewed on at least one occasion more than 4 years postoperatively; 282 teeth were examined at the recall and 198 teeth had been documented to have failed from the records. Thus, the data set contained a total of 608 teeth. The median survival time of these treated teeth was 111months (Fig. 2). On closer examination of the overall survival curve (Fig. 2), an initial fast drop, which then slowed down after approximately18months, could be identified.
Of the teeth examined, none was associated with any clinical symptoms. One hundred and sixteen teeth were deemed to have failed, for a total number of failure of 314 teeth (52%) (Table 3). The majority of the failures occurred within the first 3 years after obturation, with a mean ‘life span’ of 39 months (SD = 39), median 26.5 months (Fig. 3). Teeth with no preoperative periapical lesion had a significantly better survival outcome after root canal treatment compared to those with a periapical lesion or widened periodontal ligament (Table 2; Fig. 4). Maxillary and mandibular molars and maxillary premolars had a significantly lower survival probability than maxillary and mandibular anteriors and mandibular premolars (Table 2; Fig. 5). Teeth restored with crowns survived significantly longer than those with intracoronal plastic restorations only (Table 2; Fig. 6). The use of an all-amalgam postcore or prefabricated post was associated with a significantly higher survival probability than no post or using a cast post (Table 2; Fig.7). Although the medicament used was found to be significant in the initial analysis (P = 0.047), it became an insignificant factor when all single-visit cases were excluded (Table 2).The preoperative periapical status, tooth type, type of post and the final restoration remained significant.

Figure 2. Overall cumulative survival probability, with 95% confidence interval, for all teeth included in the analysis (N = 608).



Table 3. Observed or documented reasons for failures up to the documented date of failure or the recall.



Figure 3. Incidence of failures (n = 314) with time.



Figure 4. Survival curve as a function of the preoperative periapical status (result after Cox Regression analysis and plotted at mean of other covariables).



Figure 5. Survival curve as a function of the tooth type (result after Cox Regression analysis and plotted at mean of other covariables).



Figure 6. Survival curve as a function of the type of coronal restoration (result after Cox Regression analysis and plotted at mean of other covariables).



Figure 7. Survival curve as a function of the type of post used (result after Cox Regression analysis and plotted at mean of other covariables).

Discussion.
Long-term clinical and radiographic follow-up is essential for the assessment of endodontic treatment outcome. Whilst the process of periapical healing may often be diagnosed radiographically at the 1-year recall, individual cases may take up to 4 years for complete healing of periapical lesions (Drstavik 1996). Before this time, teeth with an incompletely resolved area were neither regarded as a failure nor success, because the data were not considered valid. Given a minimum postoperative period of 4 years, the presence of a radiolucent area of any size was regarded as a failure, with the exception of a locally widened periodontal ligament space (European Society of Endodontology 1994). These criteria for failure were rather stringent, indeed, much stricter than those used in many studies of similar or shorter periods of observation (reviewed by Friedman 1998). This may account for the relatively high percentage of cases (52%) that were deemed to have failed in this study.
The 378 teeth that were excluded from entry into the data set were distributed at a similar proportion amongst various tooth type as those being analyzed (Table 1), and thus were unlikely to affect the results here. Cases of extraction due to nonendodontic reasons were excluded for convenience in the analysis. On reviewing the data, majority of these ‘nonendodontic extractions’ were carried out 2-4 years postoperatively. Whilst one might include these cases as ‘censored’ at the time of extraction in the K-M analysis, their inclusion would merely ‘refine’ the initial part of the survival curve. The survival function beyond this early period was unlikely to be affected to any significant degree by the omission of these cases.
The different rates of decline in survival of the treated teeth before and after a postoperative period of about 18months suggest the presence of two overall processes leading to failure of the treatment, one rapid and one slow. Examination of the life span of individual failed cases confirmed this proposition; some 44% of all failures occurred in the first 18months (Fig. 3). Thereafter, the incidence of failures declined with time. Further investigation is justified to examine the factors contributing to failures in these different periods.
The median survival time of the root canal treatment was111months, which was remarkably similar to a previous but independent study in the same hospital that examined teeth treated between1990 and1995 (Cheung 2002), but the method used to estimate the ‘time to failure’ was somewhat different. The Cox Regression analysis was used in preference to the Log Rank Test, as the former is able to identify any confounding of covariables being examined (Lee & Go1997). Four factors were identified to be significant in affecting the survival function: preoperative periapical status, tooth type, the type of coronal restoration and the type of post used. We had not examined the influence of preoperative pulpal condition on the longevity of treatment because such information sometimes was missing in the patients’ records nor was it included as part of the computerized dental database at that time.
The cumulative probability of survival was significantly higher when there was no preoperative periapical lesion; this corroborates many previous findings based on the calculated percentage of successful cases (Heling & Tamse1970, Sjogren et al.1990, Smith et al.1993).However, there seemed to be a lack of effect for small versus large periapical lesions on the survival outcome in this study, despite reports suggesting a lower success rate with lesions larger than 5 mm in diameter (Storms1969, Matsumoto et al.1987). These two studies, however, had a mean review period of1-2 years only. Giventhat a long enough interval may be required for complete healing of large lesions (Drstavik 1996), it was possible that premature evaluation of periapical status might have produced an inaccurate result (Friedman1998). A recent study also indicated that the extent of preoperative lesion had the largest effect on treatment outcome (Chugal et al. 2001). However, the same observation was not apparent in the present study, nor supported in a previous report (Sjogren et al. 1990). Cases with widened periodontal ligament seemed to behave similarly as those with periapical radiolucent areas. As pathological lesions could exist in the periapical region, which might not be discernible on the periapical radiograph (Brynolf 1967), it was likely that periapical inflammation was present when the periodontal ligament space was seen to be widened radiographically.
Our results indicated that tooth type was a significant factor affecting the endodontic treatment outcome. It can be seen that the survival curves associated with different tooth types were in two clusters. Mandibular premolars and maxillary and mandibular anterior teeth were in one group which had a significantly longer survival time than other tooth types. Molar teeth demonstrated the worst survival outcome. From a clinical point of view, molars are more difficult to treat because of accessibility and complex root canal anatomy. Root- filled posterior teeth might fracture more readily if they had not been restored in such a way that crown-root fracture may be prevented (Hansen & Asmussen 1990, Gutmann 1992). It is also arguable that the criteria for success might have discriminated against molars because a multirooted tooth here was regarded as failed if any one of its roots had met the criteria for failure (Friedman1998).Mandibular premolars usually possess relatively simple canal morphology in the Chinese population (Walker 1987), which together with better accessibility could imply easier treatment and thus a greater chance of attaining a high ‘technical’ standard of treatment. Maxillary anterior teeth share a similar survival time probably because of the same reason.
Many teeth surveyed in this study did not receive any inter appointment medicament during treatment, which was a common practice at one time in the 1980s. When calcium hydroxide was used as the root canal medicament, a seemingly higher survival probability of root-filled teeth was attained, compared with other medicaments or no dressing when all cases were included in the analysis. Calcium hydroxide has been documented to be an effective disinfectant for the root canal system (Bystrom et al. 1985, Sjogren et al. 1991). However, knowing that no medicament would have been used in cases of single-visit treatment, the multivariate test was rerun with these cases removed - the medicament then became an insignificant covariable (Cox Regression, P = 0.102). The lack of significance of the medicament used in multiple-visit therapy might be related to its placement method. As calcium hydroxide paste was delivered into the canal with a K-file rotated in reverse by hand during those times at PPDH, the quality of ‘fill’ was probably not very dense (Sigurdsson et al. 1 992) which could affect its antimicrobial effectiveness. Other factors, e.g. the integrity of the temporary restoration that had not been studied because such information was often not entered in the patient’s record, might have a bearing on the effectiveness of calcium hydroxide.
The survival of root-filled teeth was enhanced when a crown restoration had been provided. On the other hand, studies that had attempted to examine the relationship between the type of coronal restoration and the endodontic treatment outcome failed to demonstrate a statistically significant association (Safavi et al. 1 987, Friedman et al. 1995). The difference may be related to the criteria for failure used in the present study. As all extractions, except for documented reasons not related to the root canal treatment, were regarded as failures of the primary treatment here, this would include cases of nontraumatic tooth fracture that had been extracted over the postoperative period. It has been reported that teeth with no subsequent restoration or restored only with amalgam or resin composite were significantly more likely to undergo extraction than teeth that had received a crown (Lazarski et al. 2001). Thus, not only does a crown restoration provide a coronal seal to prevent bacterial recontamination of the root canal system (Saunders & Saunders 1994), but it may also serve to protect the tooth from the risk of fracture.
The placement of a cast post and core had an adverse effect on the survival function of the root-filled teeth. The theoretical disruption of the apical seal (of the cold laterally compacted gutta-percha root canal filling) was unlikely to be the cause because teeth with prefabricated posts were not so affected. Most cast posts constructed in the 1980s were tapered, and the teaching then did not require the placement of crown margin at a distance beyond the tooth-core junction to provide a ‘ferrule effect’. Without this design, the post-crown assembly could fail at a much lower load and with an unpredictable pattern (Sorensen & Engelman 1990). Such nontraumatic tooth fractures had been considered as a failure because the ultimate goal of treatment, i.e. to maintain the tooth as a functional unit of the dentition, was not achieved.
The use of an integrated amalgam postcore, or coronal- radicular amalgam core (Nayyar et al. 1 980) was associated with a favourable survival outcome, possibly because a crown was of ten placed as the final restoration for these teeth and thus some form of protection was present. The prefabricated posts used were almost exclusively parallel-sided and were placed mostly in posterior teeth (anterior : posterior @ 1: 10). About 80% of them were subsequently restored with a crown. It appears that the favourable survival outcome for these two types of posts is related to the final restoration - teeth with crown restorations survived significantly longer than those with plastic restorations only.
The great majority of the 224 teeth that did not receive any form of posts here were restored with a plastic intracoronal restoration with or without bonding, and they had a significantly lower survival probability than the groups with amalgam postcores and prefabricated posts. Again, it can be noticed that the survival pattern of the different postcore groups mirrored that of the restoration type. It seems that even with the multivariate analysis used in this study, the confounding of the two variables, type of restoration and type of post could not be separated. Clinically, these two variables are closely related to each other because a crown restoration is likely to be planned with the decision for a post and core. The relatively poor survival outcome for teeth with a cast post may be related to a lack of ‘ferrule’ protection (due to the prevailing practice then), and iatrogenesis during posthole preparation.
To our surprise, the apical extent and the ‘quality’ of the root canal filling were not significant covariables affecting the survival function. The assessment of quality here was determined from radiographs alone and was scored on basis of the homogeneity of its appearance. Such evaluation has major limitations, as any buccal or lingual voids to the root filling could have been masked and hence not detected. The same may also be masked by root canal sealers with high radiopacity. The presence of voids at the apical third of the canal has been reported to associate with a lower survival probability in another survey at the same hospital (Cheung 2002). A similar result, however, is not reproduced in the present study, although there was a tendency for a longer survival for cases that did not show any void radiographically. The difference may be related to the substantial number of cases that had received no medicament in the present study. These ‘empty canals’ may have harboured residual bacteria which survived the root canal preparation process (Bystrom& Sundqvist1981), because hypochlorite irrigant was not used at that time. The effect of the location of voids could have been masked by the general presence of microorganisms. Thus, a knowledge of the prevailing treatment technique is essential for studies on treatment outcome.
The apical extent of root fillings was found to be difficult to determine at times, especially in maxillary posterior teeth which apices tended to overlap with the image of anatomical structures on the radiograph. It has been argued that the apical extent of root canal fillings bears little, if any, correlation to endodontic failure because the main determinant appears to be the presence of microorganisms (Lin et al. 1991). The lack of influence by the apical extent of the root canal filling was also reported in a study where treatment was carried out under controlled asepsis (Bystrom et al. 1987). Once microorganisms are effectively removed from the root canal system, periapical healing can usually be expected. In contrary, the apical level of obturation was shown to have a significant influence on the outcome of treatment for roots with necrotic pulps and periapical lesions in a hospital environment where treatment was carried out by dental students (Sjogren et al. 1990). Unfortunately, the preoperative pulpal status of the treated teeth was not available in our database, and thus a direct comparison with the corresponding group of teeth is not possible.
The influence of the number of treatment visits on the survival time was found to be insignificant after the Cox Regression analysis. The teaching at PPDH has always been biased towards multiple-visit therapy. The small number of cases in the single-visit group may have diminished the power of the statistical tool to substantiate a difference. In theory, once effective chemomechanical preparation is carried out and a high level of root canal disinfection is achieved with the use of calcium hydroxide, there seems little justification to extend the treatment to a third or more visit. Additional appointments can pose a risk for coronal leakage and bacterial recontamination of the root canal system due to fracture, dislodgement or disintegration of the temporary restoration.

References.

Anderson JD, Zarb GA (2000) Evidence-based dentistry: prognosis. Journal of Prosthetic Dentistry 83,495-500.
Brynolf I (1967) A histological and noentgenological study of the periapical region of human upper incisors. Odontology R evy 18 (Suppl.11),1-176.
Bystrom A, Claeson R, Sundqvist G (1985) The antibacterial effect of camphorated para-monochlorophenol, camphorated phenol and calcium hydroxide in the treatment of infected root canals. Endodontics and Dental Traumatology1,170-5.
Bystrom A, Happonen RP, Sjogren U, Sundqvist G (1987) Healing of periapical lesion of pulpless teeth after endodontic treatment with controlled asepsis. Endodontics and Dental Traumatology 3,58-63.
Bystrom A, Sundqvist G (1981) Bacteriological evaluation of the efficacy of mechanical root canal instrumentation in endodontic therapy. Scandinavian Journal of Dental Research 89, 321-8.
Chan TK (2001) Clinical and radiographic evaluation of root canal treated teeth10-20 years after treatment. MDS Thesis. Hong Kong SAR, China: The University of Hong Kong.
Cheung GSP (2002) Survival of first-time non-surgical root canal treatment performed in a dental teaching hospital. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology and Endodontics 93,596-604.
Cheung GSP, Ho MWM (2001) Microbial flora of root canal treated teeth associated with asymptomatic periapical radiolucent lesions. Oral Microbiology and Immunology16,332-7.
Chugal NM, Clive JM, SpSngberg L (2001) A prognostic model for assessment of the outcome of endodontic treatment: effect of biologic and diagnostic variables. Oral Surgery, OralMedicine, Oral Pathology, Oral Radiology and Endodontics 91,342-52.
Dahle. n G, Haapasalo M (1998) Microbiology of apical periodontitis. In: Drstavik D, Pitt-Ford TR, eds. Essential Endodontology. Malden, MA: Blackwell Science,106-30.
Engstrom B, Segerstad LH, Ramstrom G, Forstell G (1964) Correlation of positive cultures with the prognosis for root canal treatment. Odontology R evy15, 257-70.
European Society of Endodontology (1994) Consensus report of the European Society of Endodontology on quality guidelines for endodontic treatment. International Endodontic Journal 27,115-24.
Friedman S (1998) Treatment outcome and prognosis of endodontic therapy. In: Drstavik D, Pitt-Ford TR, eds. Essential Endodontology. Malden, MA: Blackwell   Science,367-401.
Friedman S, Lost C, Zarrabian M, Trope M (1995) Evaluation of success and   failure after endodontic therapy using a glass ionomer cement sealer. Journal   of Endodontics 21,384-90.
Gutmann JL (1992) The dentin-root complex: anatomic and biologic considerations   in restoring endodontically treated teeth. Journal of Prosthetic Dentistry   67, 458-67.
Halse A, Molven O (1986) A strategy for the diagnosis of periapical pathosis. Journal of Endodontics12,534-8.
Hansen EK, Asmussen E (1990) In vivo fractures of endodontically treated   posterior teeth restored with enamel-bonded resin. Endodontics and Dental   Traumatology 6, 21 8-25.
Hayes SJ, Dummer PM (1997) Late failure of root canal therapy: a diagnostic   and treatment planning challenge. Case report. International Endodontic   Journal 30,68-71.
Heling B,Tamse A (1970) Evaluation of success of endodontically treated teeth. Oral Surgery, Oral Medicine, Oral Pathology 30, 533-6.
Lazarski MP, Walker WA, Flores CM, Schindler WG, Hargreaves KM (2001) Epidemiological evaluation of the outcomes of nonsurgical root canal treatment in a large cohort of insured dental patients. Journal of Endodontics 27,791-6.
Lee ET, Go OT (1997) Survival analysis in public health research. Annual Review in Public Health18,105-34.
Lin LM, Pascon EA, Skribner J, Gangler P, Langeland K (1991) Clinical, radiographic, and histologic study of endodontic treatment failures. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology and Endodontics 11,603-1 1.
Matsumoto T, Nagai T, Ida K, Ito M, Kawai Y, Horiba Net al. (1987) Factors affecting successful prognosis of root canal treatment. Journal of Endodontics 13, 239-42.
Meyer LS (1975) Data Analysis for Scientists and Engineers. New York, NY: John Wiley & Sons, 29.
Mitchell L, Walls AWG (1991) Survival analysis in practice. Dental Update   18,125-8.
Molander A, Reit C, Dahle. n G (1996) Microbiological root canal sampling:   diffusion of a technology. International Endodontic Journal 29,163-7.
Molander A, Reit C, Dahle. n G, Kvist T (1998) Microbiological status of   root-filled teeth with apical periodontitis. International Endodontic Journal   31,1-7.
Nayyar A,Walton RE, Leonard LA (1980) An amalgam coronal radicular dowel   and core technique for endodontically treated posterior teeth. Journal of   Prosthetic Dentistry 43,51 1-5.
Drstavik D (1996) Time-course and risk analyses of the development and healing   of chronic apical periodontitis in man. International Endodontic Journal   29,150-5.
Parmar MKB, Machin D (1995) Survival Analysis: a Practical Approach. Chichester, UK: JohnWiley & Sons, 21-51.
Safavi KE, DowdenWE, Langeland K (1987) Influence of delayed coronal permanent   restoration on endodontic prognosis. Endodontics and Dental Traumatology   3,187-91.
Saunders WP, Saunders EM(1994) Coronal leakage as a cause of failure inroot   canal therapy: a review. Endodontics and Dental Traumatology10,105-8.
Sigurdsson A, Stancill R, Madison S (1992) Intracanal placement of Ca(OH)2:   a comparison of techniques. Journal of Endodontics 18,367-70.
Sjogren U, Fidgor D, Persson S, Sundqvist G (1997) Influence of infection   at the time of root filling on the outcome of endodontic treatment of teeth   with apical periodontitis. International Endodontic Journal 30, 297-306.
Sjogren U, Figdor D, SpSngberg L, Sundqvist G (1991) The antimicrobical effect   of calciumhydroxide as a short term intracanal dressing. International Endodontic   Journal 24,119-25.
Sjogren U, Hagglund B, Sundqvist G, Wing K (1990) Factors affecting the long-term   result of endodontic treatment. Journal of Endodontics16,498-504.
Smith CS, Setchell DJ, Harty FJ (1993) Factors influencing the success of conventional root canal therapy - a 5-year retrospective study. International Endodontic Journal 26,321-33.
Sorensen JA, Engelman MJ (1990) Ferrule design and fracture resistance of endodontically treated teeth. Journal of Prosthetic Dentistry 63,529-35.
SPSS Inc. (1999) SPSS Advanced Models TM10.0. Chapters 10 and 21. Chicago, IL: SPSS Inc.,83-9 (283-311).
Storms JL (1969) Factors influencing the success of endodontic treatment. Journal of Canadian Dental Association 35,83-97.
Sundqvist G, Figdor D, Persson S, SjogrenU (1998) Microbiologic analysis of teeth with failed endodontic treatment and the outcome of conservative retreatment. Oral Surgery, Oral Medicine, Oral Pathology, Oral Radiology and Endodontics 85,86-93.
Walker RT (1987) A comparative investigate of the root number and canal anatomy of permanent teeth in a southern Chinese population. PhD Thesis. Hong Kong SAR, China: The University of Hong Kong.
Weiger R, Axmann-Krcmar D, Lost C (1998) Prognosis of conventional root canal treatment reconsidered. Endodontics and Dental Traumatology 14,1-9.