|Home | About | Journals | Submit | Contact Us | Français|
Our purpose was to quantify the incidence of root canal treatment (RCT) failure and identify its predictors in root canals done in or referred from general dentistry practices in a practice-based research network (PBRN). A retrospective cohort study of 174 endodontically-treated teeth was conducted. Mean duration of follow-up was 8.6 years. Permanent restorations were ultimately placed in 89% of teeth; mean (S.D.) number of days to permanent restoration was 215.4 (609.1). Although RCT had been completed, 18% of teeth were ultimately extracted anyway. Receipt of a permanent restoration was a significant predictor of treatment failure, whether it was determined clinically or radiographically.
This study of PBRN practices suggests a higher failure rate than reported from studies in highly-controlled environments or populations with high levels of dental insurance. Also, the probability of receipt of a permanent restoration is not optimal -- and strongly predicts RCT failure. Appropriately, no RCT was done on teeth with severe periodontal bone loss.
Estimates of the success of non-surgical root canal treatment (RCT) have varied widely. Many studies have reported very high success rates, and endodontic textbooks also typically cite high rates.1-5 However, RCT outcome studies have usually been conducted in academic settings or with patient populations that have ubiquitous or high levels of dental insurance coverage. Results from these settings may not be generalizable to the more-typical settings found in community-based private practice. Therefore, RCT outcome studies done in these community settings are warranted.
In addition to variation in the settings for RCT outcomes research, there has been wide variation in how treatment failure is defined in these studies. Some studies have only used information gleaned from radiographs. Others have excluded teeth that had been extracted or had fractured, using the justification that extractions, fractures, and restoration failures were not directly caused by the RCT and therefore should not be included in assessments of RCT outcomes. Therefore, it would be instructive to conduct RCT outcome studies in community settings using multiple definitions of failure, such that the effect of the definition on conclusions can be assessed.
Practice-based research networks (PBRNs) offer an opportune venue to conduct RCT outcomes studies in community private practice settings. PBRNs have been in existence in the United States since the 1970’s.6 The purpose of these networks is to join practitioners with academic researchers in developing and answering relevant research questions that can directly impact routine clinical practice. These networks offer unique advantages both to research and quality improvement, and foster information sharing between practitioners.7-10 Traditionally, clinical research has been conducted in academic settings. However, less than one percent of Americans receive their health care in that type of setting.11 With studies initiated and developed by non-academic practitioners and conducted in their offices, such as in PBRNs, results should be more relevant to these clinicians. Therefore, these studies should lead to improved clinical treatment in a shorter amount of time in these non-academic settings, as compared to conventional clinical research done in academic settings by academic dentists.7-9 One such PBRN, called the Dental Practice-based Research Network (DPBRN), originated in Alabama and ultimately expanded to include four other regions.12 The Alabama portion of DPBRN served as the setting for this current study on outcomes of RCT.
The purpose of this study was to: 1) quantify the incidence of RCT failure in private practice settings, using three different definitions of treatment failure; 2) quantify the typical amount of time in private practice settings that it takes the RCT-treated tooth to receive a permanent restoration; 3) test the hypothesis that receipt of a permanent restoration is a significant predictor of treatment failure; and 4) test the hypothesis that other factors are significant predictors of treatment failure.
A retrospective cohort study was designed to meet these objectives. Practices in the community were identified from the Alabama portion of DPBRN. Dentist practitioner-investigators were recruited through mailings to licensed dentists. As part of enrollment in DPBRN, all practitioner-investigators complete a 101-item enrollment questionnaire about their practice characteristics and themselves. The enrollment questionnaire and other details about DPBRN are provided at http://www.DentalPBRN.org. We have demonstrated that DPBRN dentists have much in common with dentists at large.13
Practices were identified for inclusion in the current study if they were enrolled in DPBRN, had a practice address in Alabama, were general dentists, and they had answered “more than 20” to the enrollment question “How many root canals do you perform or refer to other dentists each month?”. A total of 25 practitioner-investigators who met these criteria expressed a desire to be included in the study. Of these, a total of 13 were included in the study after a visit to the practice confirmed that dental charts would consistently include the treatment information and radiographs that would be needed for the study. A follow-up date was arranged for the research assistant to screen charts for potential cases. Because most practices had hundreds of charts, a random number generator was used to identify a random start for chart review.
A research assistant, whose background included training as a certified dental assistant, was trained in chart abstraction using dental charts from dental school and private practice dental charts. Following several training sessions, she achieved standardization in proper abstraction procedures specific to this study and achieved high intra-rater reliability. This research assistant contacted all of the offices, screened the charts, abstracted and entered data, and scanned radiographs.
Dental charts were used to abstract dates of all visits in which any treatment was done on the endodontically-treated tooth. RCT was defined as ADA codes 3310, 3320, or 3330. Additionally, we also recorded tooth or areas treated; American Dental Association procedure codes14 for all procedures done on that tooth; a description of the procedure code (to ensure that the code matched its description); whether the tooth served as an abutment for a removable partial denture, overdenture, or fixed partial denture; whether the RCT was done through an existing prosthetic crown; date of data entry; and patient’s demographic characteristics. We also recorded whether the RCT was done by a general dentist or by an endodontist. Data were gathered from a total of 174 teeth in 84 subjects.
Two endodontic faculty evaluators (KRT and MJB) made measurements from radiographs taken before and after the RCT procedure. No identifying information was provided with these radiographs so as to mask the evaluators to any characteristics of the subjects. A total of 99% of the pre-operative radiographs were periapical; the remainder was panoramic. For the post-operative films, 78% were panoramic and 22% were periapical. Each evaluator made a determination whether the radiograph was of sufficient resolution to make an accurate measurement. Of the 174 RCT-treated teeth, a total of 115 teeth had radiographic measurements made. Exclusions primarily comprised either panoramic radiographs because of inadequate resolution or films with insufficient imaging of periapical structures.
The two evaluators were calibrated on the use of the periapical index (PAI; 1 – 5 scale) using an authorized reference set provided by Dag Őrstavik.15 Evaluation of the study radiographs was done independently. The evaluators jointly re-evaluated any disagreements and a consensus was reached. All radiographs were digitized before measurements were taken. From these images, the evaluators recorded the number of pre-RCT and post-RCT proximal contacts (either 0, 1, or 2); and pre-RCT and post-RCT measurements of the PAI.
Measurements of pre-RCT and post-RCT bone loss were made using a digital subtraction laboratory. A measurement was made from the crestal bone height to the root apex, which served as the numerator. The denominator was calculated by measuring the distance from the cemento-enamel junction to the root apex.
Using the pre-RCT radiograph only, the evaluators recorded whether there was radiographic evidence of caries at access (no; yes). Using the post-RCT radiographs only, the evaluators recorded whether there was any radiographic evidence of recurrent caries adjacent to the restoration that restored the RCT (no; yes); whether the RCT filling was inadequate (no; yes, if the RCT fill was more than two millimeters short or past the apex or if it had voids); the presence of a permanent restoration (no; yes, if the endodontically-treated tooth had a prosthetic crown or permanent filling); or if the restoration was inadequate (no; yes, if the restoration had radiographic evidence of an overhang or an open margin).
One measure of failure was whether or not the endodontically-treated tooth was ultimately extracted. In addition to documenting this event, the number of days elapsed between the date the RCT was completed and the extraction date was also calculated.
A second measure of failure, called “clinical failure”, was defined as having occurred if the endodontically-treated tooth was extracted or had endodontic re-treatment (ADA codes 3346, 3347, 3348, or 3410). The “clinical survival time” was defined as the number of days elapsed between the date that the RCT was completed, and the date at which clinical failure occurred or the date of the last clinical visit (in the case of no failure).
“Radiographic failure” was defined as having occurred if the endodontically-treated tooth was extracted, had endodontic re-treatment, or received a Periapical Index score of 3, 4, or 5 from the post-RCT radiograph. The “radiographic survival time” was defined as the number of days elapsed between the date that the RCT was completed, and the date at which radiographic failure occurred or the date of the last radiograph of that tooth (in the case of no failure).
All analyses were done using SAS.16 Comments about statistical significance refer to probabilities of less than 0.05. To account for the fact that individual RCTs were clustered within individual patients and dental practices - and therefore outcomes might be correlated - generalized linear mixed models (GLIMMIX procedure) were used to implement logistic regression analyses of these outcomes (extractions, clinical failure, and radiographic failure) in both Tables Tables11 and and2.2. Associations with time to failure were evaluated using an analogous multiple linear regression model, also implemented with GLIMMIX in order to account for correlated observations. The GLIMMIX procedure implements models including random effects to account for correlated observations. A pseudo-likelihood estimation approach is used to fit statistical models to data that are not necessarily normally-distributed. This approach allows the models to account for correlations among observations, as may occur when multiple observations are made on the same sampling unit (i.e., multiple teeth within the same patient), and for non-constant variability among groups. Mean age was compared between persons with and without clinical failure of teeth and between those with and without radiographic evidence of failure using t tests.
All the dentists who participated were in general practice. A total of 85% of these dentists were male; 85% were non-Hispanic Caucasian and 15% were non-Hispanic African American. They had been in practice from 11 to 39 years. These practitioner-investigators spend only a small portion of their patient contact time doing RCT procedures: 77% spend none of their time doing these procedures; 15% spend only 1% - 10% of patient contact time; 8% spend 11% - 20% of their time doing RCT procedures. The bulk of their time is spent doing restorative dentistry procedures that are not related to dental implants (e.g., amalgams, composites, crowns, bridges, posts, foundations): all reported spending at least 51% of their patient contact time doing these restorative procedures; one practice reported that the figure was more than 90%.
Practitioner-investigators reported that patients typically wait a mean (S.D.) of 9.1 (9.2) days for a treatment procedure appointment, and a mean (S.D.) of 8.9 (8.4) days for a new patient examination appointment. When asked to characterize their practice workload in the past 12 months, 20% of practitioner-investigators reported that their practice was “too busy to treat all people requesting appointments”; 30% responded “provided care to all who requested appointments, but the practice was overburdened”; 30% responded “provided care to all who requested appointments, and the practice was not overburdened”; and 20% responded “not busy enough – the practice could have treated more patients”.
The racial mix of patients in these practices varied widely. In one practice, more than 90% of patients are Black or African American. In three practices, the figure was less than 10%. The percentage of practice revenues derived from patients’ self-payments, in contrast to coming from a dental insurance source, also varied widely. Thirty percent of practices responded that 21% - 30% of revenues are derived from self-pay; 17% said 31% - 40%; 25% said 41% - 50%; 17% said 71% - 80%; and 8% said that 81% - 90% of revenues are derived from self-pay.
A total of 56% of these RCT-treated patients were female. A total of 94% were non-Hispanic Caucasian; the remainder were non-Hispanic African American. The mean (S.D.) age at the time of the RCT was 49.2 (12.9) years, with a range of 18 – 85 years old.
Seventeen percent of teeth were anterior teeth; 33% were premolars; 50% were molars. A total of 44% of teeth had the initial RCT access done through an already-existing prosthetic crown.
Three percent of teeth served as abutments for removable partial dentures at the time the RCT was initiated. None were overdenture abutments; 6% were fixed partial denture abutments at the time that the RCT was initiated.
A total of 64% of teeth had two proximal contacts at the time of initial RCT access, 31% had one proximal contact, and 4% had none.
Sixteen percent of teeth had radiographic evidence of dental caries on the pre-RCT radiograph such that there was active decay through which the RCT access was made. Twenty-three percent of teeth had a pre-RCT PAI score of 1; 29% had a score of 2; 32% had a score of 3; 15% had a score of 4; 2% had a score of 5.
The mean (S.D.) percent of bone loss at the time of the RCT was 16.7 (7.8), with a range of 0.5% to 41%. At the time of the post-RCT radiograph, this was 17.8 (8.4). The difference in bone loss between pre-RCT and post-RCT was not statistically significant.
A total of 52% of teeth had their RCT done by an endodontist; the remainder were done by a general dentist. This varied by tooth type. Thirty percent of anterior teeth RCTs were done by an endodontist, compared to 51% of premolars and 60% of molars (χ2 test; 2 df; value = 7.94; p < 0.02).
Based on the post-RCT radiograph of teeth that had not been extracted, the quality of the RCT fill was judged by the evaluators to be poor for 15% of teeth.
Antibiotics were prescribed for only 1% of the teeth at the time that the RCT was initiated.
A total of 89% of teeth ultimately received a permanent restoration. Of the teeth that received a permanent restoration, 50% had received that restoration by 37 days after the RCT had been completed. Among those teeth that received a permanent restoration, the mean (S.D.) number of days to permanent restoration was 215.4 (609.1). The range was 0 days to 6,289 days. When two outliers were excluded from the calculation (teeth with 2,210 and 6,289 days until restoration), the mean (S.D.) number of days to permanent restoration was 158.5 (285.7).
After excluding the two outliers, permanent restorations were completed sooner when the RCT was done by an endodontist (mean (S.D.) of 140.8 (239.1) days) than when the RCT was done by a general dentist (mean (S.D.) of 177.8 (329.9) days), but this difference was not statistically significant. Twenty-two of the 81 teeth treated by general dentists had the RCT done in one appointment, while 80 of the 93 teeth treated by endodontists were performed in a single appointment. This difference was statistically significant (p<.05).
Based on the post-RCT radiograph of teeth that had not been extracted, the restoration was judged to be poor by the evaluators for 2% of the teeth. Based on the post-RCT radiograph of teeth that had not been extracted, recurrent caries was present for 7% of the teeth.
Among all 174 teeth, the mean (S.D.) length of the follow-up period was 8.6 (5.1) years. When limited to those teeth that were not extracted and which had no evidence of clinical or radiographic failure, the mean (S.D.) length of the follow-up period was 8.8 (5.1) years, with a range of 1.2 – 28.6 years
A total of 18% of teeth were extracted at some point after the RCT had been completed. Among these extracted teeth, the mean (S.D.) time to extraction was 6.7 (4.5) years, with a range of 0.3 – 16.8 years.
Twenty percent of teeth were judged to have had clinical evidence of failure (i.e., extraction or endodontic re-treatment). Among these teeth with clinical failure, the mean (S.D.) clinical survival time was 7.3 (5.0) years, with a range of 0 – 18.0 years.
Thirty percent of teeth that had both pre-operative and post-operative radiographs of satisfactory quality were judged to have had radiographic evidence of failure. Among these teeth with radiographic evidence of failure, the mean (S.D.) survival time was 7.3 (4.2) years, with a range of 0.8 – 21.8. A total of 27% of teeth were extracted, failed clinically, or had radiographic evidence of failure.
Patient characteristics and tooth characteristics were analyzed to quantify their relationships to the probability of RCT failure, using three definitions of treatment failure (Table 1). Neither gender nor race of the patient were significantly associated with failure. Age of the patient was significantly associated with clinical failure and radiographic failure, but not extraction (not shown in the table). The mean (S.D.) age of the persons who had one or more teeth with clinical failure was 50.5 (11.8), compared to 52.4 (13.0) for those without clinical failure (p < 0.05; t-test). The mean (S.D.) of the persons who had one or more teeth with radiographic evidence of failure was 49.7 (10.6), compared to 54.5 (13.0) for those without radiographic evidence of failure (p < 0.05; t-test).
Whether or not a permanent restoration was received during follow-up was significantly associated with failure, as was whether or not the RCT was done on a tooth that also served as an abutment for a removable partial denture at the time that the RCT was initiated (Table 1). The number of proximal contacts at the time the RCT was initiated, and whether or not the endodontic fill was judged to be poor, were both significantly associated with failure as determined radiographically (Table 1).
Table 2 shows the results of three logistic regressions in which three different outcomes were used for analysis: (1) whether or not the endodontically-treated tooth was extracted; (2) whether or not the tooth was either extracted or had endodontic re-treatment; and (3) whether or not there was radiographic evidence of failure. All regressions were adjusted for these additional variables, although these variables were not statistically significant: subject’s age, subject’s gender, subject’s race, and tooth type.
One variable was statistically significant and had a substantive effect size (large odds ratio) regardless of which definition of failure was used: whether or not the tooth received a permanent restoration. The point estimates for its odds ratios ranged from 3.7 – 6.1, depending on the outcome used for analysis.
Other variables were also statistically significant for some definitions of failure, but are not shown in Table 2. Whether the endodontically-treated tooth was an abutment for a removable partial denture pre-operatively, was significantly associated with failure due to extraction and with clinical failure (extraction or endodontic re-treatment), with permanent restoration, subject’s age, subject’s gender, subject’s race, and tooth type also in the regression. Additionally, whether the endodontically-treated tooth had at least one proximal contact pre-operatively, was significantly associated with radiographic evidence of failure, with permanent restoration, subject’s age, subject’s gender, subject’s race, and tooth type also in the regression.
This study suggests a lower percentage of RCT success than other studies in the literature. A success rate of 90%-95% is commonly cited as a clinical expectation, and indeed, some studies have observed success rates at this level.17 A systematic review of the literature from 1966-2004 identified 306 articles that were published about outcomes of RCT,18 of which 51 articles included at least 100 teeth. The review observed that the radiographic success rate was about 82% over a five-year follow-up period; about 95% of teeth remained functional (i.e., had not been extracted) in that amount of follow-up time.
The sample size in our study exceeded the “100 teeth” threshold used in the 2004 literature review, 18 but the small sample size in this study is a limitation, although a key advantage of this study is that it used a community-based sample of practices. Traditionally, clinical research studies have been conducted only in academic settings (whose dentists, patients, and clinical circumstances may not be generalizable to non-academic settings) or studies have been conducted in settings that only included patients with dental insurance. The sample of “real-world” general dentistry practices in our study, with a mix of RCTs done by general dentists and endodontists, and with a mix of insured and un-insured patients, suggests a composite failure rate (either extraction, clinical failure, or radiographic failure) of approximately 27% during a mean follow-up period of more than eight years. Additional limitations included the retrospective nature of the design, the need to rely on panoramic radiographs for many of the post-operative assessments, and the fact that complete sets of radiographs were often not available. Nonetheless, the main conclusion of this study is that the failure rate was substantial. That is, even if no radiographs had been included in the study, the failure rate would still have been 20%, and the key predictor of this substantial failure would still have been the delay in placing the permanent restoration. The findings from this study justify further investigation in community-based practices, an investigation which should have a prospective study design and in which full measurements can be obtained using all three definitions of failure.
Many definitions have been used in the literature to define endodontic treatment success.19 Some definitions have relied on periapical changes determined from radiographs, the need for re-treatment, receipt of re-treatment, or tooth fracture. The clinical significance of these changes and their effect on subsequent treatment have not been without controversy. Some studies have used tooth retention as a key measure of treatment success.1, 2, 20-22 Clearly, a major justification for much of dental care - including RCT - is tooth retention. Using tooth retention as a key outcome measure is also consistent with studies that have used tooth retention as a measure of quality of care.23 Although the need to extract an endodontically-treated tooth may not be due directly to the RCT, the RCT nonetheless is the precipitating event that causes a chain of restorative treatments that themselves can act as direct causes of treatment failure. Approximately 18% of the teeth in this study were extracted at some point following RCT. A study by Lazarski and colleagues2 found an overall incidence of extraction following RCT of 4% after a shorter mean follow-up period – only 14.7 months. This lower rate of tooth loss may be partly due to the fact that the study population consisted entirely of subjects who had dental insurance. A large U.S. study that also comprised only persons with continuous dental insurance coverage observed a 3% tooth loss during a mean follow-up similar to our study – about eight years.1 A large study from Taiwan, which was also limited to persons with dental insurance, observed a 7% tooth loss rate after five years of follow-up.4 A community-based study from north Florida, in which only 36% of persons had dental insurance, observed a higher tooth loss rate - 19% of endodontically-treated teeth were extracted after a mean of only 24.8 months of follow-up.20 The north Florida study also confirmed in its report that RCT was never done on teeth with substantial periodontal attachment loss.
One explanation for the higher failure rate in our study is that 11% of endodontically-treated teeth never received restorative treatment by the end of the follow-up period. Temporary restorations and obturated canal systems are not impervious to bacteria, and coronal leakage has been shown to be an important factor in treatment failure.24 Many previous studies excluded from enrollment or analysis all teeth that did not receive definitive restorations. However, a lower success rate has been shown if an adequate coronal restoration is not placed,2, 25-27 although one recent study suggests confounding between preoperative periapical status and whether or not a permanent restoration is received.28 In our study, the mean time to restoration was 215.4 days after RCT among the 89% of teeth that actually received a permanent restoration at some point during follow-up. This finding is another indicator of the value of conducting observational community-based studies, and the value of including all teeth that receive RCT, not just the ones that receive permanent restorations. The time to restoration after RCT was much less than ideal, and this was a key predictor of retention of the endodontically-treated tooth.
It is disappointing to be cognizant that a commitment to treatment had been made by the patient, and the tooth had been deemed worthy of endodontic treatment by a clinician, yet many teeth ended up being extracted anyway, for whatever reason. Our study, which used populations with less dental insurance coverage and which included a community-based setting, suggests that higher failure rates should enter into treatment planning discussions when RCT treatment is pondered for these populations.
Other investigators have observed a higher failure rate among endodontically-treated teeth that were abutments for removable partial dentures, with differences varying depending on whether the removable partial denture was retained by a clasp or some other design.29, 30
While periodontal bone loss is a key factor in RCT treatment planning, and it has been associated with an increased risk for tooth loss,31 we did not find an association between tooth extraction following RCT and pre-operative bone loss. It is reassuring that – consistent with the notion that teeth are being appropriately treatment planned for RCTs in community practices by avoiding teeth with severe periodontal bone loss - none of the teeth in study had severe bone loss, with the mean bone loss being only 16%.
This practice-based study suggests a higher failure rate for RCT than reported in studies based in academic environments or in studies of populations with dental insurance. It also suggests that the length of time to initial restoration of endodontically-treated teeth (or indeed, whether a permanent restoration is ever completed) is not optimal in community settings, and suggests a key role for this factor in the ultimate success of the RCT. This underscores the importance of taking this factor into account in community settings when treatment planning for RCT.
This investigation was supported by National Institutes of Health grants R21-DE-16033, U01-DE-16746, and U01-DE-16747. The study protocol received approval by the Institutional Review Board for Human Use at the University of Alabama at Birmingham. We are grateful to Jacqueline Love, AS, CDA, who served as Research Assistant for this study. Opinions and assertions contained herein are those of the authors and are not to be construed as necessarily representing the views of the respective universities or the National Institutes of Health. The informed consent of all human subjects who participated in this investigation was obtained after the nature of the procedures had been explained fully. An Internet site devoted to details about the Dental PBRN is located at http://www.DentalPBRN.org.