Multi-state outcome analysis of treatment interventions after failure of non-surgical root canal treatment: a 13-year retrospective study

Abstract Objective To examine the factors affecting the transitions through treatment interventions after failure of non-surgical root canal treatment (NS-RCT). Methodology Insurance enrollment and claim information for enrollees of Delta Dental of Wisconsin (DDWI), USA were analyzed for 438,487 initial NS-RCT procedures to determine the effect of initial provider type and other covariates on additional treatments (no additional treatment, nonsurgical retreatment, surgical retreatment and extraction). A multi-state model was created using the “mstate” R package. Transitions between the four states identified by Code on Dental Procedures and Nomenclature were analyzed. Cox proportional Hazards regression stratified by transition type was used to estimate the effect of provider type on the risk of each transition, adjusting for covariates. Results The overall survival rates for all teeth that were treated by NS-RCT was 82.8% [95% CI 82.57%, 83.11%] at 10 years. Approximately, 7% of cases changed from the first state of initial NS-RCT during the 13-year study period with ultimately 0.9%, 0.4% and 5% of cases receiving non-surgical retreatment, surgical retreatment or extraction, respectively. Teeth are more likely to be retreated non-surgically than surgically, and to be extracted than retreated. In general, the probability of a tooth having non-surgical retreatment was higher if the initial provider was not an endodontist (Hazard Ratio (HR)=3.2). Molars were more likely to be non-surgically retreated (HR=2.0) or extracted (HR=2.8) when compared to anterior teeth. The probability of non-surgical retreatment (HR=0.93) or extraction (HR=0.50) was lower when a crown was placed within 90 days after NS-RCT. Conclusion Most teeth remained in the same state after treatment with no additional treatment transitions. When a transition occurred, it was more likely to be an extraction. Type of provider, age, location of the tooth, gender, and time to placement of final restoration significantly influence treatment transitions.


Introduction
When dental caries or trauma lead to pulp and periapical pathosis, non-surgical root canal treatment (NS-RCT) is often the most common and conservative treatment option available to save the natural tooth.
The success rates of endodontic therapy have been reported ranging from 81% to 97%. [1][2][3] Previous studies on prognosis of endodontic treatment report that the type of provider, age of the patient, type of the non-endodontist-general dentists and providers from all other dental specialties) and other covariates such as location of tooth, age of the patient or time to final restoration affect these transitions.
In this study, we used insurance claims and enrollment information from a 13-year period to examine the factors affecting the transition states.
Multi-state models (MSM) are generally used to model the outcomes in studies where participants may transition to any or all finite set of events, generally randomly, from one state to the next. 11 The models can provide predictions for multiple outcomes simultaneously. The MSMs also allow us to examine the effect of covariates on the transitions, to estimate the progression and survival rates in transition stages, and even the overall prognosis of the tooth.
Knowing the information about treatment transitions can greatly help in clinical decision making and developing a proper strategy by non-endodontists to refer these cases to specialists for better prognosis.
Selection of alternate treatment options should be done based on the best available evidence. The endodontic literature on studies at a high level of evidence regarding decision making on treatment options after failure of NS-RCT is scarce, and the consensus among dental professionals is insufficient when making decisions related to what is next after persistence of periapical lesions and/or symptoms after NS-RCT completed. 12,13 This study aimed to examine the factors affecting the transitions through treatment interventions after failure of non-surgical root canal treatment (NS-RCT).

Methodology
The subjects in this study were enrollees of Delta Dental of Wisconsin (DDWI) USA that underwent a non-surgical root canal treatment (NS-RCT) between January 1, 2000 andDecember 31, 2013. This is the same enrollment and claims database used in our previous study. 4,5 The dataset contained demographic information of the enrollees, start and end dates of dental insurance coverage, as well as all dental claims with date of service and procedures performed. DDWI is the largest private dental insurance and benefits program with more than 1.25 million enrollees.
DDWI has the largest provider network in the state of  We followed the teeth for evidence of any additional treatment interventions after the initial NS-RCT.
Each additional treatment-nonsurgical retreatment (D3346, D3347 and D3348 for anterior, pre-molar and molar teeth, respectively), surgical retreatment (apicoectomy-D3410, D3421 and D3425 for anterior, pre-molar and molar teeth, respectively) and extraction (D7140) was considered a transition state. An NS-RCT treated tooth could potentially have no additional treatment (successful), be retreated, or be extracted.
When teeth were retreated, non-surgically or surgically, we continued to follow them for further interventions.

Results
The information about 438,487 patient encounters in 325,290 subjects for NS-RCT's was included in this study after eliminating individuals that did not meet our study criteria. In total, 105,287 subjects had a  Table 1 shows that molars received most of the NS-RCTs and anterior teeth were the least likely to receive an NS-RCT. A statistically significant difference was found in the teeth type which endodontists and the non-endodontists completed NS-RCTs (p<0.001).
On average, the patients treated by endodontists were older than the patients treated by the nonendodontists. Table 2 shows the number of events for each of the four possible transitions. In total, 438,487 teeth had initial NS-RCT and 407,336 (~93%) of those teeth had no additional treatments. Extraction was the most common intervention after NS-RCT followed by non-surgical retreatment. No additional treatments were found in most teeth retreated non-surgically or surgically. Extraction was a more common additional treatment than a surgical retreatment after a nonsurgical retreatment.
In Figure 2 (cumulative hazard plot), we report the transitions between treatment states over a 13-year study period from NS-RCT completion. In general, teeth were more likely to be extracted than retreated.
The teeth which received surgical or non-surgical retreatment were more likely to be extracted than those that did not have such an intervention. Teeth were more likely to be retreated non-surgically than surgically. If a tooth had a nonsurgical retreatment and then subsequently had a surgical retreatment, then it was more likely that the surgical intervention occurred  In Table 3, we report the transition probabilities for teeth that received initial NS-RCT from endodontists and non-endodontists. The probability of teeth receiving further treatments after initial NS-RCT is very low irrespective of the type of provider in the 12-year follow-up period. Only a small number of teeth had subsequent interventions. In case of both types of providers, extraction was the most common type of intervention. The probability of a tooth with NS-RCT being retreated (non-surgically or surgically) or extracted was higher when the initial NS-RCT was completed by non-endodontists (Table 3).     removing the old filling, treating missed canals and improving any kind of shortcomings of the previous treatment is the same as the initial therapy, which is to remove the infection and create a favorable environment for healing. Deciding how to proceed after an endodontic failure is complex. The quality of the previous treatment and/or the restoration determines the further course of treatment after endodontic failure. 12,15 If the provider determines that they can improve the quality of the initial root canal therapy and navigate previously unaddressed canal space without drastically weakening the tooth structure, then the treatment decision would be to retreat non-surgically. 15 Nonsurgical retreatment is considered as the first line of treatment for an endodontic failure if the tooth is restorable. 16 This study identified a greater likelihood of a nonsurgical retreatment than a surgical retreatment. We found that approximately 13% of secondary treatments had a subsequent intervention which would indicate a failure of the treatment. This is similar to a study by Ng, Mann, Gulabivala 17 (2008) in a systematic review, which reported a success rate of 77% for secondary root canal treatment. 17 The multi-state analysis also found that both non-surgically retreated or surgically retreated teeth had similar probabilities of being extracted, which was greater than teeth that did not have secondary treatment after the NS-RCT. This finding differs from the metaanalysis by Torabinejad, et al. 13 (2009), which found that endodontic surgery offers more favorable initial success and nonsurgical retreatments have favorable long-term outcomes. 13 On the other hand, our results agree with the meta-analysis conducted by Del Fabbro, et al. 12 (2016), which found no significant differences in the long-term outcomes between surgical and non-surgical retreatments. 12 Haxhia, et al. 18 (2021) also found no differences in outcomes between nonsurgically retreated or surgically retreated teeth.
The most common intervention after NS-RCT was the extraction of the involved tooth, a finding similar to what was previously reported. 3,10,14 Among the studies which examined the reasons for tooth exactions after endodontic treatment, Touré, et al. 19 (2011) reported that the reasons for extraction were periodontal disease in 40.3% and endodontic failure in 19.3% of cases, whereas Chen, et al. 20 (2008) reported that only 10% of the extractions were due to endodontic failure. Extracting endodontically treated teeth may be due to non-restorability, patient finances, crown or root fractures, or provider philosophy. Clinicians may lack confidence in the success of retreatment therapy leading to increased pressure to replace "failed" endodontically treated teeth with implants. 21 However, Kim found that, after primary endodontic failure, the most cost-effective treatment was microsurgery.
This was followed by nonsurgical retreatment, then extraction and fixed partial denture, and the least cost-effective treatment was a single unit implant. 22  Nevertheless, in our study, more teeth with failed NS-RCT were extracted than retreated.

Surgical RT -> Extraction
When considering transitions to additional treatment states, the cumulative hazard for most transitions accumulates at an almost constant rate, implying that the risk of these events does not change over time. The transition from non-surgical to surgical retreatment has a very different shape: RT. However, the risk is very low. In a previous report published using the same dataset, Burry, et al. 4 (2016) found better treatment success when the provider was an endodontist. We found a decrease in the risk of transition from NS-RCT to non-surgical RT with an increase in age indicating a lower probability of failure among older individuals. A meta-analysis by Kojima, et al. 23 (2004) found no significant difference between age groups in endodontic success. 23 However, Ørstavik, et al. 24 (2004) had a finding similar to our study. They found that the results were better for the older age groups. They postulated that progressive reduction of pulp space with age limits space for infection and makes it easier to provide adequate canal debridement and filling. 24 We found that the risk of retreatments after NS-RCT was lower among men although they had a marginally higher risk for extraction than women. A meta-analysis and a prospective study both by Ng, et al. 25, 27 (2008Ng, et al. 25, 27 ( , 2011, as well as the Toronto Study, 27 reported no significant differences in success of NS-RCT between men and women. Our analysis also found that the presence of permanent restoration within 90 days after the NS-RCT had influenced the treatment transitions of the tooth with generally positive outcomes among those who had a post/core, and crown within 90 days of NS-RCT. This confirms findings from previous studies by Yee,et al. 5 (2018), Salehrabi andRotstein (2004) andLazarski, et al. 14 (2001).
To the best our knowledge, we presented the first application of a multi-state model to data from subjects with NS-RCT aiming to introduce the advantages of this type of analysis of outcomes of dental treatments. For our study, only transitions to a higher level of reintervention were allowed, since once a tooth receives a treatment or retreatment, the next treatment is typically more complex. For example, surgically retreated tooth will typically go for another surgical retreatment or extraction and rarely undergo a nonsurgical retreatment.
We excluded patients that had a follow-up shorter than 90 days (34,616) after their initial NS-RCT.
Having a definitive restoration within a reasonable timeframe after NS-RCT is a strong predictor of survival of endodontically treated teeth. Hence, we wanted to incorporate having a core/post and crown by 90-days as predictors. The easiest way to achieve this objective was to start counting from 90 days, since this would eliminate any ambiguity about the status of the restoration. We also excluded 3,376 patients with a failed NS-RCT in the first 90 days after the procedure which is 0.7% of teeth with NS-RCT. This translates to an annual incidence of 2.7% which is higher than the incidence of extractions during the remaining study period. We suspect this may be due to errors in NS-RCT technique (e.g., missed canals) or improper case selection (e.g., cracked teeth) which can lead to persistence or aggravation of symptoms, leading to an extraction.
This study has a very large study population, which allowed us to evaluate the true outcome of teeth including all the endodontic treatments that occurred during the tooth "lifespan" by performing the DDWI entire database for an extended period (a 13-year period). This large study population also allows the statistical analysis to detect minor departures from the null hypothesis. The immense dataset can minimize the effects of variations in treatment or providers.
It also provides a way to study tooth survival and true outcomes of teeth treated by NS-RCT in the real world. Many studies are performed in residency programs or evaluating smaller groups of private practices. An important limitation of these studies is that they are only representative of their office and the treatment and decisions by their referring dentists. 26 With this study, we have access to the true outcome of teeth treated across the entire state of Wisconsin with a broad variety of patients and providers. As a retrospective analysis of administrative data, it eliminates any provider-related biases in treatment planning decisions. This allows for this study to yield pragmatic outcomes and provides information as the treatment be provided to a large population.
In retrospective insurance studies, is impossible to have standardization of the providers or attempt to understand the rationale for a treatment decision. They cannot provide insight into the quality of treatment provided or if proper techniques were followed. The inability to understand the rationale for treatment can result in underestimated survival, since providers may be extracting teeth that are otherwise restorable or choosing not to retreat a tooth that may have a good chance of success in favor of an implant. It is also impossible to consider additional factors that may affect the survival such as the periodontal health of the patient, pulpal and periradicular diagnosis of the tooth or remaining tooth structure before the.
Additionally, considering that this study is evaluating survival, the teeth studied that have survived may not actually be a true successful treatment. For example, in instances that the teeth could have asymptomatic lesions associated with them or in cases in which patients have not received needed treatments. Finally, DDWI is a private insurance carrier in one state and the results may not be generalizable to other populations.

Conclusion
Most teeth remained in the same state after treatment with no additional treatment transitions.
When a transition occurred, it was more likely to be an extraction. Type of provider, age, location of the tooth,