Search

The search for articles was performed in PubMed/Medline, Scopus, and Cochrane Library databases. The search strategy was constructed by using controlled vocabulary and free terms around the terms dental restoration, amalgam, composite resin, inlays, onlay, survival, longevity, dental restoration failure, posterior teeth, clinical trial, clinical evaluation, longitudinal study, retrospective study, and follow-up. The search was performed in April 2014, and an automatic update was scheduled in the PubMed database up to the completion of this review, in April 2015.

Eligibility Criteria

For inclusion, full-text articles published in English, with the characteristics presented below, were considered:

• • 

  Longitudinal clinical studies, prospective and retrospective

• • 

  Posterior permanent teeth

• • 

  Direct and indirect restorations, class I, class II, inlay, onlay, overlay, and partial crown

• • 

  Amalgam, composite resins (direct and indirect), ceramics

• • 

  Three-year minimum follow-up period

• • 

  Information regarding patient factors (caries risk, bruxism, DMFT, etc) of the study population, including the criteria applied and/or the effect of patient factors (age, gender, caries risk, bruxism, DMFT, etc) on restoration survival

• • 

  Outcome: Cumulative restoration survival (percentage) or annual failure rate (AFR%) or information in text to allow the calculation (number of restorations evaluated and failed/replaced/repaired for a given period of time; life tables)

Studies not presenting the above-mentioned characteristics or presenting different outcome measurements (eg, median survival time) were not considered for this review. Studies presenting the above-mentioned characteristics and also including anterior teeth, primary teeth, post systems, full crowns, or different restorative materials were excluded if the outcome was not reported separately.

Study Selection

All retrieved titles were stored and managed in EndNote X7 software (Thomson Reuters, San Francisco, CA, USA). The articles identified in all databases were screened for duplicates that were automatically excluded. Titles and abstracts were screened by two reviewers (F.H.S., K.C.) independently. If the abstract was missing, the full-text article was subjected to appraisal. Disagreements were identified and discussed until a consensus was reached. References of eligible articles and reviews on restoration survival were hand searched to detect other potential studies of interest, which were screened in the same way.

Evaluation

The articles meeting the inclusion criteria were subjected to critical appraisal, which was carried out by one reviewer (F.H.S.) and checked by another (K.C.). Data were extracted using a pilot-tested table, in duplicate, and included country, clinical setting, study design, follow-up period, number of patients included, drop out, patient-related factors, number of operators, number of restorations originally included and followed, restorative material type, cavity design or number of restored surfaces, tooth, survival/AFR%, factors influencing restoration survival, size of the effect of patient-related variables, and statistical analysis performed. The survival/AFR% was either extracted from the article or calculated by the authors of this review according to information given on live tables or on length of follow-up and number of restorations evaluated and failed. To estimate the mean AFR% of the restorations, the following formula was applied: (1 − y)z = (1 − x), in which y expresses the AFR and x the total failure in z years.²²

Data Synthesis and Management

Data collection was organized into two separate parts for analysis. First, articles reporting on patient-related factors were searched for the criteria applied to classify the individuals into groups. This information was organized into one table according to the reported patient factors. For the second part, only studies that included patient factors in the analysis of the outcome (restoration survival, failure rate/failure distribution) were selected. Detailed information of these studies was organized into tables, including the significance of all investigated variables and the size of the effect for patient-related variables (when available). Some of the included studies had data on restorations placed in anterior teeth, primary teeth, and full crowns. In such cases, the extraction of data for the present review did not include those samples.

Methods to Assess Patient Risk (Part 1)

Studies addressing patient-related factors and the methods used by the authors to assess the individuals are described in Table 1 and included caries risk, caries activity, caries severity, number of restorations, oral hygiene or oral health, salivary parameters and bacterial levels, bruxism/parafunctional habits, erosion, periodontal status, attrition of the tooth structure, and smoking habits. Twenty-six studies reported to have assessed the caries risk of the patients, which was based, in most of the reports, in the present/past caries experience.^{15,16,22-33,36,42,43} Objective parameters for defining the caries risk profile were often set according to the number of new caries lesions leading to restorations in a definite period of time. In this sense, a high caries risk was established when one or more new lesions occurred per year in Opdam and others (2010)²² or two or more in a three-year period in van de Sande and others (2013).¹⁵ In Jokstad and Mjor (1991)⁴² and Nordbo and others (1998),⁴³ a high caries risk was determined when two or more lesions occurred per year, while in Fasbinder and others (2005),³⁶ the placement of four or more restorations in the previous year should have been reported by the patient. In several articles, the caries risk was reported to have been estimated by the treating clinician by the evaluation of clinical information regarding incipient caries lesions and former caries histories as well as sociodemographic data.^16,24-35 The study of Laegreid and others (2012)³⁷ was the only one reporting the use of a caries-risk assessment computer software tool (Cariogram Program⁶⁰) to classify the patients into risk groups. Although not using a specific tool, van Dijken (1991,1994),^38,39 van Dijken and others (1998),⁴⁰ Aberg and others (1994),⁴¹ and Pallesen and Qvist (2003)⁵² reported a number of indicators that were taken into account to determine the caries risk, such as oral hygiene, intake of fermentable carbohydrates, salivary microbial counts, salivary flow rates, and buffer values. In these studies, patients presenting three or more out of six negative factors were assessed as high caries risk. Other variables that can be related to the caries risk of the patient were also used, such as DMFT/DFT,^19,45,61 number of total or new restorations per patient,^46-49 caries severity,⁴⁵ caries experience at earlier ages,⁴⁴ salivary parameters, and microbiologic counts.^54,55

Table 1 Description of Methods Presented in the Studies Regarding the Assessment of Patient Factors

Abbreviation: CFU, colony-forming unit.

View Fullscreen

The assessment of bruxism or parafunctional habits in the study populations was mentioned in nine reports.^{15,29,51-53,56-58,62} When stated, the methods used to estimate this condition were based in the examination of clinical signs (eg, wear facets)^{15,29,53,56,58} and by self-report questionnaires.^15,52,56

Gender was investigated in several reports,^* as well as was the age or age range of the patients.^† A few other patient-related factors were mentioned in the studies with lower frequencies, namely, erosion and periodontal status,⁵⁸ attrition,⁵⁹ oral health, oral hygiene or plaque levels,^45,51-53,63 socioeconomic status,⁴⁹ and smoking habits.⁵³

Effect of Patient Risk Factors in Restoration Survival (Part 2)

Characteristics of the Studies

The effect of patient-related variables on survival of restorations, along with other variables, was investigated in 27 studies. General characteristics of the studies are presented in Table 2, and detailed information is presented in Table 3. Studies were grouped according to the restorative material used and included amalgam (six studies),^{19,42,53,64,68,69} amalgam and composite resin (three studies),^22,47,49 composite resin (10 studies),^‡ sandwich restorations (two studies),^16,34 and ceramics (six studies).^{56-58,62,65,66} Most studies (21) were undertaken in European countries, and 52% (14) were prospective trials. Regarding the clinical setting, 10 studies were undertaken in private clinics, seven in faculty clinics, five in public dental health facilities, one in a dental school, one in the dental clinic of a defense agency, and two in more than one type of clinical facility. The quality/failure of restorations was assessed with the criteria for the clinical evaluation of dental restorative materials for use by the United States Public Health Service (USPHS), or modified USPHS (11), the standards of quality of dental care used by the Californian Dental Association (1), clinical history extracted from patients' files (5), other predefined clinical criteria (4), and the association of methods (6), for example, by using the Fédération Dentaire Internacionale clinical criteria for the evaluation of restorations and the clinical history. The restorations were placed in both premolar and molar teeth in most of the investigations (23), filling small, moderate, and extensive cavities. One study included practically only premolar teeth (98%),⁴⁸ and three others included exclusively molars (Table 2).^37,68,69

Table 2 General Characteristics of Selected Studies According to the Investigated Materials

Abbreviations: CDA, Californian Dental Association evaluation criteria; E, experimental; FDI, Fédération Dentaire Internacionale evaluation criteria; M, molar; PM, premolar; O, observational; P, prostective; PBR, practice-based research; R, retrospective; RCT, randomized controlled trial; USPHS, United States Public Health Service evaluation criteria; USPHS-m, modified USPHS.

View Fullscreen

Table 3 Information Regarding the Size of Selected Studies, Patients' Age, Survival (%), and Annual Failure Rate (AFR%)^a

Abbreviations: AFR%, annual failure rate; NA, not applicable, retrospective studies.

^a 

Underlined information (survival and AFR) represents numbers that were calculated by the authors of this review, using data provided in the article.

^b 

Follow-up time with regard to survival/AFR analysis.

^c 

Numbers presented here are only for the variables of interest (excluding anterior teeth, primary teeth, and full crowns). In two studies, the number of patients for the variables of interest could not be determined.

^d 

Total number of patients involved in the trial is presented.

^e 

The study included amalgam (4.6%), but the analysis was performed only for resin composite restorations (81.5%), and therefore the extracted data relate to resin composite.

^f 

Only one operator was included in the analysis, and therefore the extracted data relate to him.

View Fullscreen

The follow-up times are presented in Tables 2 and 3. The first refers to the maximum period to which restorations were followed, and in Table 3, the follow-up is given according to the period used in the survival analysis (survival%; AFR%) in the original article or the period was selected by the authors of this review, taking into account the number of restorations remaining in life tables in one case.⁵³

Regarding the size of the studies (Table 3), two were large prospective trials undertaken in public dental health centers, with high numbers of patients (1873⁴⁵ and 2881⁴⁸), restorations (3286⁴⁵ and 4355⁴⁸), and operators (27⁴⁵ and 115⁴⁸) involved. The dropout of patients ranged from 0⁴⁹ up to 41%,⁴² and in most prospective studies, dropouts varied between 8 and 22%.^{34,46,54,58,61,62,64,65} Concerning the age group of the participating patients, several studies (21) had a wide age range. The studies of Roberts and Sherriff (1990),⁶⁹ Soncini and others (2007),⁴⁹ and Pallesen and others (2013)⁴⁸ included only children⁴⁹ or children and adolescents.^48,69

Effect

Regarding the statistical method in the studies, information was retrieved concerning the use of univariate or multivariate data analysis (Table 4). Statistical significance of all investigated variables (yes/no) in each study is shown in Table 4. The size of the effect (odds ratio/hazard ratio) for patient-related variables is displayed in Table 5 from available studies.

Table 4 Statistical Significance (Yes⁺/No⁻) of the Investigated Variables on Restoration Survival and the Statistical Method Applied (ie, Univariate [U] or Multivariate [M] Analysis)^a

^a 

The effect for factors presented here are only for the variables of interest (excluding anterior teeth, primary teeth, and full crowns).

^b 

Caries and other caries-related variables.

^c 

Others: oral hygiene, Al-Samhan and others (2010)⁵⁰ and Kopperud and others (2012)⁴⁵; patient as a factor, Bottenberg and others (2009)⁶⁷; heavy smoking and poor oral health, Smales (1993).⁵³

^d 

Decayed, missing, filled teeth.

^e 

Number of dentists per patient.

^f 

Number of restorations per patient.

^g 

In the three-year analysis.

^h 

Caries severity.

ⁱ 

Adhesive system.

^age Age of the operator.

View Fullscreen

Table 4 Statistical Significance (Yes⁺/No⁻) of the Investigated Variables on Restoration Survival and the Statistical Method Applied (ie, Univariate [U] or Multivariate [M] Analysis)^a (ext.)

View Fullscreen

Table 5 Statistical significance (P) and Hazard Ratio (HR)/Odds Ratio (OR) of Patients' Risk Factors Assessed in the Studies

Abbreviation: DMFT, decayed, missing, filled teeth.

^a 

Estimate coefficient from Cox regression model estimated with the bootstrap technique.

View Fullscreen

Sixteen studies included the patient age or age range in the analysis, which was found to be significant in 47% (seven) of the studies.^{42,45,48,50,53,67,68} Pallesen and others (2013)⁴⁸ investigated several variables influencing the survival of class I and II composite restorations in a large sample of children/adolescents. The study reported that among the patient-related factors, only the age range influenced the results, with adolescents showing a hazard ratio of 0.43 compared with younger children (5-11 years; Tables 4 and 5).⁴⁸ Also, in Kopperud and others (2012),⁴⁵ younger patients at baseline influenced negatively the survival of composite restorations. When age was categorized into over/under 30 years, no effect on composite restoration survival was found in van de Sande and others (2013),¹⁵ but lower survival rates for amalgam restorations were observed for patients older than 30 years in Plasmans and others (1998).⁶⁸ Two studies also reported lower survival rates in patients older than 41⁵³ and 45 years⁵⁰ when compared with other age groups. In this last study, the hazard ratios for patients younger than 15 years and older than 45 years were not significantly different.⁵⁰

Regarding gender, 23.5% (four of 17) reports found a significant effect on survival, showing more failures for men in three studies^37,62,64 and for women in one.⁵⁰

The caries risk profile or related variables (DMFT, number of restorations, and caries severity or activity) was present in the analysis of 15 studies, and a significant effect on survival was reported for high-caries-risk individuals (or related variables) in 66.7% (10) of these studies.^§ These studies included amalgam,^{19,22,42,47,49} resin composite,^{15,22,45,47-49,54} and sandwich^16,34 restorations. For individuals classified as having high caries risk, the hazard ratio ranged from 2.45 to 4.40 compared with low-risk individuals.^15,16,34 Kubo and others (2011)⁴⁶ evaluated the retreatment risk and did not find a significant effect on survival for class I and II composite restorations. In the study by Kopperud and others (2012),⁴⁵ no effect of caries severity on survival of class II composite restorations was found, but a higher DMFT score was significantly related to lower restoration survival. The study by Laegreid and others (2012),³⁷ in which the Cariogram Program was used to estimate the caries risk, reported no effect on survival of extensive composite restorations according to different risk profiles. Also, Lundin (1990)⁶¹ reported that no correlation was found when caries experience (DFT) and failure rates were compared between different composites.

The effect of bruxism or parafunctional habits was significant in three of six reports in which this factor was investigated. Studies reporting a significant effect included amalgam,⁵³ resin composite,¹⁵ and partial-crown ceramic restorations.⁶² Patients presenting high caries risk and bruxism were found to present a hazard ratio of 8.31 compared with low-risk patients in van de Sande and others (2013).¹⁵ The other three studies reported no effect of this variable on survival of inlay/onlay ceramic restorations.^56,57,62

Bottenberg and others (2009)⁶⁷ analyzed the patient as a factor and found a significant contribution of this variable on general failures of composite restorations. Patient oral hygiene had a significant effect on survival of composite restorations in the study of Al-Samhan and others (2010),⁵⁰ but the effect of this variable was not significant in the study of Kopperud and others (2012),⁴⁵ and neither was oral health significant in the survival of amalgam restorations, as reported by Smales (1993).⁵³

Caries

In the caries disease process, multiple risk indicators/predictors may be needed to establish a graded risk status and future caries prediction.⁷¹ Certainly, the collection of several variables is important to correctly identify risk indicators in each patient, guiding preventive and treatment strategies at the individual level.⁷² Nonetheless, when investigating restoration survival, the use of simplified measures may provide a good estimate of the disease activity when the restoration is placed and in follow-up evaluations. Visible cavitation or caries into dentin identified by radiographic examination was shown to significantly correlate with several caries risk factors.⁷³ Caries lesions leading to restorations within a three-year period was one of the correlated items,⁷³ which is similar to the criteria applied in some of the included studies reporting a significant influence on restoration survival.^15,22,42 Although the included studies reported different methods and cutoff points (Table 1), most of them were able to show an influence of caries-related variables on restoration failure (Table 4). Decayed, missing, filled teeth–surfaces (DMFT-S), representing past caries experience, has been used as a predictor variable and has shown that higher caries experience in the past correlates with caries increment.⁷⁴ Also, individuals presenting a higher level of caries disease (component D from the DMFT index) at the age 15 were more likely to have failed restorations when they were 24 years old.⁷⁵ Three of the included studies have used DMFT/DFT, and two reported a significant effect on restoration survival.^19,45 The other study reported that no correlation was found for DFT and failure rates, but the statistical method was not described in the article. In addition, most of the patients were dental students, which could have influenced the results.⁶¹ For studies on restoration survival, the use of cumulative scores as a single indicator may overestimate the caries risk. The increment in DMFS/DFS on a given interval of time should also be given, because it would be comparable to new restorations and cavitated lesions as reported in other studies.

Identifying high-caries-risk patients when the restoration is placed may provide a good estimate of individuals at higher risk of restoration failure.⁷⁶ A large retrospective cohort study, with seven years of follow-up, showed that high-caries-risk patients developed more primary dentin lesions as well as secondary caries compared with patients classified as no/low risk at baseline. At baseline, the most marked differences between these groups were the number of dentin lesions (0.45 for low risk vs 3.1 for high risk) and secondary caries (0.07 for low risk vs 1.0 for high risk).⁷⁴

Oral health and oral hygiene were evaluated in three studies,^45,53,50 and one reported a significant contribution of this variable in restoration survival.⁵⁰ Although individuals presenting a high level of biofilm accumulation throughout life may be more prone to oral health problems in adult life, as reported in a birth cohort study,⁷⁷ other variables should be jointly evaluated when investigating restoration survival. Still, the observation of biofilm accumulation and gingival bleeding indexes during the follow-up of patients is essential to observe their compliance to treatment.

Another variable investigated in the studies and included in the present review was the number of restorations per patient (regardless of the reason for placement) at a given period of time. Individuals with more restorations were shown to experience more failures than individuals with fewer restorations in two reports,^47,49 but this variable was not significantly related with restoration survival in two other studies.^46,48 In this sense, considering all the above-mentioned reasoning, registering the number of cavitated caries lesions, dentin caries from radiographic evaluations, or interventions due to caries within a two- to three-year period seems a straightforward method to identify patients at high risk.⁷²

Bruxism and Occlusal Loading

The general mechanisms—friction, corrosion, and stress⁷⁸—that can affect sound tooth structures in the form of noncarious tooth surface lesions may also affect restorations. Tooth wear and bruxism are multifactorial conditions, sometimes overlapping each other because mixed mechanisms may be involved.⁷⁹ For both, physiologic and pathologic distinctions should be identified, since bruxism habits may be seen as a normal activity,^80,81 and tooth wear is part of a normal physiologic process.^82,83 The assessment of these conditions usually takes into account objective clinical evaluation and subjective self-reported information.^83,84

Most of the studies included in the present review, assessing bruxism, have not objectively stated the cutoff points applied to determine the condition. Thus, a direct comparison of methods is not feasible. Among the studies evaluating ceramics, no significant effect on the failure rates for inlay/onlay restorations was found.^56-58 However, for extensive partial crowns, a significant effect for bruxism was shown in restoration survival.⁶² Regarding other materials, only two studies have investigated the effect of bruxism, and in both cases, this variable significantly influenced the survival of amalgam⁵³ and composite¹⁵ restorations. Other reports were found presenting information regarding bruxism behavior only in the discussion of the results, where more failures were seen in bruxing patients.^41,59,65

A review on bruxism prevalence in adults showed that several flaws in the studies were related to the lack of valid criteria to diagnose this condition.⁸⁵ A grading system was proposed by Lobbezoo and others (2013)⁸⁶ in which bruxism should be registered as “possible,” “probable,” or “definite.” These distinctions should be made according to the assessment strategy, namely, solely by means of self-report information with questionnaires (possible), by the use of questionnaires and clinical evaluation (probable), and, for a definite diagnosis, by the use of both preceding evaluations plus an electromyographic recording (awake bruxism) or polysomnography (sleep bruxism).⁸⁶ These distinctions seem useful for the awareness of clinicians and researchers that bruxism may be under- or overestimated, especially when only one method is applied.⁸⁷ In addition to this grading system, the severity of bruxism should be part of the assessments.⁸⁵ Questionnaires designed with this purpose should include response options other than simply “yes” or “no,” such as proposed by Liu and others (2014)⁸⁸ for tooth wear assessment, in which “mostly,” “sometimes,” and “never” were included. As a fourth response option, “not aware” could also be added. For the clinical evaluation, specifically concerning the intraoral examination, indexes should be used to grade the severity of clinical signs.

In addition, little is known about the effect of high occlusal loads and stress concentration on tooth surface loss and on restorations, except that several mechanisms may be involved.⁷⁹ In vitro studies on occlusal load frequently focus on abfraction on premolar teeth, and stress concentration in the cervical area was shown to slightly increase when an occlusal restoration is present.⁸⁹ Probably the cavity configuration as well as the axis and force of applied loads will generate distinct stresses on different teeth. Material properties,⁹⁰ the occlusal load, and the cavity type⁹¹ were shown to influence stress concentration patterns. This might be particularly relevant for restorations placed in patients presenting high occlusal stress risk, due to bruxism, parafunctional habits, heavy occlusal loading, or severe tooth wear. Hence, for practical reasons, the measurement of clinical signs regardless of the name of the condition seems advisable when investigating restoration survival.

Ideally, assessment strategies developed with this purpose should be appropriate for use in research trials but specially by practicing dentists.Factors taking long periods to influence restoration survival are probably more suitable for practice-based research in which patients are usually not particularly selected, as seen in several of the studies included in the present review. With time, moderate to severe conditions will be identified by the patient and/or the clinician during routine dental appointments and should be clearly distinguished in clinical files. For example, the degree of attrition of the tooth structure as reported by Felden and others (2000)⁵⁹ may serve to measure tooth structure loss regardless of the etiological factor(s).

Gender and Age

Few studies have found a significant influence of gender on survival of restorations, and this variable is probably related to others. Men, in general, may have stronger bite forces than women,⁹² which could contribute to more failures due to fatigue of the material or bonding interfaces, leading to fracture and debonding and increased failure rates. As discussed by Schulz and others (2003),⁶⁶ the combination of a patient effect, such as unfavorable loading, and an inadequate material dimension may have contributed to a higher failure rate in men observed in their study. The presence of bruxism and parafunctional habits may overcome the influence of gender, and therefore gender should not be an isolated variable when evaluating restoration survival. In addition, women are more concerned with their health and they attend dental services more regularly.⁹³ Such an aspect is important because it has been demonstrated that individuals having regular dentist visitations during the life course may exhibit better oral health outcomes.⁹⁴

The same line of reasoning may be valid when considering the influence of age on restoration survival. Other factors, such as caries risk in younger individuals⁴² or more complex restorations and greater tooth structure loss after several restorative interventions in older individuals, may superimpose the effect of age. Pallesen and others (2013)⁴⁸ observed, among children and adolescents, a higher intervention rate for younger individuals at baseline. The authors discussed that findings could be related to differences in caries risk and the more difficult cooperation of younger children during treatment procedures. So although age may present a significant effect, polarized for the very young and more mature patients,^45,48,50 the analysis of the contribution of age on restoration survival, as it is for gender, should not be seen under an isolated perspective.

Other Patient-Related Variables

Socioeconomic status and educational level may also influence restoration survival,⁷⁵ but no longitudinal evaluations investigating the effect of socioeconomic vulnerability were found. Although Soncini and others (2007)⁴⁹ characterized the participants according to ethnicity, household income, and educational level of the caregiver, this information was used primarily to verify the equal distribution of the materials (amalgam and resin composite) for each of the displayed characteristics. It is also mentioned that the statistical model was adjusted for some patient factors if they were statistically significant or changed the effect (10% or more) of the restoration material. Since for permanent teeth, the model was adjusted only for number of restorations in the mouth, the other factors (age, sex, and socioeconomic status) presumably did not influence restoration survival. However, the collection and reporting of these data are of importance because they provide the characterization of the sample population. In fact, when evaluating the survival of restorations in specialized private practices, a more favorable environment may be displayed because patients with a higher socioeconomic background usually attend these facilities,^63,95 especially considering countries where the dental health system does not rely on public coverage.⁹⁶ Thus, a better general and oral health may be expected, with lower chance of failure, and restoration survival may be overestimated for the general population, where individuals with different economic backgrounds are included.¹⁰

Statistical Analysis

One important point to be raised when investigating patient risk factors for longevity of restorations is the use of appropriate statistical analysis. Generally, a descriptive analysis of interest variables is recommended, followed by the analysis of associations between each evaluated patient factor and failure of restorations, often called univariate analysis. From the 27 selected studies, 10 have analyzed factors associated with longevity of restorations only in a univariate way (Table 4). This strategy does not consider the complex interrelationships that exist between all covariates investigated. For example, it is well established in the literature that dental caries is a multifactorial condition, affected by socioeconomic, behavioral, and tooth factors, among others. In this way, to investigate the real effect of caries on longevity of restorations, it is strongly recommended to adjust its effect by other variables that are associated with both caries and longevity of restorations, which can act as confounders of this association, using multivariate methods. An increasing tendency to improve the quality of the analysis using multivariate models is observed among included studies. Another aspect that requires attention when patient factors are investigated is data organization. Most of articles on longevity of restorations considered that all variables are at the same level of organization, ignoring the complex nature of dental studies, where variables from surfaces/teeth/patients are analyzed together. Generally, more than one restoration is evaluated per patient. In this case, the assumption of independence between observations (restorations) leads to errors in data analysis and interpretation of results. Restorations are clustered within patients. This means that an important correlation exists between restorations of the same patient. The use of methods that ignore this correlation may lead to incorrect results, increasing the probability of rejection of the null hypothesis (ie, finding statistically significant results when none are present in the data).⁹⁷ This problem is present in most of articles that use a survival analysis approach, by conventional Cox regression models. To deal with data organization, multilevel models are the appropriate method that adjusts the results by correlation existing between restorations from the same patient.⁷⁵ Recent studies on longevity of restorations have used Cox regression models with shared frailty to investigate patient risk factors.^15,63 These models for survival analysis are analogous to multilevel regression models with random effects and consider the intragroup correlation being recommended for future studies.

Final Considerations

The selection of patients for particular treatment alternatives is often restricted to certain risk profiles. Recommendations for restorative techniques according to patient-related conditions are made, regardless of sound evidence to support the clinical decision.⁹⁸ Interestingly, the methods used to estimate the risk, meaning the criteria applied for patient inclusion/exclusion, are frequently missing, and the description for patient exclusion is often limited to “poor oral hygiene” or “patients with bruxism were excluded.” While in retrospective evaluations, investigators may be limited to work with information available in the clinical records, in prospective studies, the characterization of the sample population (by means of indexes, self-reported information, and cutoff points) should be far more complete and available for the reader, even if data will be presented only descriptively. A recent report reinforced the need to use guidelines when reporting clinical studies, to increase the completeness and transparency of biomedical research. Inadequate reporting of research may lead to wasted research resources, increasing the risk for publication inaccuracy or biased data, with implications for health care decisions.⁹⁹

A survey among general dentists in Kuwait showed that the dentist's choice regarding direct restorative materials is influenced by factors such as oral hygiene, numbers of restorations in the mouth, and cavity size.² However, no strong evidence exists to support the use of a particular material for either situation.¹⁰⁰ Material choice seems to be related to dentists' preference, country, and cultural trends.^5,101-104 Future investigations should deal with individuals' particularities and risk factors, assisting the clinical decision making for materials and techniques in challenging situations.

It is noteworthy that studies evaluating the survival of restorations have been mostly focused on the comparison between materials, including a very restricted group of patients. Aiming for studies more easily translated to daily clinical practice, investigations including patients with different socioeconomic and education backgrounds, with different levels of caries and occlusal stress, should be encouraged. Another interesting approach would be to set new prospective studies on longevity of dental restorations recruiting only volunteers/patients at high risk, considering that these risk situations would be the utmost challenge for the restorations. Also, since the events experienced during the life course may affect a series of oral health outcomes in other periods of life,¹⁰⁵ they should be considered during the design and evaluation of studies reporting the longevity of posterior restorations. Considerable time in clinical practice is spent on replacing failed restorations,¹⁰⁶ with a high cost for the individuals and for health systems.^107,108 Restorations are replaced/repaired, and in the near future, they tend to fail again,^109,110 because the dentist is treating the consequences instead of the causes for failures.⁷⁴ Therefore, the investigation of factors related to patients is crucial to change their current status, increase the survival of restorative procedures, and cut costs.