INTRODUCTION

Clinicians are beholden to provide safe and effective restorations with optimal longevity, even more today than in the past, given that patients are living longer and retaining more of their natural teeth. Dental restorations, however, have a limited clinical durability,¹ with the replacement of failed restorations accounting for >60% of all operative dental procedures.² When restorations need replacement, tooth structure is lost, with the patient entering into an accelerating repetitive restorative cycle involving larger restorations, weakening of remaining tooth tissues and structures, and more complex treatments. The ability to predict longevity of restorations of different materials and types is therefore central to longitudinal patient care, let alone patient understanding of treatment provision. To date, however, there is no generally accepted simple but meaningful means of aggregating data on restoration longevity. It is the aim of this paper to propose a means whereby this goal may be achieved.

MATERIALS AND METHODS

Inclusion Criteria

To determine which papers should be included in the review, the following criteria were used:

• Studies of at least three years duration.

• Studies containing >30 restorations at baseline.

• Studies published in English or with an English translation.

• Studies from the past 30 years.

• Studies reporting on occlusal and occlusoproximal restorations in adult (>18 years of age) patients.

• Studies including “open sandwich” technique restorations, preventive resin (sealant) restorations, and tunnel restorations were excluded.

From the selected studies, the following information was extracted: name of the first author and the year of publication of the study, the length of the study in years, the number of restorations examined at the end of the study, the number of failed restorations, and the mean annual failure rate, determined by dividing the number of failures at the end of the study by the number of restorations remaining in the study divided by the number of years of observation and expressing as a percentage.

To calculate the Normalized Failure Index, the number of restorations examined in each study was multiplied by the annual failure rate to give the failure index (Figure 1). The failure indices from all the included studies were then added together, and this figure was divided by the total number of failed restorations to give the Normalized Failure Index (Figure 2). To demonstrate the use of the Normalized Failure Index, publications for composite and amalgam restorations that fulfilled the inclusion criteria were used.

View Figure

• Download as Powerpoint
• Download Figure

View Figure

• Download as Powerpoint
• Download Figure

RESULTS

The data collected in respect of amalgam and composite restorations are presented in Tables 1^12-22 and 2^{12,17,18,20,23-42}, respectively, together with the values obtained for the failure index. When the data for all the included studies were combined (indicated by the figures in the bottom line of Tables 1 and 2), the Normalized Failure Index for amalgam was calculated (2616 ÷ 2542) to be 1.0, with the corresponding value for composite (5988 ÷ 1979) being 3.0.

Table 1:  Amalgam Studies

View Fullscreen

Table 2:  Composite Studies

View Fullscreen

DISCUSSION

This study introduced the concept of the Normalized Failure Index and used it to examine the survival rates of amalgam and resin composite restorations placed in load-bearing situations in posterior teeth—with the data having been obtained from a review of controlled clinical studies dating back 30 years to the time that composites began to be used in load-bearing situations in posterior teeth. The papers included in the investigation were those that a group of dentists deemed “best available” studies, namely, studies of at least three years' duration and involving >30 restorations placed at the commencement of the study. Of the studies available, there were more addressing resin composites than amalgam. The composite studies, however, tended to include smaller numbers of restorations, and, overall, fewer composite restorations were included, although this was still substantial (n=1979). The reason for this difference may be considered to be multifactorial, ranging from arrangements to undertake clinical assessments of the many composite restorative systems introduced in the last 30 years to the cost differential between clinical studies of amalgams and composite systems. Furthermore, the amalgam studies tended to be older and longer given that amalgams, unlike composite systems, were not subject to such change and, as a consequence, the subject of extended investigations.

The large numbers of restorations included in the present investigation, together with the calculation of the Normalized Failure Index, may be considered to provide an overview of the performance of amalgams and composite systems in controlled clinical studies. In considering the overall Normalized Failure Index values, these indicated that composite systems, when assessed by this approach, performed less well than amalgams when used in similar, controlled situations. The results, however, do not concur with those published in 2004 in a review of the literature by Manhart and others⁴³ in which the mean values for failure of restorations of differing types were calculated, with the mean annual failure rate for amalgam restorations being 3.0% and 2.2% for direct composite restorations, both placed in stress-bearing situations. Such calculations, which did not take into account the numbers of restorations in each study (as the Normalized Failure Index does), indicated that resin composite restorations in posterior teeth performed better than amalgam restorations, contrary to the findings of the present work. In this respect, it is to be remembered that failure rates for occlusal and occlusoproximal restorations in controlled clinical studies tend to be much lower than in practice-based studies in which materials are applied under circumstances that prevail in the general practice setting.⁸

The results of the study by Manhart and others⁴³ indicated that composite systems had a higher mean annual failure rate than amalgams when overall mean values were calculated. It could be argued, however, that this simple calculation does not provide a valid perspective, as the different size of the studies (ie, the number of restorations included in each evaluation) is not taken into account. A novel approach was therefore adopted. The product of multiplying the annual failure rate by the number of restorations included for examination at the end of the study was termed the failure index. The failure index for all the papers was totaled and divided by the total number of restorations, to provide an index: the Normalized Failure Index. In this way, studies including larger numbers of restorations have a greater impact on the proposed comparative measure of success in clinical service. This approach of applying the Normalized Failure Index may be considered also to produce a more valid measure of performance than a simple mean age of failure, as used in many studies, given that the impact of different sizes of clinical investigations may be minimized through the normalization process. It could be argued that Kaplan-Meier statistics offer similar advantages. It is, however, suggested that the findings of such statistics and the Normalized Failure Index, may be viewed as complementary to give an informed overview of performance.

In the present work, if the average of the data presented as the failure index is calculated, the result is 1.22% for amalgam, while the Normalized Failure Index is 1.03. For composite, the average of the calculated failure index is 3.07%, while the Normalized Failure Index is 3.03. There are therefore marked similarities in the outcome of the two approaches. In this respect, it is important to recognize that normalizing data is a form of averaging, but with a provision to limit the effect of “outlying” data. In the absence of outlying data in a data set, normalized values will inevitably be very similar to average values. When outlying data are present, the average value will be somewhat misleading relative to typical values, while normalized values should be more representative. This is what we are attempting to achieve with the Normalized Failure Index: an approach that may often give values very similar to average values but a more representative value than the average when there may be outlying data in the data set or the data set is heterogeneous.

One difficulty in the calculation of the Normalized Failure Index arises if one paper is included that has a much greater number of restorations than other studies. In the first draft of the present work, the paper by Martin and Bader⁴⁴ was included, as it presented the results of 3764 large amalgam restorations. These restorations were placed by a large number of dentists (n=74) who were operating under routine clinical conditions as part of an insurance scheme (Kaiser Permanente, Portland, OR), so it may be considered to represent a wealth of data from a large database with a broad spectrum of patients, and this should not be ignored. However, it differs from the other papers insofar that it used Kaplan-Meier statistical methodology, with success of the individual restorations included being defined as an absence of replacement of the restoration, whereas in the majority of the other studies included in the present work, US Public Health Service (USPHS)–type criteria were utilized. It may therefore be considered that the Normalized Failure Index should be applied only to studies in which the understanding of failure is the same. In this regard, the Normalized Failure Index brings together data from studies that are of similar type/methodology but that may incorporate internal heterogeneity (ie, heterogeneity within the specific designs of the included papers). In the Martin and Bader paper, criteria for success are fundamentally different from USPHS criteria. In this respect, an interesting development from the present study would be to carry a calculation of the Normalised Failure Index on studies using data from large third-party databases, such as the work by Martin and Bader⁴⁴ and Lucarotti et al.¹

The authors of the present work acknowledge that long-term randomized clinical trials (RCTs) and subsequent meta-analyses are required to fully assess the value of restorative materials for posterior teeth. Such studies are, however, very resource intensive and often begin to yield meaningful results only when test materials, such as resin composite and adhesive systems, have been superseded by new generations of materials. As a consequence, there is an increasing tendency to view such studies as having limited value as an assessment of efficacy, in particular, when conducted in environments other than general practice. Furthermore, it is the opinion of the authors that it is unlikely that there will be RCTs in sufficient numbers to appropriately investigate the many varied new materials and restorative systems anticipated in future years, in particular, RCTs in the real-world setting of general dental practice, the environment in which materials find their principal application. As a consequence, the Normalized Failure Index may be considered to be a useful means to bring together data on restoration longevity from cohort studies, which are more common and in many ways more realistic and manageable in the settings in which the test materials will eventually be used. As with any index, let alone the normalizing of data, it is accepted that there are certain limitations and loss of detailed data, as included in reports of individual studies. In this regard, the Normalized Failure Index is proposed as a readily calculable new tool for summarizing the results of a number of studies that use similar methodologies. A limitation may be that values for mean failure rates, as used in the calculation of the Normalized Failure Index, do not reveal different patterns of failure, such as a number of early failures followed by a limited tendency to failure, as compared to good early performance with subsequent increasing tendency to fail (N. Opdam, personal communication). In addition, potentially substantial studies should now be commissioned to investigate its statistical validity since there is no precedent for its calculation.

An alternative method of assimilating data from a number of studies is the meta-analysis. It would be helpful to be able to compare the Normalized Failure Index with a meta-analysis for purposes of validation. The authors are not, however, aware of such an analysis having been carried out on the longevity of amalgam compared to composite systems. In addition, meta-analyses can suffer from a shortage of homogeneous comparable data from a variety of studies, with many candidate studies having to be excluded through the application of various inclusion and exclusion criteria. As such, meta-analyses are therefore generally restricted, by the nature of the analysis and the need for homogeneity of the data, to only a small number of studies. This approach has the benefit of strict criteria but invariably leads to the exclusion of large amounts of valuable data of clinical relevance. The concept of the Normalized Failure Index is to allow inclusion of a much larger number of studies that may be more heterogeneous.

Finally, the finding, according to the Normalized Failure Index, that composite resins performed less well than dental amalgam in occlusal and occlusoproximal preparations is considered to be of particular note. With the increasing application of resin composites in minimally invasive preparations, it will be of considerable interest and importance to monitor the ways in which resin composite materials may begin to be found to realize their potential in the restoration of posterior teeth and to further modify the teaching and practice of operative dentistry accordingly.

CONCLUSION

A novel method of assimilating results from a number of studies on restoration longevity, the Normalized Failure Index, has been described. According to the Normalized Failure Index, when applied to a selection of studies conducted over the last 30 years, resin composites were found to perform less well than dental amalgam in the restoration of posterior teeth.