Restoration longevity is an important issue in dentistry and especially for the Dental Materials community. The importance of a dental restoration with a good longevity is obvious for the patients who wants to keep their teeth functioning without problems throughout their lifetime. For manufacturers, high longevities suggest that their restorative materials have appropriate physical properties and will be accepted (and purchased) by dentists for clinical use. The researcher uses longevity as a clinical criterion to establish the quality of restorative materials or techniques and differences between them. Moreover, for public health care organizations and insurance firms, a long lasting restoration reduces the expenditure on dental care.

For materials scientists the longevity of a dental restoration is often related to the dental materials involved. It is assumed that the strongest, most wear resistant materials, when placed by an excellent operator with a good adhesive technique, will provide the most successful long lasting restorations. When results from dental restorations placed by general practitioners show poor outcomes in cross-sectional studies, it may be assumed that some general practitioners have used sub-optimal techniques and placed restorations in a quick and sloppy manner. This could explain the too common disappointing outcomes of restorations placed in general practice. But is this really the case or could we be doing something wrong in our clinical research designs and interpretation of results?

Recently I received for review a longevity paper based on a huge insurance database. From that database, the declaration of a certain restoration and the next restorative intervention (new declaration of a restoration in the same tooth) was used as a method to determine restoration longevity. In a huge database of several thousands of restorations, this can be considered as an appropriate method as it provides – within certain limitations – the lifetime of that specific restoration. The large number of restorations included compensates for flaws and mistakes in the data acquisition. The problem arises when the median lifetime of all those ‘failed’ restorations is then considered as representative for the entire population of all restorations placed. Earlier cross-sectional studies were based on this method and led to the conclusion that restorations placed by general practitioners have a high replacement rate with median survival of 6 years for posterior composite and 9 years for amalgam . In 2012 our research group published a paper suggesting that this method is misleading and should not be used for longevity analysis, as data of all restorations still in function are missing. This leads to a gross underestimation of actual longevity . We addressed the need for the Kaplan–Meier method in which data of all failed and surviving restorations are included with still functioning restorations being censored at the end of the observation time. Insurance datasets and cross-sectional surveys in general practices on failed restorations that do not have data on still functioning restorations are therefore unsuitable for longevity determination. As an example in the above mentioned paper, that I was requested to review, authors were concerned as they concluded that, over the years, the quality of restorations placed by dental practitioners was decreasing as the median survival of restorations was lower for the more recently placed restorations. This is inevitable as, for a group of restorations placed 2 years ago, the maximum age of a failed restoration cannot be more than 2 years, while for restorations placed 10 years ago this will be 10 years. As this study design does not provide data on the still functioning restorations, it ignores the fact that, for restorations placed only 2 years ago, most will be still be in function, while for restorations placed 10 years ago, a larger proportion will have failed already. So the often-presented conclusions, based on cross-sectional surveys and insurance data, that longevity of restorations by general practitioners is low or even the (absurd) conclusion that the quality of dentists work is diminishing, are reached via this incorrect analysis.

For an appropriate longevity dataset and analysis, all data from placed restorations, failed and not failed should be available. The many cross-sectional studies that have been published in the past have contributed to a negative image of the quality of restorative work by dental practitioners, but this seems unjustified. Data from a practice based research network in the Netherlands suggest that the median survival-time of more then 350.000 direct restorations placed by practitioners averages at 10 years but varies among dentists from 7 to 20 years. Amalgam and composite materials show comparable outcomes. This is very different from results of cross-sectional studies that typically show a median survival of 6 years for composite and 9 years for amalgam .

Another recent development in longevity analysis is the growing appreciation that, not only material properties, but other factors, such as the patient and dentist, probably have a greater influence. Patients with a high caries risk have more than double the risk of restoration failure, compared with a low risk patient . Other patient factors such as bruxism are also relevant for restoration survival . However, many clinical studies are not taking those risk factors into account as investigated in a recent review . In many clinical trials patients are included ‘in need for a restoration’ and are randomized for material A or B. If, in a RCT, the risk factors of patients are not included, a considerable inclusion bias may be present in the design, which may lead to false conclusions. A typical example is the ongoing debate on the performance of posterior composite restorations compared to amalgam. Recently, a Cochrane review was published on this issue. As only limited randomized clinical trials are available on this issue, the outcome of the review was mainly based on two NIH funded studies that compared the longevity of amalgam and composite resin restorations . Due to the fact that the health effects of the restorations were the main topic of investigation, the inclusion criterion was that the amalgam or composite restorations should be the first restorations in the mouth, which led to the design that children were included. The reason for placing a first restoration in children is almost always primary caries, so the conclusion that a high caries risk population was treated is obvious, but nowhere is this mentioned, neither in the papers, nor in the Cochrane review resulting from it . The Cochrane review concluded that amalgam has a better chance for survival compared to composite, but is this a justified conclusion? Should it not be that when relatively small restorations are placed in caries active patients, amalgam has a better longevity? So actually we really don’t know from these RCTs what the result of the comparison will be in the (majority) of adult people with large restorations that were placed in the past when they were young and caries active due to high sugar consumption and absence of fluoride in the toothpaste. Now these often very motivated patients have a low caries risk and the predominant reason for failure is fracture of a restored tooth and not caries.

Most review papers end with the conclusion that more randomized clinical studies are necessary to build the body of evidence. But who is going to do the comparative study between amalgam and resin composite for adults, including caries risk assessment? More than 5 years of observation time is needed to have results, preferably 10 years or more. Can we find funding for such a project, and more than this, are we going to find patients that want to have a resin composite or amalgam at random for their defective restoration in the first molar? This remains very doubtful!

In contrast, most of the randomized clinical trials in dental materials are ‘going the easy way’ and are comparing not very exciting differences between groups, like two adhesives or the newest material from a manufacturer compared to the predecessor, preferably placed in low-risk patients. Observation times do not exceed more then 3 years (as a newer version will be on the market by then) and obviously materials have similar behavior. But that is not the information the clinician needs. Dentists need evidence whether a crown is a better option then a direct restoration, whether a full ceramic bridge is as good as a metal supported bridge! They want to know what to do in high caries risk patients or for those bruxing patients that rapidly destroy all type of restorations! Meanwhile researchers exclude high-risk patients from the study population, probably because the sponsoring manufacturer is anxious to have an outcome as high as possible.

Isn’t it time to reconsider this? Shouldn’t we question the often proclaimed ‘Mantra’ that randomized clinical trials are the best design for investigating clinical restoration performance? Dental restorations are not a new drug supplied in a pill. We don’t have placebos and as a result, blind testing and evaluating is virtually impossible. Relevant effects are not found within a couple of months, it needs often 10 or more years to expose differences: not very feasible for a clinical trial. Should we continue to exclude high-risk patients to have the best possible results? Shouldn’t we do more research in specific risk groups, because that may help the profession to find solutions for difficult cases?

That would also mean that the Cochrane review as the ultimate ‘Evidence’ is not the best way to find that evidence. Most long-term clinical studies are retrospective or not controlled and randomized and therefore excluded from the review. But aren’t we being delusional if we only rely on RCTs?

If we want to have answers on all these clinical questions, we should reconsider our clinical trial designs. Recognizing the problems of RCTs we should probably look for other – more practice based – designs that are suitable for answering our research questions, accepting that randomization and high levels of standardization are then not possible, but have the advantage of including high-risk patients, large numbers of restorations and the possibility to investigate the probably predominant factor, which is the dentist.