Article
I was recently researching the topic of restoration longevity in the UK. Why, you might ask! Because, in my view, being able to measure restoration longevity/survival is central to providing patients with realistic expectations (thereby avoiding medico-legal complications) and ultimately for maintaining faith in our profession.
The literature took me back to a publication in 1969, which I identified as the first UK paper published on this important subject, with there being little prior to that. Readers will have realized that, prior to the arrival of computer technology to aid the compilation of data, this was a time-consuming and laborious task, therefore, it was not frequently undertaken. In that regard, one could consider that the analysis of data from patients' dental records was a labour of love by enthusiasts in the field. One such enthusiast, in 1969, was Douglas Allan, an academic in the Sutherland Dental School, which is what they then called the dental school in Newcastle upon Tyne. For his ‘statistical survey’, he used ‘unselected patients who attended for conservation’, and included 887 restorations.1 All filled teeth were critically examined by one examiner (the author), and the length of service of the restoration charted in years. At that time there were no clear guidelines as to what actually could be defined as failure of a filling, hence he defined the ‘rules’ for his project which are likely to surprise readers, namely, that a filling was charted as failed if it required extension, or if the tooth was extracted for orthodontic or prosthetic reasons. He used two categories for failure: (1) failures due to poor restorative technique (classified as true failures); and (2) failures not due to poor restorative technique (false failures). The most prevalent failure was ‘lack of extension’ (in the way decreed by GV Black (25%) and ‘faulty marginal integrity’(41%)). Allan's statistical analysis indicated no difference between class I and II amalgams and the equivalent gold inlays – possibly also a surprise to readers. His analysis provided no survival data, but he added that ‘it is the way that a material is employed is more related to its duration than the actual material used’, something that is still true today.
Move on 2 years, and, in another labour of love, Robinson2 analysed the records of 43 patients from his London practice who had attended continually from 1949 to 1969. This involved ‘detailed examination of the records' – think about going through the notes of these patients year on year, and recording the data, presumably by pen and paper! He also published ‘rules’ to indicate the failure of a restorations. Again, all restorations in an extracted tooth were deemed to have failed, while occlusal restorations that became part of a subsequent class II were also considered failures, although in discussion, he admitted that these criteria were perhaps harsh! Results indicated that 112 amalgam restorations had failed at 20 years, 33 remaining in place. Robinson also pointed out that this was a special, dentally conscious group of patients, not necessarily representative of the whole population. But, it was a start in restoration survival assessment.
Allan was obviously a glutton for punishment (by data), or, like me, he believed that the ability to measure restoration longevity was central to successful clinical practice. He therefore set up a more elegant project when records from a practice in NE England were made available. He analysed the records of 47 continually attending patients from 1951 to 1971 and 31 such patients from 1954 to 1969.3 His analysis indicated that half of the amalgam restorations were lost at 8 years and ‘nearly’ 90% at 20 years, producing graphs to illustrate these points. In total, he followed 241 amalgam restorations and 53 silicate restorations for 20 years.
What a lot of work! He later wrote4 – ‘my interest in the duration of dental restorations was awakened in the early 1950s, on seeing the genuine need to replace so much work’, ‘so the gathering of statistics commenced as a one-man research project’. Furthermore, he alerted clinicians and researchers to the difficulty of deciding between what was a viable restoration and one that had failed – easy readers say, we make decisions on that every day! But, are those decisions consistent day after day and, upon what criteria are they based? Allan alluded to this problem, writing there are myriad causes of failure, ‘these varying between acts of God to decisions involving extraction prescribed by an orthodontist’!
The north-east of England was obviously a hotbed for the analysis of restoration longevity, given that another UK publication5 on the subject described the evaluation of restorations at an NHS practice in NE England, where Neil Patterson, the author worked. He defined failure of a restoration if (1) all or part of the restoration was removed and/or replaced; (2) if endodontic treatment was carried out, necessitating removal of the restorations; and (3) if the tooth was crowned or extracted. Patterson examined the dental records of 200 patients who had attended the practice regularly for 10 years or more, thereby including 2344 amalgam, 546 silicate and 130 composite restorations – the first time that composite had been mentioned in this context. The 50% survival time of class I amalgams (n=854) was 8 years, class II (n=1490) was 8 years, not statistically significantly different. The median survival time of class III silicate restorations was 5 years, and class V around 8 years, while the small sample of composite restorations provided 50% survival at 4.5 years.
Allan4 stated that he realized that he could ‘never produce finite figures’, but these studies were groundbreaking, providing, as they did, meaningful figures for success times of one type of restoration versus another. Again I add, what a lot of work in the pre-computer era!
At this point, it may be appropriate to mention silicate cement, because few readers (unless they are of my vintage!) will have used it or heard of it. Other than acrylic resin, with its many deficiencies, it was the only tooth-coloured restorative material available in the 1960s and 1970s, before resin composite became available. This comprised a fluoro-alumina-silicate glass (similar to that used in early glass ionomer cements), mixed with phosphoric acid. It suffered badly from dissolution in saliva (not a great property for a restorative material!) but released large amounts of fluoride into solution around the restoration. Therefore, although the fillings could be seen to have dissolved, recurrent caries adjacent to the restoration was rare. The arrival of resin composite materials hastened its demise.
The first time that I found the word computer mentioned with regard to restoration survival was in a 1997 publication from Australia.6 In this, the authors stated that all restorative treatment for 100 patients (who had attended continuously for 12 years) was encoded and entered into a computer database, and all data subjected to ‘numerous error checking procedures’. As one might expect as a result of computerization, the number of restorations was greater than in earlier studies (1728 amalgams, 458 resin composites, 399 crowns, 71 gold castings and 275 glass ionomers), and the recording of other factors was facilitated, such as patient age, patient frequency of attendance, change of dentist, dentist's time since graduation. The authors presented median survival times (13.8 years for castings, 26.0 years for crowns, 22.5 years for amalgams, 16.7 years for composites, 13.8 years for castings and too few glass ionomer failures to calculate the median), but identified factors that would not have been possible without computerization, for example, that younger dentists had better results with resin composite than older dentists. In that regard, the majority of composite restorations being placed by the younger dentists, presumably because they had learned these at dental school.
Given the difficulties in keeping a group of patients (with their restorations) together with a researcher and clinician for periods of more than 5 years, Elderton7 realized that large administrative databases were of value. These have the advantage of big numbers, even if factors like the state of the patients' mouths prior to treatment was missing. He examined the records of selected patients who had been examined as part of an Adult Dental Health Survey, finding (perhaps surprisingly) no significant difference in survival of 1-, 2- and 3-surface amalgams and 50% survival of ‘routine’ amalgam and synthetic restorations being 4.5–5.0 years. In the discussion of these results, he intimated that these were on the low end of expectations and ‘add fuel to the argument that redeployment of resources towards prevention should receive greater consideration than at present’. This is a recurring theme, even to the present day.
Other researchers have gleaned useful data using large administrative bases. Among these were:
It can be seen from the foregoing that the analysis of computer databases facilitated the inclusion of large numbers of restorations in the studies, this being light years away from the early efforts of 1969 and subsequent years. I, along with my colleague, Dr Steve Lucarotti, analysed data from the payment claims of NHS dentists in England and Wales, on a dataset that included over 13 million restorations.11 We used Robinson's Rules, as they seemed to have been widely accepted in many other studies, so our results were a study of re-intervention on a restoration, rather than absolute longevity. However, this facilitated comparison between different restorative techniques and materials and the sheer size of the dataset allowed the effect of the restoration on the survival (to extraction) of the restored tooth, which is arguably more important than the survival of the restoration. The principal findings from the extensive analysis of the data, published in the British Dental Journal in 2018 can be summarized as follows11 with regard to preserving the lifespan of the restored tooth:
Studies from 50 years ago began the essential work on restoration survival: recent methods of analysis have built on that and can also assess factors that influence survival. Where do we go from here?
A group of dentists in the Netherlands are an example of what can now be done. Laske and colleagues12 published data from electronic patient files from 67 general dentists. This included 358,548 restorations in 75,556 patients, with the annual failure rate (AFR) varying between 2.3% and 7.9%, mean 4.6% at 10 years, with restorations in molars having higher AFR, and the AFR of composites being 4.4%, amalgam 5.1%, and GI 11.1%. Overall, the 10-year failure rate was 3.8%, but varied between practices (2–5%): age of patient, gender, number of surfaces, operator, tooth type and endodontically treated teeth significantly influenced survival. This is an example to all of us when a group of dentists band together to collect data. This surely could be done in the UK, where there are increasing numbers of large corporates (but, are they more interested in the bottom line than the longevity of restorations placed under their auspices?) and other large groups of dentists, such as Denplan?
Finally, if Allan and others over 50 years ago could produce comparative data on restoration survival 25 years before computers came to help with the task, it is not beyond software manufacturers and/or individual dental practices to produce scaled-down systems (like the one we used for the analysis of the 13 million dataset), which could measure the time to re-intervention of restorations placed in one's own practice. Being able to advise patients on the potential survival of treatments that they might receive is surely the most powerful information that a clinician can present to a patient. And, it's much easier today than it was in 1969!