Dental caries paradigms in diagnosis and diagnostic research

2006 The Authors. Journal compilation
2006 Eur J Oral Sci

Eur J Oral Sci 2006; 114: 263–277 Printed in Singapore. All rights reserved

European Journal of Oral Sciences

Focus

Dental caries paradigms in diagnosis and diagnostic research Baelum V, Heidmann J, Nyvad B. Dental caries paradigms in diagnosis and diagnostic research. Eur J Oral Sci 2006; 114: 263–277.
2006 The Authors. Journal compilation
2006 Eur J Oral Sci In this article, the fundamentals of caries diagnosis are reviewed from the three component perspectives, namely the strategy, the logics, and the tactics. Strategy concerns the objectives of the diagnostic process (i.e. why we diagnose caries). The logics describe how we assemble and evaluate the information collected and how this leads to an assessment of diagnostic value. Finally, tactics are about how we collect the information necessary to arrive at a correct diagnosis. We argue that the hithertodominant essentialistic caries paradigm should be replaced by a nominalistic caries concept. This allows us to circumvent the problem of a lack of a caries gold standard and to proceed in caries-diagnostic research to find the diagnostic methods that result in the best health outcomes for our patients. We also demonstrate the limitations of the medical model when attempting to understand caries diagnosis, and adhere to the Bader & Shugars caries script model. It is concluded that the current caries profile, characterized by lower prevalence and extent, and slower progression, has increased the need for an academic strengthening of the dental curriculum with respect to diagnostic reasoning and clinical decision-making processes.

It is well known that there is a substantial variation between dentists in the way they practice clinical diagnosis and therapeutic decision-making relating to caries (1–7). This variation is usually accepted as reflections of Ôthe art of dentistryÕ (i.e. the natural variation in the best clinical judgment of individual dentists concerning individual patients) (4). However, the existence of such nontrivial variation raises the issue of under- and over-utilization of dental procedures and services (8). If two dentists consistently apply different diagnostic regimens and, as a result, provide different treatment or prevention procedures for similar patients, it must be concluded that one dentist is more efficient than the other (4). In times where patient-initiated litigation is more common than ever before (9), it is important for the dental profession to address the roots of this variation in order to improve clinical decision-making. While accumulated personal clinical experience cannot be denounced in caries diagnosis, it is increasingly necessary to admit the caries-diagnostic process to rational and formal reasoning. Current times are characterized by remarkable declines in the caries experience of many of our populations (10–15), while new and increasingly refined caries-diagnostic methods and techniques continue to be offered to, and recommended for use in, clinical practice (16–25). As late as 2002, concerns were expressed that these early detection technologies may be used to unnecessarily and overaggressively Ôdrill and fillÕ, using Black’s Ôextension for preventionÕ preparations (26).

Vibeke Baelum1, Jens Heidmann2, Bente Nyvad3 1

Department of Community Oral Health and Pediatric Dentistry, 2Section of Computer Technology, and 3Department of Dental Pathology, Operative Dentistry and Endodontics, School of Dentistry, Faculty of Health Sciences, University of Aarhus, Denmark

Vibeke Baelum, Department of Community Oral Health & Pediatric Dentistry, School of Dentistry, University of Aarhus, Vennelyst Boulevard 9, DK 8000 Aarhus C, Denmark Telefax +45–86–136550 E-mail: [email protected] Key words: accuracy; caries; clinical decisionmaking; diagnosis; reliability Accepted for publication May 2006

The word diagnosis has two different senses (27) that must be held apart. On the one hand the diagnosis refers to the disease or state of the patient in question; while on the other hand it refers to the maneuvers the professional goes through to reach a conclusion on the state of the patient (i.e. the diagnostic process). In dentistry, we know what we are looking for when we examine the teeth of a patient: signs and symptoms that we attribute to dental caries. This implies that classical differential diagnosis, which is a critical issue in medicine, is not so dominant in dentistry (1). This has led to the suggestion (1) that the repetitious search process, commonly referred to as Ôcaries diagnosisÕ, would better be described as caries ÔdetectionÕ, a proposition that has been seconded by many (28–31). This article attempts to scrutinize the caries-diagnostic process by dissecting it into its three components: the strategy; the logics; and the tactics (27). The strategy is determined by the objectives of the diagnostic process, which deal with the question of why we want to diagnose caries. The logics of the process are about how to assemble the various pieces of information gathered to arrive at a diagnostic assessment. Finally, the tactics of the diagnostic process are about the procedures we use to obtain the necessary information for arriving at a correct diagnosis. It is our hope that this article will provide dental clinicians and researchers alike with tools enabling them to address systematically the caries-diagnostic component necessary for improved clinical decision-making.

264

Baelum et al.

The strategy What are the objectives of caries diagnosis?

Why do we want to diagnose caries? This question may seem provocatively trite, as we would probably all argue that we want to do so to ensure the best dental health outcome for the patient. However, the means used to achieve this goal vary tremendously, as dental professionals and dental researchers alike do not have a universally shared and common understanding of caries. A recent publication (26) listed some of the many connotations associated with the term ÔcariesÕ, such as the infectious disease itself, a cavity or cavitation, demineralization, a white spot lesion, a brown spot lesion, radiographic caries, histologic caries, infected dentin, affected dentin, stain, etc., and yet other connotations include chemical dissolution of the dental hard tissues (32), a disease process (31, 33), active caries (34–36), arrested caries (36, 37) and inactive caries (34, 38). So many different understandings of dental caries are not helpful when attempting to clarify the caries-diagnostic process and improve clinical decision-making. However, the connotations listed above fall, broadly speaking, into one of two classes: one focuses on the caries disease process; and one focuses on particular caries signs and symptoms. These two approaches reflect two different disease paradigms, namely the essentialistic and the nominalistic (39, 40) (Fig. 1). The apparent co-existence of these two lines of thinking reflect the longstanding problem, in both medical and dental sciences, to define, in a logical manner, what is meant by the term ÔdiseaseÕ (41). Essentialistic vs. nominalistic views on caries

In brief, the essentialistic disease concept holds that diseases exist independently of the criteria we use to describe them (39–41). In essentialism, the disease is conceived of as a sort of engine that converts the disease causes into signs and symptoms of disease (Fig. 1). This viewpoint is implicit in statements such as, ÔDental caries is an infectious disease that manifests itself by…Õ (42), or ÔThe disease, dental caries, attacks the hardest structure in the body…and destroys it just as a cancer may destroy…Õ (43). These statements dictate that a disease

essence exists – dental caries – which, when it attacks a person, causes signs and symptoms such as white spot lesions, cavities, or pain. Essentialism is implicit in colloquial speech (39, 40), and a manifest example of this is the Encyclopedia Britannica definition of caries as a ÔLocalized disease that causes decay and cavities in teethÕ (44). In contrast, the nominalistic disease concept dictates that diseases do not exist independently of their signs and symptoms. The disease name is a label attached to a group of individuals (or teeth or tooth surfaces) who share certain defining characteristics (i.e. the signs and symptoms) (Fig. 1). In the nominalistic view, we choose our caries-diagnostic classification system to reflect biologically sound principles and the most effective management options (34). How essentialism fails

The essentialistic disease concept fails in several important aspects, all illustrated in dental caries: The word ÔcariesÕ originates in Latin and means ÔrotÕ. The diagnosis of caries used, for example by Fauchard (45, 46), in the 18th century, was entirely nominalistic (i.e. a name given to describe the presence of a macroscopic sign, in this case a rotten tooth). However, increased knowledge and new techniques permitted the study of tooth decay in much greater detail, leading us gradually to adopt defining characteristics from other scientific fields, including microbiology (infection), chemistry (dissolution), pathology (hypomineralization), and bio-imaging (subsurface lesion, dentinal caries). This shows how the defining criteria for ÔcariesÕ have changed as a function of scientific development, and how increased knowledge has led to changes in the description of caries. This contradicts the essentialistic idea that dental caries is a disease which exists independently of defining criteria. Notwithstanding, the caries-diagnostic literature holds many statements to the effect that diagnosis is defined as the determination of disease, but not as the determination of the signs and symptoms thereof (28).

The Ôcaries processÕ: disease or cause?

Fig. 1. Essentialistic and nominalistic disease concepts.

Caries is a disease and caries has causes. When asked about the causes of caries we may produce a long list of factors relating to broad concepts, such as biofilm, diet, saliva, behavior, knowledge, values, beliefs, psycho-social resources, economic circumstances, and community characteristics (47–49), and we conclude that caries has a multifactorial etiology (47, 50). We also know that caries lesions arise because the metabolic activities in the microbial deposits covering the tooth surface at any given time (47) produce minute pH fluctuations at the interface between the tooth surface and the microbial deposits. When the resulting episodes of mineral losses or gains are integrated (i.e. accumulated over time), discernible caries lesions may arise (51). This sequence of events is often referred to as the caries pro-

Caries paradigms in diagnosis

cess (47), and implies the existence of a ubiquitously occurring disease process, the signs of which may remain subclinical or may progress to become clinically visible. This is precisely the feature which has led to the statement that this process cannot be prevented but it can be controlled to the extent that a visible caries lesion never develops (33). Thereby, the caries process may be regarded as a natural phenomenon occurring wherever microbial deposits cover a tooth surface (51). Undoubtedly, many will understand the Ôcaries processÕ as the engine that converts the causes of caries into its signs and symptoms, hence apparently supporting an essentialistic disease paradigm. However, this viewpoint fails because what is described as the process in fact represents the causes of caries lesions (47, 49). Therefore, when we refer to Ôthe caries processÕ, we should make explicit that we understand this term as a synonym for the entire complex constellation of component causes that interact to produce caries lesions, whether clinically discernible or not. The caries ÔtruthÕ – a question of scale

Another point of failure of the essentialistic disease concept relates to the difficulties encountered when attempting to distinguish the carious from the normal. The essentialistic view implies that caries exists as a fixed essence, which is independent of signs and symptoms, whereby the distinction between carious and non-carious is necessarily static. Increasingly detailed study of caries lesions has revealed a whole spectrum of lesions that forms a continuum from the barely discernible change at the ultrastructural level to obvious cavitated lesions (33,52). Proponents of the essentialistic view are basically looking for the start of this continuum. The identification of caries therefore depends on the resolution of the available instruments. This effectively precludes a fixed-caries ÔtruthÕ being stated in terms of normal (sound), as opposed to diseased (carious): from a theoretical viewpoint, it is always possible to conceive instruments with even better resolution (i.e. a finer scale of measurement), but this will merely result in changes in the measurements. Thereby, the ÔtruthÕ about caries will depend on the instruments used for observation. Determination of the extent of mineral loss from a tooth surface is akin to searching for the answer to a seemingly trivial question raised by Mandelbrot (53), ÔHow long is the coastline of Great Britain?Õ. It turns out that there is no single answer to the question, no single fixed ÔtruthÕ, because the length of the British coastline increases infinitely as the scale of measurement decreases. It is thus a fact that Mandelbrot (53) considered the length of geographic curves undefined (i.e. without a statement of truth). Applied to the caries lesion this merely shows us that a ÔtruthÕ about the extent of mineral loss from the tooth surface cannot be defined without reference to a particular scale of measurement. It is therefore meaningless to continue to adhere to the common belief in a reference of ÔtruthÕ for caries – the Ôgold standardÕ.

265

How nominalism may help us

No essentialistic caries definition, no matter how elaborate, can answer the question ÔWhat is caries?Õ, because there are no diseases, only sick people (39). However, a nominalistic caries definition by a given set of diagnostic criteria X can answer the question, ÔWhat shall we call the patients, teeth or tooth surfaces that fulfill the set of criteria X?Õ The nominalistic disease concept for caries has two advantages. The first is that it is patient-centered. The health outcome for our patients – rather than some elusive ÔtruthÕ – is what concerns us (39, 54), and the key element in nominalism is to employ a diagnostic classification system that reflects biologically sound principles and the most effective management options (34), while doing as little harm as possible. The other advantage of a nominalistic caries concept is that we may avoid the many discussions about the Ôtrue gold standardÕ defining characteristics for dental caries to use in caries-diagnostic research (55). The caries diagnostic literature is characterized by the use of a host of different Ôgold standardsÕ for caries (for review see (55–57), which merely demonstrates that the approach is not feasible (58). The ÔtruthÕ about a caries process that we cannot observe is, of course, elusive. Adherence to the nominalistic caries concept circumvents the difficulties by turning the problem upside down, merely asking the question, ÔWhich are the defining criteria for caries that will result in the best health outcomes for our patients, given current knowledge about the treatment options?Õ. It follows that the best diagnostic test methods and criteria are those that maximize the health outcome by virtue of their implications for control and management. New diagnostic test methods and criteria must be considered as useless if they do not result in improved health outcomes compared with the standard alternatives. It finally follows that the diagnostic study paradigm for the future makes use of the randomized controlled clinical trial study design to test whether a new diagnostic method or approach actually results in better health outcomes than the traditional one(s) (59). What is carious and what is sound?

Rose (60) eloquently pointed out that the only real truth is that we all have a touch of disease. As shown above, dental caries is no exception. In a caries-diagnostic scenario, where one of the key purposes of the diagnostic activities is to distinguish between the sound and the sick, it is therefore mandatory not only to have a very clear understanding of what is meant by ÔnormalÕ, ÔhealthyÕ or ÔsoundÕ, but also why we think this is so. Several different definitions of ÔnormalÕ may be encountered within the dental and medical research framework (61). The Gaussian definition (62, 63) is based on the assumption that the measurements made in

266

Baelum et al. Table 1 The diagnostic test matrix for a dichotomous test result (T) in the diagnosis of caries C Known caries status ¼ gold standard

Test result T+ T)

Fig. 2. The Gaussian and percentile definitions of ‘normal’ as opposed to ‘abnormal’.

a group of normal or healthy persons – the reference population – will fit a Gaussian distribution (Fig. 2). Thereby, subjects with measurements outside the range defined by the mean ± 1.96 standard deviations are defined as ÔabnormalÕ. While the closely related percentile definition (Fig. 2), occasionally used to define high-risk groups for caries (64), may circumvent the problem with the distributional assumption, both definitions have drawbacks because they imply a fixed frequency for any disease associated with the test parameter. It is also just a question of carrying out a sufficient number of independent tests before a person will be deemed sick (65–67). These distributional definitions are not widely used in caries-diagnostic research, but can be encountered (68) and may gain increasing popularity as quantitative diagnostic methods continue to be developed. The risk-factor definition defines the normal range for a disease-associated parameter by the levels that are compatible with no additional risk of disease development or progression. The use of baseline caries status to define high- and low-risk groups for caries (64) may be seen as an example. Moreover, salivary flow rate is inversely related to the caries incidence, but the salivary flow rate does not seem to exert an effect until it is below a certain threshold value (69). Accordingly, a threshold has been set, for the unstimulated flow rate, at < 0.1 ml min)1 for the diagnosis of hyposalivation as a caries-risk factor (70, 71). However, diurnal variations in the salivary flow rate (72), and other influences, may make the relevant threshold value vary across individuals (69, 71). The diagnostic approach to the definition of ÔnormalÕ holds that this is defined by the range of measurements, below which disease, with known probability, is absent. Disease presence is, conversely, defined by the range of measurements beyond which disease is present. This definition necessitates a Ôgold standardÕ for disease, leading to the calculation of accuracy parameters (sensitivity and specificity) and predictive values (Table 1), and is the definition that has hitherto dominated caries research (56, 57, 73).

Caries present

Caries absent

True positive (TP) False negative (FN)

False positive (FP) True negative (TN)

Sensitivity ¼ probability of positive test result, given that caries is present ¼ TP/(TP+FN). Specificity ¼ probability of negative test result, given that caries is not present ¼ TN/(FP+TN). Bayes’ theorem: PV+ ¼ probability of caries, given positive test result ¼ Prevalence · Sensitivity/[Prevalence · Sensitivity + (1–Prevalence) · (1–Specificity)]. PV– ¼ probability of caries not present, given negative test result ¼ (1–Prevalence) · Specificity/[ (1–Prevalence) · Specificity + Prevalence · (1–Sensitivity)].

Finally, a therapeutic approach to the definition of ÔnormalÕ exists. The upper limit for normal test values is the level beyond which specific treatments for the disease have been shown to do more good than harm. This definition is intuitively appealing because it is focused to prognostic considerations (59), does not necessitate a gold standard, and avoids labeling persons as diseased unless they are actually going to be treated. Sackett & Haynes (65) use hypertension as an example to illustrate how brilliantly the therapeutic definition may work. In the early 1960s there was evidence for a beneficial effect of antihypertensive treatment only when the diastolic blood pressure exceeded 130 mmHg. In the latter part of the 1960s, additional randomized trials provided evidence that treatment instituted already at 115 mmHg was beneficial, and in 1970 there was further evidence leading to yet a lowering of the threshold value to 105 mmHg. Contemporary definitions of hypertension have gone even further and take into account not only the blood pressure per se but also other patient factors, such as age, gender, cholesterol levels, blood sugar levels, and smoking (65). The very short summary of this discussion is that at least five different approaches to the definition of ÔnormalÕ, ÔsoundÕ or ÔhealthyÕ, as opposed to diseased, may be identified, one of which – the diagnostic approach – has completely dominated caries-diagnostic research. Unfortunately, this approach may have led caries-diagnostic research into a blind alley. We suggest that the only definition of ÔnormalÕ which holds real appeal in the context of dental caries is the therapeutic definition. This is also the only definition that links itself directly to the basic rationale – the strategy – underpinning caries diagnosis. We want to diagnose caries because we believe that in so doing we may bring about a benefit to the patient by means of some form of therapeutic intervention (39). This is precisely why a patient-centered nominalistic caries concept is highly preferable.

Caries paradigms in diagnosis

The logics Diagnostic uncertainty in problem-solving and screening

The foundations for caries-diagnostic testing were laid centuries ago with the observation that people who complained about toothache usually presented cavities in the affected teeth. The statement that people with toothache usually have large cavities in the aching teeth shows a fundamental aspect of the diagnostic process. It yields a probabilistic answer; the word ÔusuallyÕ implies a probability of cavitation somewhere between 0 and 1, although much closer to 1 than to 0. This means that there is some measure of uncertainty connected with a diagnosis, and that the diagnosis could be wrong. However, in this mainly pain-relief era, the ÔtruthÕ of the diagnosis was easily established when the treatment turned out to solve the patient’s problem. Diagnostic uncertainty was therefore not an issue of major concern in the early days of dentistry. Even though the entire clinical spectrum of dental caries has been fairly well understood since the latter part of the 19th century (74), problem solving in the form of pain relief and functional restoration of the severely decayed teeth remained the essential issues in dentistry until a caries decline emerged during the latter part of the 20th century (15). At this point in time, the ideas of earlier detection, earlier intervention, and prevention gradually began to crystallize as desirable and feasible ways forward in the control and management of caries. While dentists had been recommending routine (typically 6-monthly) (75) check-ups for dental health for more than a century (76), adherence to such routine recall schemes had become considerable at the end of the 20th century (77–79), even though the shallow scientific basis (75, 80) continued to prompt criticisms (76, 81). However, this chain of events has changed the content of the dental visit away from its focus on symptom relief to a focus on early disease detection in asymptomatic people. When people go for dental check-ups, they go to be reassured that all is well. Therefore, the dental visit contents have changed from

267

being predominantly curative to having a major screening examination component, and this development is likely to continue as the caries decline trickles into the middle-aged and older sections of our populations. This chain of events has fundamentally altered the underpinnings for the selection of caries-diagnostic methods. Whereas the ÔtruthÕ of the diagnosis in the painrelief era could readily be established by means of the symptom relief resulting from the treatment, such quick answers are no longer possible. Other ways of establishing the caries-diagnostic ÔtruthÕ have therefore become necessary. The ensuing search for the ÔtruthÕ about caries has focused on the demineralization processes prior to cavity formation, prompting a host of new diagnostic developments and new gold standards for the evaluation of caries-diagnostic tests (55, 56). However, the beneficial effects of earlier intervention (e.g. by restoring radiographically recorded dentin lesions with no clinical evidence of cavitation) are based on assumptions rather than evidence (82).

The gold standard paradigm in caries-diagnostic research

We have already suggested that the current thinking in caries diagnosis fails because it depends on the existence of a caries ÔtruthÕ, which is separate from the signs and symptoms, and widespread use of the diagnostic approach to the definition of ÔnormalÕ, ÔsoundÕ, and ÔhealthyÕ. The Phase 1 question (83) of diagnostic research (Table 2) can only be answered on the basis of knowledge of the gold standard truth about caries, making it straightforward to calculate the test accuracy parameters, sensitivity and specificity. These accuracy parameters are often considered universal constants that may be used to assess the predictive values of the test in different population settings using Bayes’ theorem, but this view is erroneous (84). For test results belonging to an ordinal (e.g. radiographic depth) or a continuous (e.g. laser fluorescence) measurement scale, the values for sensitiv-

Table 2 The architecture of diagnostic research

Phase 1

2

3

4

Reference of truth (gold standard) required?

Question

Study base

Do test results in affected patients differ from those in normal subjects? Are patients with certain test results more likely to have the disease? Do test results distinguish persons with and without the disease? Do people who undergo the diagnostic test fare better than untested people?

Subjects with known disease status

Yes

Subjects with known disease status

Yes

Subjects with unknown disease status

Yes

Subjects with unknown disease status

No

Adapted from Sackett & Haynes (65).

268

Baelum et al.

ity and specificity may be calculated for relevant threshold values and plotted as the sensitivity against 1– specificity to obtain the receiver operating characteristics (ROC) curve. The ROC curve is often used to summarize the global performance of the diagnostic test (85–87) by means of calculating the area under the curve (AUC). The AUC can be interpreted as the probability that the diagnostic test value for a randomly selected carious tooth exceeds the diagnostic test value for a randomly selected sound tooth (88). Even so, ROC curves per se have limited implication for clinical diagnostic decision-making. The curves are derived connecting the points defined by the coordinates (1–specificity, sensitivity), obtained when changing the diagnostic threshold options for the phenomenon tested. However, clinical diagnosis can utilize only one of the alternative thresholds, corresponding to one of the defining points (1–specificity, sensitivity), which brings the clinical question back to the traditional trade-off between sensitivity and specificity. This trade-off involves a good deal more than the sheer numerical size of the accuracy parameters. Hence, considerations must be given to the long-term health outcome consequences of false-positive, respectively, false-negative diagnoses, which cannot be assumed to be equal. Moreover, the absolute number of false diagnoses is vitally important, as it is a fact that even a high specificity may result in a number of false-positive diagnoses which exceeds the number of false-negative diagnoses resulting from a modest sensitivity. This contrast is aggravated the lower the caries prevalence. Predictive values are more interesting than accuracy parameters because they provide the answers to the diagnostic Phase 3 question (Table 2) (83): ÔAre persons with positive test results more likely to have the disease than persons with other (negative) test results?Õ. However, in caries-diagnostic research it seems quite commonplace to calculate the predictive values positive (PV+) and predictive values negative (PV)) {Bayes’ theorem: PV+ ¼ probability of caries, given a positive test result ¼ Prevalence · Sensitivity/[Prevalence · Sensitivity + (1–Prevalence) · (1–Specificity)]; and PV– ¼ probability of caries not present, given a negative test result ¼ (1–Prevalence) · Specificity/[(1–Prevalence) · Specificity + Prevalence · (1–Sensitivity)]} on the basis of exactly the same data table that gave rise to the accuracy parameters sensitivity and specificity. This calculation will provide answers to the Phase 2 question of diagnostic research (Table 2), namely whether patients (teeth) with certain test results are more likely to have the disease, but this is no more than a redirection of the Phase 1 question. At best, it shows that the test may show diagnostic promise under profoundly unrealistic conditions. The important distinction between the Phase 2 and the Phase 3 question lies in the issues of spectrum bias and transferability of the diagnostic test (84). In caries-diagnostic research, the Phase 1 and Phase 2 questions about sensitivity, specificity, and predictive values, are typically answered on the basis of extremely artificial populations, for example consisting of sets of extracted teeth with caries lesions of varying depths (58, 89), which have limited relevance for real life clinical dentistry. The disease spec-

trum is heavily biased, and the population investigated typically comprises many more severe lesions and fewer sound teeth than is the real scenario (58). The diagnostic settings used are typically laboratory settings that are very far from those encountered in clinical practice. While data generated in this way can be used to provide hints about the possibility of a test being useful, they cannot be presumed to have bearings for real populations examined under everyday clinical practice conditions. It is our view that it is altogether impossible to provide answers to caries-diagnostic Phase 3 questions under the auspices of the essentialistic gold standard paradigm. While we might establish the clinically and diagnostically relevant populations, we would typically still be prevented from establishing their gold standard caries status, as this, in all probability, would involve invasive (histological or excavation) procedures. The work-up bias that may thus be introduced owing to the difficulties in establishing the gold standard status of the test-negatives will typically result in inflated sensitivity and underestimated specificity (84). Facing such difficulties, it has been quite common, in caries-diagnostic research (42, 55, 56), to select another diagnostic test and to use this test to provide the gold standard result. However, this merely leads to the absurd situation that a diagnostic method, which is the test method in one study, may be the gold standard method in another. No diagnostic method evaluated in this way can be demonstrated to perform better than the comparison method, even if this is indeed the case. From predictive probabilities to diagnosis

For clinical patient management purposes, it is necessary to convert diagnostic disease probabilities into hard decisions of whether or not to act. The ideal diagnostic test either pushes the disease probability towards 1, in which case disease is definitively present, or towards 0, in which case disease is ruled out. However, diagnostic tests are rarely good at both ruling in and ruling out disease (90). As shown in the example presented in Table 3, a negative test result may be used to rule out caries when the caries prevalence is 10%, whereas a positive test result provides little diagnostic guidance. Conversely,

Table 3 The results of applying a diagnostic test, T, with sensitivity 0.80 and specificity 0.80, to a random sample of size 1,000 from a population with a 10% caries (C) prevalence Known caries status

Test result T+ T– Caries distribution

PV+ ¼ 80/260 ¼ 0.31. PV– ¼ 720/740 ¼ 0.97.

C+

C–

Test result distribution

80 20 100

180 720 900

260 740 1,000

Caries paradigms in diagnosis Table 4 The results of applying a diagnostic test, T, with sensitivity 0.80 and specificity 0.80, to a random sample of size 1,000 from a population with a 70% caries (C) prevalence Known caries status

Test result T+ T– Caries distribution

C+

C–

Test result distribution

560 140 700

60 240 300

620 380 1,000

PV+ ¼ 560/620 ¼ 0.90. PV– ¼ 240/380 ¼ 0.63.

Fig. 3. Selection of the threshold value that minimizes the total number of misclassifications.

when the caries prevalence is high (70%) (Table 4) we may use a positive test result to rule in caries, but cannot really trust a negative result. For diagnostic tests with outcomes belonging to ordinal or continuous scales, a threshold value must be selected that is used to distinguish between the presence and absence of disease (Fig. 3). When the threshold chosen is the intersection point between the two probability density curves for sound and carious teeth, as shown in Fig. 3, the total number of misclassifications (false positives + false negatives) is minimized. However, choosing a combined minimum is sensible only if the clinical consequences of a false-positive diagnosis are the same as those of a false-negative diagnosis. This is not likely to be the case. The health outcome consequences of overlooking a frank cavity (i.e. omission or postponement of treatment) are fundamentally different from the consequences of diagnosing a cavity where none existed (i.e. performing operative treatment where nonoperative intervention was the treatment of choice). A simulation study thus clearly showed that high specificity (i.e. a limited number of false-positive diagnoses) is crucial for identifying health gains, whereas the number of overlooked (false-negative) lesions plays a very limited role (82).

269

Logics in caries diagnosis – a way out of the blind alley

The above considerations suggest that in caries-diagnostic research, it may be wise to skip both the Phase 2 and Phase 3 questions and proceed to the crucial Phase 4 question (Table 2), which is so closely related to the nominalistic caries concept, ÔDo patients who undergo the new caries-diagnostic test (and its consequences for treatment) fare better than patients who undergo the standard test?Õ. In the case of caries, where we focus on diagnostic tests and methods for the early detection of asymptomatic disease, the Phase 4 question can only be addressed by means of follow-up of patients randomized to undergo the new diagnostic test or the standard procedure. To the best of our knowledge, such studies have yet to be carried out for caries-diagnostic methods. The essentialistic gold standard tradition underpinning the Phase 1–3 questions (Table 2) is based on only one aspect of the validity concept – criterion validity. However, validity has other aspects (91), such as construct validity and content validity. We suggest that the most feasible way forward is to consider the concept of predictive validity. Hence, a diagnostic test result may be said to have predictive validity if it can predict a subsequent event that is a consequence of the phenomenon being tested. An example of this approach applied to caries-diagnostic research was provided by Nyvad et al. (92), who developed a set of caries-diagnostic criteria (Table 5) based on the nominalistic caries concept. These criteria were used to diagnose caries in a clinical trial of the effect of fluoride toothpaste. As fluoride exerts its cariostatic effect in the presence of actively ongoing caries lesions (32, 93, 94), one of three predictions assessed was that surfaces diagnosed with active noncavitated caries lesions would be more likely to be diagnosed as inactive, or even appear sound at followup, than would surfaces diagnosed as inactive at baseline. Indeed, the predictions could be confirmed and this was one of the findings supporting the validity of these criteria (92). Unlike other caries-diagnostic criteria, the Nyvad criteria are characterized by a direct correspondence between the caries classification and the treatment decisions (Table 5). Nyvad et al. (34, 38, 92, 95), as well as other researchers (96), have long stressed the importance of using our insight in the causal caries processes and the resulting management options to establish and use car-

Table 5 Principal structure of the NYVAD criteria for caries diagnosis Diagnostic category

Suggested intervention

Sound Inactive, non-cavitated Active, non-cavitated Inactive, cavitated Active, cavitated

Adapted from a previous publication (38).

None None Non-operative None (operative) Operative

270

Baelum et al.

ies-diagnostic criteria that are keyed to the treatment strategies. Likewise, Pitts and coworkers (30, 97, 98) have advocated that a link should be established between the caries diagnoses and recommended lesion-management options.

The tactics A wealth of caries-diagnostic options

The caries-diagnostic options available to the dentist do, broadly speaking, fall into one of three groups, which may be designated the Classics, the Newcomers, and the Prospects. The Classics comprise clinical inspection, including fiber optic transillumination (FOTI), and bitewing radiography, including digital radiography (for review see refs 19, 20, 38, 42, 56, 74, 99–103). The Newcomers encompass laser fluorescence (DIAGNOdent), quantitative laser fluorescence (QLF), and the electrical caries monitor (ECM) (16, 22, 23, 25, 104–108). The Prospects are based on techniques such as multiphoton imaging, thermography, infrared fluorescence, optical coherence tomography, ultrasound, and terahertz imaging (18,109). To this long list of methods, we should add the fact that each may cover several different sets of criteria. The group of Classic methods amply illustrates this. Hence, Ismail (74) was recently able to identify 29 different sets of clinical caries-diagnostic criteria that have been reported in the literature between 1966 and 2000. Moreover, within these criteria the actual maneuvers undertaken during the visual-tactile examinations may also vary significantly [e.g. with respect to the use of explorers (110), or in the perceived necessity to clean and dry the teeth prior to the examination (74)]. The tactics employed to date when selecting among the methods, the criteria, and the maneuvers, seem to be governed more by longstanding tradition than by solid evidence or by considerations about the health outcomes of different diagnostic options. As an example, Ismail (74) identified what was termed a ÔgulfÕ between the approaches to caries diagnosis taken in Europe and those taken in the USA. While European caries diagnosis and diagnostic research has focused on the early signs of caries, thereby favoring control of the disease process in its early stages, caries diagnosis in the USA has focused on cavitation and explorer catch, thereby favoring reliability and comparability of findings and operative intervention (74). In any case, the mere co-existence of so many different methods and criteria for the diagnosis of dental caries emphasizes an urgent need to reconsider carefully the questions posed earlier, namely ÔWhat do we understand by caries?Õ, ÔWhy are we concerned with caries?Õ, and ÔWhat can we do about caries?Õ It is thus necessary to re-appraise the many traditions that have been uncritically carried forward over time. The tradition of additional diagnostic yield

Bitewing radiography was introduced as an adjunct to the visual-tactile caries examination as early as 1925 by

Raper (43) using the argument that it allows disclosure of approximal cavities that remain undetected by visualtactile examination alone. This diagnostic yield argument for adding bitewing radiography to the visual-tactile caries examination still operates (111), although the focus has been expanded to include earlier manifestations of approximal surface lesions (112–114), and ÔhiddenÕ caries in occlusal surfaces (115–117). Indeed, with few exceptions (118, 119), studies comparing the diagnostic yield of a visual-tactile caries examination and a radiographic caries examination have concluded that substantially more caries lesions are detected in approximal surfaces (111, 112, 120–123) and occlusal surfaces (124–129) using bitewing radiographs than by visual-tactile examination alone. However, in the land of the blind, the one-eyed is king, and the additional yield of bitewing radiography is apparent only when the clinical caries criteria used have been restricted to concern cavitated lesions only. When the visual-tactile caries examination also comprises the non-cavitated stages of lesion formation, the added value of bitewing radiography is no longer obvious (118, 119). Moreover, in lowcaries populations, such as many contemporary populations, bitewing radiography may only, to a very limited extent, disclose radiographic lesions extending into dentine (113, 114), and the number of false-positive diagnoses may be substantial (130). These observations strongly suggest that the tactics used in contemporary caries diagnosis should be redefined to emphasize full exhaustion of the visual-tactile caries examination before the application of any adjunct caries diagnostic method is considered. Different diagnostic methods tell different stories

It is often ignored that observations made based on a bitewing radiograph are fundamentally different from observations made during the visual-tactile caries examination. Bitewing radiographs may disclose the depth of lesion penetration into the dental hard tissues, but there is far from a one-to-one relationship between the radiographic lesion depth and the clinical presentation. Although it is often assumed that radiographic evidence of caries extending into dentin is synonymous with cavitation, the validity of this assumption is only supported by the results of two small studies, each reporting on fewer than 10 radiographic dentin lesions (131, 132). Some studies suggest a cavitation frequency for radiographic dentin lesions in the range of 75 to 90% (133–136), while several large studies indicate a substantially lower cavitation frequency, ranging from 28 to 65% (137–141). Too much reliance on the observation of a radiographic dentin lesion will thus increase the number of intact surfaces receiving restorations. To add to this problem, a study of the variation among dentists in radiographic caries diagnoses (130) has indicated that dentists generally produce many false-positive diagnoses of dentin caries when examining approximal or occlusal surfaces whose ÔtrueÕ state is either sound or caries confined to the enamel. Overall, the false-positive fraction was 21% for approximal surfaces, and more than 70% of

Caries paradigms in diagnosis

the dentists produced at least three false-positive diagnoses of 16 possible (130). The many adverse effects associated with restorations, as a result of the high frequency of iatrogenic damage to neighboring teeth (142– 144) and limited restoration longevity (145–148), are further links in a chain of unfortunate events that may be broken only if much more stringent criteria are adopted for the use and interpretation of observations on bitewing radiographs. The effect of diagnostic error

A further objection to the uncritical addition of clinical and radiographic observations stems from the fact that both diagnostic methods are error-prone owing to less than perfect inter- and intraexaminer reliability (38, 118, 149). In such error-prone circumstances, the simple addition of the results of different diagnostic methods, or the repetition of methods (e.g. by frequent recall examinations) serve to increase the total number of diagnostic errors made. In caries-diagnostic research it is commonplace to use the kappa value and one of its associated normative scales [e.g. the Landis & Koch scale (150)] to evaluate the reliability or agreement of diagnostic test methods, using arbitrary descriptors such as Ôalmost perfectÕ or ÔfairÕ. Only rarely is further consideration given to the clinical implications of the less-than-perfect reliability, even though this may have considerable implications when diagnoses have irreversible treatment consequences, such as the placement of a filling. The example in Table 6 presents the results that may be obtained when an examiner has examined the same children twice, 2 wk apart. Kappa is 0.68 (95% confidence interval ¼ 0.56; 0.80), and, depending on the scale used, this agreement may be described as ÔsubstantialÕ or ÔgoodÕ (151). However, the use of such standard adjectives does not convey whether the inconsistencies observed have clinical implications, and what these might be. Table 6 shows that the number of cavities found on each examination occasion was very similar (42 and 43 cavities, respectively). However, the crucial problem stems from the fact that only 30 of the 55 cavities diagnosed in at least one of the two examinations concerned the same teeth. Hence, 25 cavities were diagnosed (and potentially filled) that could not be reproduced at a different examination (or by another examiner, in case of interexaminer reliability assessment). Unfortunately, the example provided in Table 6 is

Table 6 Example of data-table arising when assessing the reliability of tooth-based caries diagnoses made at the cavity level Second examination

First examination No cavity Cavity Total

No cavity

Cavity

Total

465 12 477

13 30 43

478 42 520

271

realistic (38, 118, 149), and illustrates that it can indeed be risky to undergo routine dental examinations too often. A famous tonsillectomy example published more than 70 yr ago (152, 153), and later confirmed for other treatments (154), may be used to illustrate further the potential problem involved. The tonsillectomy study was based on a screening of 1000 11-yr-old-children for the need for tonsillectomy. Children who were deemed negative on the first physical examination were re-examined by another physician, and this scheme continued for three rounds. The study clearly demonstrated that undergoing more screening examinations will cause more disease to be found, and the absurd end result was that only 65 of 1000 children would still remain untreated after three routine screening examinations. Diagnosis – probabilistic deduction or pattern recognition?

In medicine, the process leading to a diagnosis for a patient presenting with a set of signs and symptoms of unknown origin is often idealized as a classical probabilistic hypothetico-deductive process (Elstein, quoted in ref. 155). In this process, a set of alternative hypotheses are generated, which might explain the patient’s signs and symptoms. These hypotheses are subsequently tested through examination and other forms of diagnostic tests, and this differential diagnostic process terminates when observations allow one of the alternative hypotheses to be selected as the most probable. When the diagnosis has been clarified, the process of treatment selection begins, although this is usually quite straightforward once a diagnosis has been reached. It is often assumed that this probabilistic differential diagnostic scheme also applies to caries diagnosis (156). Proponents of this viewpoint hold that each patient is evaluated by inspecting the surface of each tooth present for the signs of dental hard tissue pathology, followed by differential diagnostic considerations (e.g. dental fluorosis, enamel opacities), leading to a decision concerning the probability of caries, a caries diagnosis (156). However, as previously alluded to, the typical problem in dental diagnosis it not Ôwhat the patient hasÕ but Ôwhether the patient has cariesÕ, and there are many indications that the differential diagnostic model cannot be transferred to dentistry (1, 155). Where the physician is primarily oriented towards differential diagnosis and to a lesser extent to treatment planning, the reverse seems to apply to dentistry, where treatment planning considerations outweigh differential diagnostic issues. This illustrates the fundamental logical difference between the classical hypothetico-deductive differential diagnostic process invoked when examining a sick patient and the activities carried out in routine check-up examinations (67). Whereas the classical differential diagnostic process is oriented towards the solution of an existing problem, the routine dental screening examination is too cumbersome to be completed de novo for each single tooth surface in each single patient, and dentists therefore resort to rather different tactics (1), in the form of pattern recognition.

272

Baelum et al.

Pattern recognition: the caries script model Even though the scientific evidence about the tactics used by dentists in caries diagnosis and treatment planning remains relatively scant (157), an intuitively very appealing model has been suggested by Bader & Shugars (1). They argue that caries-diagnostic decisions are determined by means of a process of pattern recognition using an inventory of different caries scripts. Caries scripts are mental images, highly summarized versions of the dentist’s cumulative experience with similar clinical presentations, and they implicitly incorporate the relevant differential diagnostic considerations. Caries scripts are directly linked to a treatment decision (e.g. Ôthis-clinical-picture-demands-this-interventionÕ) (158). The caries script model is in line with studies implicating pattern recognition as a prominent feature in the diagnosis and treatment planning process (158, 159), just as pattern recognition is known from the field of medical diagnosis (160). The dentist is probably not even fully aware of all the details of this complex web of cues and signs associated with a lesion judged to need a filling (159), which may explain why there is a notable difference between the treatment thresholds stated by dentists and the treatments actually carried out (159, 161). The caries scripts may be short and succinct, based on only a few prominent details, such as Ôexplorer catch in the fissure system of a permanent molarÕ or Ôradiographic lesion extending into dentinÕ, or they may comprise a detailed scenario, such as (1): ‘a two-surface amalgam restoration in a molar exhibiting no radiological evidence of caries, but having several microfractures of the occlusal margin that have resulted in shallow ditching, an adult patient with moderate plaque accumulation, a cavitated lesion elsewhere in the mouth, and several amalgam restorations, not placed by the examining dentist, which have imperfect margins’. As the above scenario implies, a decision for intervention does not require caries to be evident, but may be made merely based on the expectation of future caries development. Scripts are taught at dental school

Anyone who has witnessed practitioners discussing differences in treatment decisions, or the relative merits of different treatment options, will have noted the dominance of arguments referring to personal experience or preference, illustrating how substantially different their caries scripts may be. However, this is a natural consequence of the way dentistry is taught. Dentistry is taught under the auspices of the expert clinicians (162) who, in the absence of more formal academic training in the issues involved, resort to reliance on personal experience and preference, and seek to convey their own views and biases to the students. The dental students may thereby often be trained by inconsistent or even contradictory expert clinicians (162). Even so, these conflicts are usually explained to students as manifestations of the Ôart of dentistryÕ (i.e. the best clinical judgment by the individual

dentist concerning the individual patient under the particular circumstances). Dental students are not encouraged to question this argument and have no option but to try to understand the particulars of each clinical situation and to incorporate these in their own mental inventory of clinical scripts. It is no wonder that many dentists value expert opinions and clinical experience much higher than scientific evidence and evidence-based practice guidelines. Are caries scripts amenable to change?

Caries scripts may be influenced by dentist characteristics such as age and experience, skill and diligence, and tolerance for uncertainty and knowledge, as well as by dentist biases concerning the perceived utility of restorations, treatment preferences, diagnostic techniques used and experience with outliers and, finally, by practice characteristics such as how busy it is, size, delivery system, equipment, guidelines, and personnel (1, 163– 169). While experiential influences on caries scripts are the best documented (163, 164, 169–171), research indicates that it is also possible to stimulate dentists actively to re-interpret the existing salient factors which determine their caries scripts. Heidmann and coworkers (172) observed a change in the restorative treatment criteria used by dentists in the Danish Child Dental Health Services, such that only 18% of the restorations inserted in 1978–79 would have been inserted had the 1984–85 criteria been employed. This rather dramatic change in the course of only 6 yr was attributed to a combination of experiences (poor outcomes of restorative treatment) and new scientific evidence for slow caries progression. Corroborating these results, the treatment criteria for caries employed by dentists in the Public Dental Service in Oslo, Norway, were noted to change over a 17-yr period to the extent that only 16% of the fillings made in 1979 would have been inserted using the criteria employed in 1996 (173). The tentative explanations for this change included an in-service lecture programme on the prognosis of approximal caries lesions coupled with increasing numbers of patients per dentist, demands for increasing productivity, and more delayed or postponed recall appointments. Taken together, these factors seem to have increasingly pushed dentists to adopt a let-us-observe attitude and, not least, to tolerate uncertainty (173), an interpretation that is corroborated by the results of other studies (174). While it may seem quite paradoxical that learning about uncertainty will enhance consistency, there are analogous examples from other branches of dentistry illustrating that clinical experience and expertise does not guarantee the most consistent diagnoses. In a study of the reproducibility of pocket depth and clinical attachment level measurements (175), it was observed that the experienced periodontist and the prosthodontist performed worse than the periodontal epidemiologist. This contrast was interpreted as resulting from the periodontists tending not only to Ôread too muchÕ into their examinations, but to do so somewhat inconsistently (175).

Caries paradigms in diagnosis Evidence-based caries scripts are needed

In the days when caries ravaged our populations, dental clinicians were faced with the insuperable task of attempting to keep pace with the rate of development of possibly pain-inflicting caries cavities. Overlooking a deep lesion would in all probability result in the patient soon presenting again in the dental office, this time in pain, which perhaps explains why many dental clinicians remain haunted by the ghost of overlooking caries lesions. Conversely, surgical treatment of a non-cavitated lesion was not seen as a major error, because such lesions were invariably progressing, making it just a matter of short time before the lesion would have become cavitated anyway. However, those days are long gone. The caries problem has diminished tremendously in most European and US populations, and in a majority of settings, it is no longer a major problem to keep pace with the rate of caries lesion development and progression. Better insight into and public awareness of the causes of caries, the use of fluoridated toothpastes and topical fluorides in caries control, and improved material and social circumstances (176, 177), have contributed to a marked and continued decline in caries prevalence and extent, and to a slower progression of caries. In most western populations, the reality is that there is a much greater risk of causing adverse health outcomes by means of overzealous operative intervention than by overlooking caries lesions. The public has become increasingly aware of the importance of dental check-up visits, and major sections of the populations have adopted regular dental attendance patterns. This means that even though a lesion may be overlooked at one appointment, there will be a next appointment during which the lesion may be found. Moreover, the risk that the lesion, in the meantime, progresses to the extent that a different treatment modality is indicated (e.g. operative rather than non-operative) is, at worst, moderate. Translated to the arena of caries diagnosis, this means that the consequences of diagnostic errors have changed from the false-negative (i.e. the lesion overlooked) being the major error, to the false-positive (i.e. overtreatment) being the key problem. The dental teaching institutions and the dental educators, whether clinical ÔmastersÕ or academics, carry a major responsibility for improving clinical decisionmaking relating to caries. There is room for much improvement in the understanding of the complexity of decision-making related to caries diagnosis and treatment planning, and this review has indicated some of the critical points. A final warning must be issued. The greater the caries decline in the populations we serve, the more aberrant will be the patients seen in our dental schools. This means that the caries scripts taught to dental students may be rather more biased towards mechanical and technical intervention than would have been the case had they been taught dentistry in free-living populations. The implications for curriculum-planning should be obvious, and the importance cannot be

273

overestimated of students being provided with a solid evidence base on caries-diagnostic strategies and biologically sound management options to counterbalance the technical and restorative focus of traditional dentistry.

Concluding remarks In this article we have presented a new approach to caries diagnosis and caries-diagnostic research based on the following principles: (i) (ii)

(iii)

(iv)

(v)

A nominalistic caries concept: caries is defined by its signs and symptoms. A therapeutic approach to the definition of ÔsoundÕ and ÔcariousÕ. The relevant caries diagnostic categories are those that reflect biologically sound management options, and result in the best health outcomes for our patients. The best caries-diagnostic method and criteria are those that reflect the best management options. Contemporary caries-diagnostic methods should therefore allow for a distinction between cavitated and non-cavitated, and active and inactive, caries lesions. A reassessment of the validity of the additional diagnostic yield concept. Adding methods means adding errors The caries script model for caries-related clinical decision-making. In order to improve clinical decision-making, we must understand the decision-making process and the salient factors influencing decisions.

If these principles are adhered to, we may circumvent the following problems: • different understandings of dental caries; • the unfortunate use of gold standards for caries; and • the co-existence of a multitude of caries diagnostic criteria and methods.

References 1. Bader JD, Shugars DA. What do we know about how dentists make caries-related treatment decisions? Community Dent Oral Epidemiol 1997; 25: 97–103. 2. Shugars DA, Bader JD. Cost implications of differences in dentists’ restorative treatment decisions. J Public Health Dent 1996; 56: 219–222. 3. Bader JD, Shugars DA. Variation, treatment outcomes, and practice guidelines in dental practice. J Dent Educ 1995; 59: 61–95. 4. Bader JD, Shugars DA. Variation in dentists’ clinical decisions. J Public Health Dent 1995; 55: 181–188. 5. Bader JD, Shugars DA, McClure FE. Comparison of restorative treatment recommendations based on patients and patient simulations. Oper Dent 1994; 19: 20–25. 6. Lewis DW, Kay EJ, Main PA, Pharaoh MG, Csima A. Dentists’ variability in restorative decisions, microscopic and radiographic caries depth. Community Dent Oral Epidemiol 1996; 24: 106–111.

274

Baelum et al.

7. Bader JD, Shugars DA. Agreement among dentists’ recommendations for restorative treatment. J Dent Res 1993; 72: 891–896. 8. Hartshorne J, Hasegawa TK. Overservicing in dental practice – ethical perspectives. J S Afr Dent Assoc 2003; 58: 364–369. 9. Renshaw J. After the first 125 years of the BDJ where might clinical dentistry be heading? Br Dent J 2005; 199: 331–337. 10. Marthaler TM. Changes in dental caries 1953–2003. Caries Res 2004; 38: 173–181. 11. Marthaler TM, O’Mullane DM, Vrbic V. The prevalence of dental caries in Europe 1990–95. Caries Res 1996; 30: 237–255. 12. Poulsen S, Scheutz F. Dental caries in Danish children and adolescents 1988–97. Community Dent Health 1999; 16: 166– 170. 13. Poulsen S. Dental caries in Danish children and adolescents 1988–94. Community Dent Oral Epidemiol 1996; 24: 282–285. 14. Hugoson A, Koch G, Slotte C, Bergendal T, Thorstensson B, Thorstensson H. Caries prevalence and distribution in 20–80-year-olds in Jo¨nko¨ping, Sweden, in 1973, 1983 and 1993. Community Dent Oral Epidemiol 2000; 28: 90–96. 15. Glass RL. The first international conference on the declining prevalence of dental caries. J Dent Res 1982; 61: 1304–1360. 16. Van Der Veen MH, De Josselin De Jong E. Application of quantitative light-induced fluorescence for assessing early caries lesions. Monogr Oral Sci 2000; 17: 144–162. 17. Jovanovski V, Lynch E. Analysis of the morphology of oral structures from 3-D co-ordinate data. Monogr Oral Sci 2000; 17: 73–129. 18. Colston BWJr , Everett MJ, Sathyam US, Dasilva LB, Otis LL. Imaging of the oral cavity using optical coherence tomography. Monogr Oral Sci 2000; 17: 32–55. 19. Analoui M, Stookey GK. Direct digital radiography for caries detection and analysis. Monogr Oral Sci 2000; 17: 1–19. 20. Wenzel A. Bitewing and digital bitewing radiography for detection of caries lesions. J Dent Res 2004; 83: C72–C75. 21. Wenzel A. Digital radiography and caries diagnosis. Dentomaxillofac Radiol 1998; 27: 3–11. 22. Lussi A, Hibst R, Paulus R. DIAGNOdent: an optical method for caries detection. J Dent Res 2004; 83: C80–C83. 23. Haak R, Wicht MJ. Caries detection and quantification with DIAGNOdent: prospects for occlusal and root caries? Int J Comput Dent 2004; 7: 347–358. 24. Lussi A, Firestone A, Schoenberg V, Hotz P, Stich H. In vivo diagnosis of fissure caries using a new electrical resistance monitor. Caries Res 1995; 29: 81–87. 25. Longbottom C, Huysmans M-CDNJM. Electrical measurements for use in caries clinical trials. J Dent Res 2004; 83: C76– C79. 26. Young DA. New caries detection technologies and modern caries management: merging the strategies. Gen Dent 2002; 50: 320–331. 27. Murphy EA. The diagnostic process. In: Murphy EA, ed. The logic of medicine. Baltimore, MD: Johns Hopkins University Press, 1997; 281–313. 28. Ten Bosch JJ, Angmar-Mansson B. Characterization and validation of diagnostic methods. Monogr Oral Sci 2000; 17: 174–189. 29. Nyvad B. Diagnosis versus detection of caries. Caries Res 2004; 38: 192–198. 30. Pitts NB. Are we ready to move from operative to nonoperative/preventive treatment of dental caries in clinical practice? Caries Res 2004; 38: 294–304. 31. Pitts NB. Modern concepts of caries measurement. J Dent Res 2004; 83: C43–C47. 32. Ten Cate JM, Larsen MJ, Pearce EIF, Fejerskov O. Chemical interactions between the tooth and oral fluids. In: Fejerskov O, Kidd E, eds. Dental caries: the disease and its clinical management. Oxford: Blackwell Munksgaard, 2003; 49–69. 33. Kidd EAM, Fejerskov O. What constitutes dental caries? Histopathology of carious enamel and dentin related to the action of cariogenic biofilms. J Dent Res 2004; 83: C35– C38.

34. Nyvad B, Fejerskov O. Active root surface caries converted into inactive caries as a response to oral hygiene. Scand J Dent Res 1986; 94: 281–284. 35. Spencer AJ, Lewis J. The relationships between prevalence of caries, percentage with active caries, and caries severity. Aust Dent J 1987; 32: 441–445. 36. Miller WA, Massler M. Permeability and staining of active and arrested lesions in dentine. Br Dent J 1962; 112: 187–197. 37. Wilding RJ, Solomon CS. Arrested caries: a review of the repair potential of the pulp-dentine. J Dent Assoc S Afr 1996; 51: 828–833. 38. Nyvad B, MacHiulskiene V, Baelum V. Reliability of a new caries diagnostic system differentiating between active and inactive caries lesions. Caries Res 1999; 33: 252–260. 39. Wulff HR. What is understood by a disease entity? J R Coll Phys Lond 1979; 13: 219–220. 40. Scadding JG. Essentialism and nominalism in medicine: logic of diagnosis in disease terminology. Lancet 1996; 348: 594– 596. 41. Scadding JG. Diagnosis: the clinician and the computer. Lancet 1967; 2: 877–882. 42. Ismail AI. Clinical diagnosis of precavitated carious lesions. Community Dent Oral Epidemiol 1997; 25: 13–23. 43. Raper HR. Practical clinical preventive dentistry based upon periodic roentgen-ray examinations. J Am Dent Assoc 1925: Sept., 1084–1100. 44. Britannica Concise Encyclopedia. Caries. Encyclopædia Britannica. Premium Service 2006, Chicago, IL: Encyclopaedia Britannica, http://www.britannica.com/ebc/ article-9359826.Accessed January 20. 45. Fauchard P. The surgeon dentist or treatise on the teeth, Vol. II. New York: Milford House Inc., 1969. 46. Fauchard P. The surgeon dentist or treatise on the teeth, Vol. I. New York: Milford House Inc., 1969. 47. Fejerskov O, Manji F. Reactor paper: risk assessment in dental caries. In: Bader JD, eds. Risk assessment in dentistry. Chapel Hill: University of North Carolina Dental Ecology, 1990; 215–217. 48. Holst D, Schuller AA, Aleksejuniene J, Eriksen HM. Caries in populations a theoretical, causal approach. Eur J Oral Sci 2001; 109: 143–148. 49. Holst D. Causes and prevention of dental caries. a perspective on cases and incidence. Oral Health Prev Dent 2005; 3: 9–14. 50. Fejerskov O. Changing paradigms in concepts on dental caries: consequences for oral health care. Caries Res 2004; 38: 182–191. 51. Manji F, Fejerskov O, Nagelkerke NJD, Baelum V. A random effects model for some epidemiological features of dental caries. Community Dent Oral Epidemiol 1991; 19: 324– 328. 52. Featherstone JDB. The continuum of dental caries – evidence for a dynamic disease process. J Dent Res 2004; 83: C39–C42. 53. Mandelbrot B. How long is the coast of Britain? Statistical self-similarity and fractional dimension. Science 1967; 156: 636–638. 54. Baelum V, Lopez R. Defining and classifying periodontitis: need for a paradigm shift? Eur J Oral Sci 2003; 111: 2–6. 55. Huysmans M-CDNJM, Longbottom C. The challenges of validating diagnostic methods and selecting appropriate gold standards. J Dent Res 2004; 83: C48–C52. 56. Hintze H. Radiography for the detection of dental caries lesions. Aarhus: University of Aarhus, 2004. 57. Wenzel A, Hintze H. The choice of gold standard for evaluating tests for caries diagnosis. Dentomaxillofac Radiol 1999; 28: 132–136. 58. Bader JD, Shugars DA, Bonito AJ. A systematic review of the performance of methods for identifying carious lesions. Public Health Dent 2002; 62: 201–213. 59. Lijmer JG, Bossuyt PM. Diagnostic testing and prognosis: the randomised controlled trial in diagnostic research. In: Knottnerus JA, eds. The evidence base of clinical diagnosis. London: BMJ Books, 2002; 61–80.

Caries paradigms in diagnosis 60. Rose G. The strategy of preventive medicine. Oxford: Oxford University Press, 1992. 61. Murphy EA. The normal, and the perils of the sylleptic argument. Perspect Biol Med 1972; 15: 566–582. 62. Haffajee AD, Socransky SS, Dzink JL, Taubman MA, Ebersole JL, Smith DJ. Clinical, microbiological and immunological features of subjects with destructive periodontal diseases. J Clin Periodontol 1988; 15: 240–246. 63. Haffajee AD, Socransky SS, Dzink JL, Taubman MA, Ebersole JL. Clinical, microbiological and immunological features of subjects with refractory periodontal diseases. J Clin Periodontol 1988; 15: 390–398. 64. Hausen H. Caries prediction – state of the art. Community Dent Oral Epidemiol 1997; 25: 87–96. 65. Sackett DL, Haynes RB. The architecture of diagnostic research. In: Knottnerus JA, eds. The evidence base of clinical diagnosis. London: BMJ Books, 2002; 19–38. 66. Murphy EA. Disease and normality. In: Murphy EA, ed. The logic of medicine. Baltimore, MD: Johns’ Hopkins University Press, 1997; 137–158. 67. Murphy EA. The diagnostic process and multiple screening techniques. Methods Inf Med 1972; 11: 8–14. 68. Shinohara T, Takase Y, Amagai T, Haruyama C, Igarashi A, Kukidome N, Kato J, Hiray Y. Criteria for a diagnosis of caries through the DIAGNOdent. Photomed Laser Surg 2006; 24: 50–58. 69. Tenovuo J. Salivary parameters of relevance for assessing caries activity in individuals and populations. Community Dent Oral Epidemiol 1997; 25: 82–86. 70. Sreebny L. Saliva–salivary gland hypofunction (SGH). FDI Working Group 10. J Dent Assoc S Afr 1992; 47: 498–501. 71. Bardow A, Nyvad B, Nauntofte B. Relationships between medical intake, complaints of dry mouth, salivary flow rate and composition, and the rate of tooth demineralization in situ. Arch Oral Biol 2001; 46: 413–423. 72. Flink H, Tegelberg A˚, Lagerlo¨f F. Influence of the time of measurement of unstimulated human whole saliva on the diagnosis of hyposalivation. Arch Oral Biol 2005; 50: 553–559. 73. Wenzel A, Verdonschot EH. Some considerations in the evaluation of diagnostic tests in dentistry. Dentomaxillofac Radiol 1994; 23: 179–182. 74. Ismail AI. Visual and visuo-tactile detection of dental caries. J Dent Res 2004; 83: C56–C66. 75. NHS National Institute For Clinical Excellence. Dental recall. Recall interval between routine dental examinations. London: National Institute for Clinical Excellence, 2004. 76. Sheiham A. How often should people go for routine dental recalls? Community Dent Health 2004; 21: 257–259. 77. Hayward RA, Meetz HK, Shapiro MF, Freeman HE. Utilization of dental services: 1986 patterns and trends. J Public Health Dent 1989; 49: 147–152. 78. Kronstro¨m M, Palmqvist S, So¨derfeldt B, Vigild M. Utilization of dental health services among middle-aged people in Sweden and Denmark. Acta Odontol Scand 2002; 60: 276–280. 79. Petersen PE, Kjoller M, Christensen LB, Krustrup U. Changing dentate status of adults, use of dental health services, and achievement of national dental health goals in Denmark by the year 2000. J Public Health Dent 2004; 64: 127–135. 80. Beirne P, Forgie A, Clarkson JE, Worthington HV. Recall intervals for oral health in primary care patients (Review). The Cochrane Database of Systematic Reviews 2005; Issue 2: 000–000. Art no. CD004346.pub2. DOI: 10.1002/ 14651858.CD004346.pub2. 81. Sheiham A. Is there a scientific basis for six-monthly dental examinations? Lancet 1977; 2: 442–444. 82. Downer MC, Moles DR. Health gain from restorative dental treatment evaluated by computer simulation. Community Dent Health 1998; 15: 32–39. 83. Sackett DL, Haynes RB. The architecture of diagnostic research. Br Med J 2002; 324: 539–541. 84. Guggenmoos-Holzmann I, Van Houwelingen HC. The (in)validity of sensitivity and specificity. Stat Med 2000; 19: 1783–1792.

275

85. Hintze H, Christoffersen L, Wenzel A. In vitro comparison of Kodak Ultra-speed, Ektaspeed, and Ektaspeed Plus, and Agfa M2 Comfort dental x-ray films for the detection of caries. Oral Surg Oral Med Oral Pathol Oral Radiol Endod 1996; 81: 240–244. 86. Hintze H, Wenzel A, Jones C. In vitro comparison of D- and E-speed film radiography, RVG, and Visualix digital radiography for the detection of enamel approximal and dentinal occlusal caries lesions. Caries Res 1994; 28: 363–367. 87. Hintze H, Wenzel A. Influence of the validation methods on diagnostic accuracy for caries. A comparison of six digital and two conventional radiographic systems. Dentomaxillofac Radiol 2002; 31: 44–49. 88. Lee WC. Probabilistic analysis of global performances of diagnostic tests: interpreting the Lorenz curve-based summary measures. Stat Med 1999; 18: 455–471. 89. Bader JD, Shugars DA, Bonito AJ. Systematic reviews of selected caries diagnostic and management methods. J Dent Educ 2001; 65: 960–968. 90. Verdonschot EH, Bronkhorst EM, Burgersdijk RCW, Ko¨nig KG, Schaeken MJM, Truin GJ. Performance of some diagnostic systems in examinations for small occlusal carious lesions. Caries Res 1992; 26: 59–64. 91. Last JM. A dictionary of epidemiology. Oxford: Oxford University Press, 2001. 92. Nyvad B, MacHiulskiene V, Baelum V. Construct and predictive validity of clinical caries diagnostic criteria assessing lesion activity. J Dent Res 2003; 82: 117–122. 93. Ellwood R, Fejerskov O. Clinical use of fluoride. In: Fejerskov O, Kidd E, eds. Dental caries: the disease and its clinical management. Oxford: Blackwell Munksgaard, 2003; 189–222. 94. Fejerskov O, Thylstrup A, Larsen MJ. Rational use of fluorides in caries prevention. A concept based on possible cariostatic mechanisms. Acta Odontol Scand 1981; 39: 241–249. 95. Nyvad B, Fejerskov O. Assessing the stage of caries lesion activity on the basis of clinical and microbiological examination. Community Dent Oral Epidemiol 1997; 25: 69–75. 96. Bader JD, Brown JP. Dilemmas in caries diagnosis. J Am Dent Assoc 1993; 124: 48–50. 97. Pitts NB, Longbottom C. Preventive care advised (PCA)/ operative care advised (OCA) – categorising caries by the management option. Community Dent Oral Epidemiol 1995; 23: 55–59. 98. Pitts NB. Diagnostic tools and measurements – impact on appropriate care. Community Dent Oral Epidemiol 1997; 25: 24–35. 99. International Caries Detection and Assessment System (ICDAS) Coordinating Committee. Rationale and evidence for the International Caries Detection and Assessment System (ICDAS II) (Draft). Indianapolis: Indiana Conference, 2005; 1–43. 100. International Caries Detection and Assessment System (ICDAS) Coordinating Committee. Criteria manual (draft). International Caries Detection and Assessment System (ICDAS II). Indianapolis: Indiana Conference, 2005; 1–31. 101. Pitts NB. ÔICDASÕ – an international system for caries detection and assessment being developed to facilitate caries epidemiology, research and appropriate management. Community Dent Health 2004; 21: 193–198. 102. Ismail AI, Sohn W. A systematic review of clinical diagnostic criteria of early childhood caries. J Public Health Dent 1999; 59: 171–191. 103. Ekstrand KR. Improving clinical visual detection – potentials for caries clinical trials. J Dent Res 2004; 83: C67–C71. 104. Pretty IA, Maupome´ G. A closer look at diagnosis in clinical dental practice: Part 5. Emerging technologies for caries detection and diagnosis. J Can Dent Assoc 2004; 70: 540–540i. 105. Tranæus S, Shi X-Q, Angmar-Ma˚nsson B. Caries risk assessment: methods available to clinicians for caries detection. Community Dent Oral Epidemiol 2005; 33: 265–273. 106. Bader JD, Shugars DA. A systematic review of the performance of a laser fluorescence device for detecting caries. J Am Dent Assoc 2004; 135: 1413–1426.

276

Baelum et al.

107. Ku¨hnish J, Heinrich-Weltzien R. Quantitative light-induced fluorescence (QLF) – a literature review. Int J Comput Dent 2004; 7: 325–338. 108. Stookey GK. Optical methods – quantitative light fluorescence. J Dent Res 2004; 83: C84–C88. 109. Hall A, Girkin JM. A review of potential new diagnostic modalities for caries lesions. J Dent Res 2004; 83: C89–C94. 110. Hamilton JC, Stookey G. Should a dental explorer be used to probe suspected carious lesions? J Am Dent Assoc 2005; 136: 1526–1532. 111. Bloemendal E, De Wet HCW, Bouter LM. The value of bitewing radiographs in epidemiological caries research: a systematic review of the literature. J Dent 2004; 32: 255–264. 112. Kidd EAM, Pitts NB. A reappraisal of the value of the bitewing radiograph in the diagnosis of posterior approximal caries. Br Dent J 1990; 169: 195–200. 113. Hintze H, Wenzel A. Clinically undetected dental caries assessed by bitewing screening in children with little caries experience. Dentomaxillofac Radiol 1994; 23: 19–23. 114. Hintze H. Screening with conventional and digital bite-wing radiography compared to clinical examination alone for caries detection in low-risk children. Caries Res 1993; 27: 499–504. 115. Weerheijm KL, Kidd EAM, Groen HJ. The effect of fluoridation on the occurrence of hidden caries in clinically sound occlusal surfaces. Caries Res 1997; 31: 30–34. 116. Weerheijm KL, Gruythuysen RJM, Van Amerongen WE. Prevalence of hidden caries. J Dent Child 1992; 59: 408–412. 117. Sawle RF, Andlaw RJ. Has occlusal caries become more difficult to diagnose? Br Dent J 1988; 164: 209–211. 118. MacHiulskiene V, Nyvad B, Baelum V. A comparison of clinical and radiographic caries diagnoses in posterior teeth of 12-year-old Lithuanian children. Caries Res 1999; 33: 340–348. 119. MacHiulskiene V, Nyvad B, Baelum V. Comparison of diagnostic yields of clinical and radiographic caries examinations in children of different age. Eur J Paediatr Dent 2004; 5: 157–162. 120. Poorterman JHG, Aartman IH, Kalsbeek H. Underestimation of the prevalence of approximal caries and inadequate restorations in a clinical epidemiological study. Community Dent Oral Epidemiol 1999; 27: 331–337. 121. Hopcraft MS, Morgan MV. Comparison of radiographic and clinical diagnosis of approximal and occlusal dental caries in a young adult population. Community Dent Oral Epidemiol 2005; 33: 212–218. 122. De Vries HCB, Ruiken HMHM, Ko¨nig KG, Van’t Hof MA. Radiographic versus clinical diagnosis of approximal carious lesions. Caries Res 1990; 24: 364–370. 123. Llena-Puy C, Forner L. A clinical and radiographic comparison of caries diagnosed in approximal surfaces of posterior teeth in a low-risk population of 14-year-old children. Oral Health Prev Dent 2005; 3: 47–52. 124. Poorterman JHG, Weerheijm KL, Groen HJ, Kalsbeek H. Clinical and radiographic judgement of occlusal caries in adolescents. Eur J Oral Sci 2000; 108: 93–98. 125. Weerheijm KL, Groen HJ, Bast AJJ, Kieft JA, Eijkman MAJ, Van Amerongen WE. Clinically undetected occlusal dentine caries: a radiographic comparison. Caries Res 1992; 26: 305–309. 126. Ricketts DNJ, Kidd EAM, Smith BGN, Wilson RF. Clinical and radiographic diagnosis of occlusal caries: a study in vitro. J Oral Rehabil 1995; 22: 15–20. 127. Creanor SL, Russell JI, Strang DM, Stephen KW, Burchell CK. The prevalence of clinically undetected occlusal dentine caries in Scottish adolescents. Br Dent J 1990; 169: 126–129. 128. Kidd EAM, Naylor MN, Wilson RF. Prevalence of clinically undetected and untreated molar occlusal dentine caries in adolescents on the Isle of Wight. Caries Res 1992; 26: 397–401. 129. Fracaro MS, Seow WK, McAllan LH, Purdie DM. The sensitivity and specificity of clinical assessment compared with bitewing radiography for detection of occlusal dentin caries. Pediatr Dent 2001; 23: 204–210.

130. Espelid I, Tveit AB. A comparison of radiographic occlusal and approximal caries diagnoses made by 240 dentists. Acta Odontol Scand 2001; 59: 285–289. 131. Mejare I, Grondahl HG, Carlstedt K, Grever AC, Ottoson E. Accuracy at radiography and probing for the diagnosis of proximal caries. Scand J Dent Res 1985; 93: 178– 184. 132. Rugg-Gunn AJ. Approximal carious lesions. A comparison of the radiological and clinical appearances. Br Dent J 1972; 133: 481–484. 133. De Araujo FB, Rosito DB, Toigo E, Dos Santos CK. Diagnosis of approximal caries: radiographic versus clinical examination using tooth separation. Am J Dent 1992; 5: 245–248. 134. De Araujo FB, De Araujo DR, Dos Santos Ck De Souza MA. Diagnosis of approximal caries in primary teeth: radiographic versus clinical examination using tooth separation. Am J Dent 1996; 9: 54–56. 135. Akpata ES, Farid MR, Al-Saif K, Roberts EAU. Cavitation at radiolucent areas on proximal surfaces of posterior teeth. Caries Res 1996; 30: 313–316. 136. Mejare I, Malmgren B. Clinical and radiographic appearance of proximal carious lesions at the time of operative treatment in young permanent teeth. Scand J Dent Res 1986; 94: 19–26. 137. Bille J, Thylstrup A. Radiographic diagnosis and clinical tissue changes in relation to treatment of approximal carious lesions. Caries Res 1982; 16: 1–6. 138. Thylstrup A, Bille J, Qvist V. Radiographic and observed tissue changes in approximal carious lesions at the time of operative treatment. Caries Res 1986; 20: 75–84. 139. Pitts NB, Rimmer PA. An in vivo comparison of radiographic and directly assessed clinical caries status of posterior approximal surfaces in primary and permanent teeth. Caries Res 1992; 26: 146–152. 140. Lunder N, Von Der Fehr FR. Approximal cavitation related to bite-wing image and caries activity in adolescents. Caries Res 1996; 30: 143–147. 141. Ratledge DK, Kidd EAM, Beighton D. A clinical and microbiological study of approximal carious lesions. Part 1: The relationship between cavitation, radiographic lesion depth, the site-specific gingival index and the level of infection of the dentine. Caries Res 2001; 35: 3–7. 142. Qvist V, Johannessen L, Bruun M. Progression of approximal caries in relation to iatrogenic preparation damage. J Dent Res 1992; 71: 1370–1373. 143. Medeiros VAF, Seddon RP. Iatrogenic damage to approximal surfaces in contact with Class II restorations. J Dent 2000; 28: 103–110. 144. Lussi A, Gygax M. Iatrogenic damage to adjacent teeth during classical approximal box preparation. J Dent 1998; 26: 435–441. 145. Qvist V. Longevity of restorations in primary teeth. In: Hugoson A, Falk M, Johansson S, eds. Consensus conference on caries in the primary dentition and its clinical management. Jo¨nko¨ping, Sweden: The Institute for Postgraduate Dental Education, 2002; 69–83. 146. Qvist V, Laurberg L, Poulsen A, Teglers PT. Longevity and cariostatic effects of everyday conventional glass-ionomer and amalgam restorations in primary teeth: three-year results. J Dent Res 1997; 76: 1387–1396. 147. Wendt LK, Koch G, Birkhed D. Replacements of restorations in the primary and young permanent dentition. Swed Dent J 1998; 22: 149–155. 148. Brantley CF, Bader JD, Shugars DA, Nesbit SP. Does the cycle of restorations lead to larger restorations? J Am Dent Assoc 1995; 126: 1407–1413. 149. Hintze H, Wenzel A, Danielsen B, Nyvad B. Reliability of visual examination, fibre-optic transillumination, and bitewing radiography, and reproducibility of direct visual examination following tooth separation for the identification of cavitated carious lesions in contacting approximal surfaces. Caries Res 1998; 32: 204–209. 150. Landis JR, Koch GG. The measurement of observer agreement for categorical data. Biometrics 1977; 33: 159–174.

Caries paradigms in diagnosis 151. Byrt T. How good is that agreement? Epidemiology 1996; 7: 561–561. 152. American Child Health Association. Physical defects: the pathway to correction. New York: American Child Health Association, 1934. 153. Bakwin H. Pseudodoxia pediatrica. N Engl J Med 1945; 232: 691–697. 154. Ayanian JZ, Berwick DM. Do physicians have a bias toward action? A classic study revisited. Med Decis Making 1991; 11: 154–158. 155. Kress GC. Toward a definition of the appropriateness of treatment. Public Health Rep 1980; 95: 564–571. 156. Verdonschot EH, Angmar-Ma˚nsson B, Ten Bosch JJ, Deery CH, Huysmans MC, Pitts NB, Waller E. Developments in caries diagnosis and their relationship to treatment decisions and quality of care. Caries Res 1999; 33: 32– 40. 157. Kay EJ, Blinkhorn AS. A qualitative investigation of factors governing dentists’ treatment philosophies. Br Dent J 1996; 180: 171–176. 158. Maupome´ G, Sheiham A. Clinical decision-making in restorative dentistry. Content-analysis of diagnostic thinking processes and concurrent concepts used in an educational environment. Eur J Dent Educ 2000; 4: 143–152. 159. Kay EJ, Nuttall NM, Knill-Jones R. Restorative treatment thresholds and agreement in treatment decision-making. Community Dent Oral Epidemiol 1992; 20: 265–268. 160. Wulff HR, Gøtzsche PC. Rational diagnosis and treatment. Evidence-based clinical decision-making. Oxford: Blackwell Science, 2000. 161. Lewis DW, Kay EJ, Main PA, Pharaoh MG, Csima A. Dentists’ stated restorative treatment thresholds and their restorative and caries depth decisions. J Public Health Dent 1996; 56: 176–181. 162. Ismail AI, Bader JD. Evidence-based dentistry in clinical practice. J Am Dent Assoc 2004; 135: 78–83. 163. Grembowski D, Fiset L, Milgrom P, Conrad D, Spadafora A. Does fluoridation reduce the use of dental services among adults? Med Care 1997; 35: 454–471. 164. Grembowski D, Milgrom P. The influence of dentist supply on the relationship between fluoridation and restorative care among children. Med Care 1988; 26: 907–917.

277

165. Brennan DS, Spencer AJ. The role of the dentist, practice and patient factors in the provision of dental services. Community Dent Oral Epidemiol 2005; 33: 181–195. 166. Dome´jean-Orliaguet S, Tubert-Jeannin S, Riordan PJ, Espelid I, Tveit AB. French dentists’ restorative treatment decisions. Oral Health Prev Dent 2004; 2: 125–131. 167. Friedman JW, Atchison KA. The standard of care: an ethical responsibility of public health dentistry. J Public Health Dent 1993; 53: 165–169. 168. Grytten J. The effect of supplier inducement on Norwegian dental services; some empirical findings based on a theoretical model. Community Dent Health 1991; 8: 221–231. 169. Bader JD, Shugars DA. Understanding dentists’ restorative treatment decisions. J Public Health Dent 1992; 52: 102–110. 170. Espelid I, Tveit AB, Riordan PJ. Radiographic caries diagnosis by clinicians in Norway and Western Australia. Community Dent Oral Epidemiol 1994; 22: 214–219. 171. Kay EJ, Locker D. Variations in restorative treatment decisions: an international comparison. Community Dent Oral Epidemiol 1996; 24: 376–379. 172. Heidmann J, Helm S, Helm T, Poulsen S. Changes in prevalence of approximal caries in 17-year-olds and related restorative treatment strategies over a 6-year period. Community Dent Oral Epidemiol 1988; 16: 167–170. 173. Gimmestad AL, Holst D, Fylkenes K. Changes in restorative caries treatment in 15-year-olds in Oslo, Norway, 1979– 96. Community Dent Oral Epidemiol 2003; 31: 246–251. 174. Choi BCK, Jokovic A, Kay EJ, Main PA, Leake JL. Reducing variability in treatment decision-making: effectiveness of educating clinicians about uncertainty. Med Educ 1998; 32: 105–111. 175. Fleiss JL, Mann J, Paik M, Goultchin J, Chilton NW. A study of inter- and intra-examiner reliability of pocket depth and attachment level. J Periodont Res 1991; 26: 122–128. 176. Nadanovsky P, Sheiham A. Relative contribution of dental services to the changes in caries levels of 12-year-old children in 18 industrialized countries in the 1970s and early 1980s. Community Dent Oral Epidemiol 1995; 23: 331–339. 177. Nadanovsky P, Sheiham A. The relative contribution of dental services to the changes and geographical variations in caries status of 5- and 12-year-old children in England and Wales in the 1980s. Community Dent Health 1994; 11: 215–223.