Researchers have created several epidemiologic indexes for assessing the severity of DF.⁷ Dean’s index (DI), which was developed in 1934, is still in use.⁸ However, because this index covers the entire range of DF severity with only four levels of severity, researchers developed other important indexes: the Thylstrup and Fejerskov index (TFI),⁹ the Tooth Surface Index of Fluorosis¹⁰ and the Fluorosis Risk Index.¹¹

	TFI.

TOP ABSTRACT TFI. MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
The TFI⁹ is the only index that attempts to correlate the clinical appearance with the pathological changes in the tissue,¹² and, therefore, it is a useful tool when evaluating DF severity in epidemiologic studies. However, despite its numerous advantages, the TFI, as well as the other DF indexes cited, is not a continuous scale. All current DF indexes use ordinal scales and, therefore, the scores should be considered only arbitrary points along a continuum of change.¹³

A preferred method would be to express the severity of DF in a continuous scale by showing the position of different observations relative to each other and the extent to which one observation differs from another.¹⁴ These properties are extremely important when the correlation between two factors is studied, as well as when evaluating the cause-and-effect relationship between two variables. Thus, the importance of creating a continuous scale for assessing the severity of DF is evident.

The visual analog scale (VAS) is a 100-millimeter continuous scale originally developed for pain measurement,¹⁵ but it also is used to measure alertness after sleep, attitudes toward the environment, quality of life and anxiety.¹⁶ Among its qualities, the VAS is completed rapidly, produces ratio data (that is, a linear scale), has high sensitivity to change, is easy to score and has good construct validity.¹⁷

We describe the development and validation of a VAS for DF. In addition, we describe the findings of a clinical (in a DF-endemic area) and a laboratory (using third molars from Canada and Brazil) protocol comparing the VAS with the TFI.

	MATERIALS AND METHODS

TOP ABSTRACT TFI. MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
We organized this study into three phases: the development of the VAS for DF, the validation of the VAS for DF, and clinical and laboratory use of the VAS for DF. The University of Toronto Ethical Committee approved this study, which was conducted in 2001 and 2002.

Development of the VAS for DF. The first step in developing the VAS was to observe the way DF is perceived by the population (that is, by dentists and nondentists) in a continuous scale. To accomplish this, one of us (A.P.G.F.V.) showed 23 photographs of teeth with different levels of DF (as measured by the TFI) to 23 dentists (graduate students at the University of Toronto) and 23 laypeople who had a college education and were residents of Toronto.

Photographs were taken of patients’ anterior teeth in areas of endemic DF, as well as from dental textbooks. The examiners were asked to grade the amount of discoloration and malformation in the photographs on a 100-mm VAS. The left end of the scale was marked "best you can imagine," and the right end of the scale was marked "worst you can imagine." The examiners received an information sheet and signed a consent form to participate in the study. To avoid any type of bias, we did not tell examiners that the photographs represented cases of DF or that we planned to use their measurements for the creation of a VAS for DF. The photographs were shown in random order to each examiner.

We calculated the VAS score for each photograph, as perceived by each examiner, using a 100-mm ruler underneath the VAS and measuring the distance between the left end of the scale and the mark made by the examiner. We calculated the mean values, the standard deviations (SDs) and the standard errors (SEs) for each photograph. The sample size calculation for this study was based on the results of a pilot study (involving 16 subjects and seven photographs). The mean SD for the photographs in the pilot study was 1.6 centimeters. We set the desired (that is, clinically significant) SE at 0.35 and calculated the sample size as 21.

Validation of the VAS for DF. After constructing the scale, we performed a series of statistical tests to validate the new VAS for DF. A valid scale is one that allows accurate inferences about what is being measured.¹⁸ Reliability is an indication of the consistency of scores over time, the consistency between scores, or the consistency of scores across different tasks or items that measure the same thing. Reliability usually is assessed by the reliability coefficient, which is defined as the proportion of variance in the scores related to true variance between the objects being measured.¹⁹

Interexaminer reliability. We used the intra-class correlation coefficient (ICC) to assess the level of agreement among the 46 examiners.²⁰ The ICC represents the proportion of variance in the scores related to the true variance between the objects being measured—in this case, photographs illustrating DF. We also used ICC to determine whether any differences existed in the way dentists perceived DF versus the way nondentists perceived it.

Intraexaminer reliability. We asked five of the examiners who were not dentists to repeat the test for the seven photographs (40 days apart). We used ICC to evaluate the intraexaminer reliability for each of these five examiners.

Validation of each examiner (n = 46) against the two gold standards (TFI and DI). We used Spearman rank correlation (r_s) to correlate the results of the VAS for DF with the TFI and DI. We chose the Spearman rank correlation because the TFI and DI are ranked data.

Validation of mean VAS for DF scores against TFI and DI. We also used the Spearman rank correlation to assess the validity of the VAS for DF in relation to the TFI and DI using the mean scores of all 46 subjects for each photograph.

Clinical use of the VAS for DF. To test the performance of the VAS for DF, two dentists (A.P.G.F.V. and Ivan Junior, D.D.S.) examined 54 children and adolescents (aged 2 through 16 years) living in an area of endemic DF (Sobral, CE, Brazil) (Table 1). Guardians and/or patients consented to a clinical examination. Using the photographs that were used to create the VAS for DF, one of us (A.P.G.F.V.) calibrated the second dentist for the VAS.

Laboratory use of the VAS for DF. We analyzed unerupted third molars collected from hospitals and private dental offices in Montreal, Toronto and Fortaleza, Brazil, for fluoride concentration (dentin and enamel) and severity of DF. We measured the fluoride concentration using instrumental neutron activation analysis, and we measured the severity of DF using the TFI and the VAS for DF. Our aim was to determine which scale was more sensitive to the changes occurring in the tooth structure as a result of the fluoride concentration; we accomplished this by correlating the dentin and enamel fluoride concentration with the two DF scales. We obtained enamel and dentin samples from a coronal midsection of the teeth, as described elsewhere.²¹

	RESULTS

TOP ABSTRACT TFI. MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Development of the VAS for DF. We used the mean values, SD and SE for each photograph to create the VAS for DF visual indicators, which are photographs of different degrees of DF severity distributed throughout the scale (Figure). These indicators guide the user and help in the precise use of the scale. We created them because of the lack of set criteria for each of the scale points (which is different from other DF severity scales).

View larger version (28K): [in this window] [in a new window]   Figure. Visual analog scale for dental fluorosis with visual indicators.

Intraexaminer reliability. To recapitulate, five examiners conducted their two analyses 40 days apart. The ICC showed excellent intraexaminer correlations (Table 2).

View this table: [in this window] [in a new window]   TABLE 3 SPEARMAN RANK CORRELATION FOR EACH EXAMINER ON THE VAS FOR DF* VERSUS THE TFI AND DI.

Clinical use of the VAS for DF. Using the VAS for DF, two dentists examined 54 children. The results of these examinations are shown in Table 1. The correlation between the VAS for DF and the TFI was high (r_s = .885, P < .001), and the correlation between the two examiners was excellent (r_s = .922, P < .001). The ICC between the two examiners showed excellent reliability (ICC = .96). The ICC between the two DF scales (TFI and VAS) was good (ICC = .6).

Laboratory use of the VAS for DF. We used only those teeth that had a complete (56 percent) or almost complete (44 percent) root. The enamel fluoride concentration ranged between 32 and 303 parts per million, and the dentin fluoride concentration ranged between 101 and 549 ppm. TFI values ranged between 0 and 4, while VAS for DF values ranged between 3 and 55 mm. The maximum TFI value for unerupted teeth was 4.^24,25

The study results revealed significant correlation between the enamel fluoride concentration and the VAS for DF values (r = .237, P = .035), while only a correlation trend (that is, P value ranged between .05 and .1) was seen between the enamel fluoride concentration and the TFI values (r = .191, P = .092). We found a stronger correlation between the dentin fluoride concentration and the VAS for DF values (r = .441, P = .001) than between the dentin fluoride concentration and the TFI values (r = .370, P = .001).

	DISCUSSION

TOP ABSTRACT TFI. MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Continuous versus noncontinuous scales. The currently used indexes for DF are based on discrete, ordinal data to evaluate the varying degrees of severity of DF, and they require less robust, nonparametric tests for analysis. This makes quantification more difficult, and the resulting statistical analyses are potentially less precise. Conversely, the continuous VAS for DF is amenable to more robust and meaningful parametric analyses. In fact, the results we found using the VAS for DF were promising and showed consistently high correlations between examiners. For many conditions and situations, discrete descriptors fail to capture subtle variations. We can postulate that by using a continuous scale for DF, researchers and clinicians can more readily evaluate subtle changes, yielding more accurate and reliable results.

The concept behind VAS measurements compared with ordinal measurements can be explained in two ways. First, verbal descriptors of ordinal measurements may fail to describe exactly what is experienced or perceived by the observer (the patient or clinician). Second, discrete numeric-rating scales can impose artificial categories on an area of study that is best evaluated by a continuum.²⁶

For example, a commonly used scale for DF is the DI. This scale uses the values of 0 to 4 to indicate tooth changes related to DF (0 = unaffected teeth and 4 = most severe cases of DF). However, a score of 4 is not indicative of a condition four times as severe as a condition that received a score of 1. This lapse of continuity restricts the use of robust statistical analysis when evaluating the data. In a continuous scale (such as the VAS for DF), a measurement of 4 mm compared with a measurement of 1 mm reflects a fourfold increase in the severity of the condition, as perceived subjectively by examiners.

Another problem that arises in noncontinuous scales (such as the DI) is the placement of borderline subjects. If a tooth has characteristics that place it between a score of 3 and 4, a score of 3.5 cannot be given, and the observer is forced to decide arbitrarily which score the tooth will receive. This introduces bias and, consequently, inaccuracy into the data.

The main advantages of the VAS for DF over the ordinal scales for DF (such as TFI, DI, Tooth Surface Index of Fluorosis) are the continuity of the scale, its simplicity and its precision. The examiner has to look only at the tooth in question and mark the VAS at the appropriate point to indicate the severity of DF observed. The continuity of the scale favors its use in correlations between DF severity and other parameters, such as fluoride concentration and tooth properties (for example, tooth microhardness, hydroxyapatite crystal size, ultrasound velocity), and it allows the use of more robust parametric statistical tests.

Use of visual indicators. However, the continuity of the scale restricts the creation of criteria for each point on the scale. Unlike other scales, the VAS for DF does not have set criteria for each of its scale points, as it would be virtually impossible and meaningless to establish criteria for each of the 100 points on the scale. Thus, instead of creating an entire set of criteria, we developed visual indicators of DF for the scale. These indicators guide examiners through the scale and help them use it precisely.

This absence of criteria initially can be seen as a limitation of the scale, owing to the perception that written criteria improve validity and reliability. However, Rozier⁷ performed a critique of several indexes used for DF severity that incorporated set criteria. He reported the percentage agreement and statistic (intraexaminer and interexaminer) for several studies that used these indexes. The percentage agreement varied between 65 and 100 percent, and the statistic varied between 0.35 and 0.93.

When TFI was evaluated (using 125 buccal photographs), the intraexaminer reliability was 76 and 85.6 percent and the interexaminer agreement for the two sessions was 72 and 79.2 percent.⁷ These values are comparable to the values presented here for the VAS for DF, which suggests that the absence of set criteria does not interfere with the validation and reliability of the scale.

The results of this study showed that the Spearman rank correlation between the VAS for DF and traditional indexes (TFI and DI) was good (.6 < r_s < .8). It is important to note that this type of correlation is desirable when creating a new scale or index.¹⁸ An excellent correlation (r_s > .8) would have indicated that both indexes were measuring exactly the same phenomenon (note that only a good correlation was seen between the VAS and the traditional indexes). A poor correlation (r_s < .21) between the new and old indexes would have indicated that the differences between the indexes were too large to be meaningful, and that they probably were measuring different factors.

Thus, the good correlation found in our study demonstrated that the scale measured what it was supposed to measure (that is, DF severity), but not in the same manner as did the traditional scales. This finding reinforces the validity of the VAS for DF and its usefulness.

To cover a broader range of DF severity, it is important to prove the usefulness of the scale throughout its range (from no DF to severe DF). To achieve this, we chose an area of endemic DF to test the VAS. However, because of limitations related to the area of endemic DF (a small community district in Brazil’s countryside), the number of children examined was limited.

The severity of DF is known to be greater in the permanent dentition than in the primary dentition.²⁷ During clinical use of the VAS for DF, we used both primary and permanent dentitions. Because we performed the examinations to evaluate clinical use of the scale and did not intend to perform an epidemiologic evaluation of the population examined, we did not anticipate any methodological error.

The results of our laboratory study are important to show the higher sensitivity of the VAS for DF to tooth changes related to fluoride concentration. These findings affirm the usefulness of the scale in research and clinical settings.

Apart from being valid and reliable, a good index also needs to be cost-effective, easy to implement and well-accepted by patients and examiners. Although this study was not designed to analyze these factors, we are confident that the VAS for DF also fulfills these requirements. The scale was easy to administer, and examiners and patients did not report experiencing any problems with it.

	CONCLUSION

TOP ABSTRACT TFI. MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

 
Our analysis of the data presented above shows that the VAS for DF is valid, has excellent reliability and is ready to be used on a large-scale basis in clinical trials. We realize that some health care professionals may be skeptical about a scale that has no set criteria. However, we are confident that the simplicity, precision and usefulness of this scale will overcome these initial concerns.