| Sign In to gain access to subscriptions and/or personal tools. |
TO THE EDITOR:Department of Public Health Sciences "G. Sanarelli" University "La Sapienza" P. le Aldo Moro 5 00185 Rome, Italy From an eight-year longitudinal study, Li and Wang (2002) analyzed the relationship between baseline caries on primary teeth at the age of 3 to 5 years and follow-up caries on permanent teeth in a cohort of 362 children. They found these variables to be statistically associated and the presence of caries in primary teeth to be predictive of caries in permanent teeth.
Since the numbers of true-positive, false-positive, true-negative, and false-negative subjects were not directly available from the text, I have extrapolated them to build a 2 x 2 table and re-calculate the caries-predictive power of caries on primary teeth. The data I have used were caries prevalence on permanent teeth (40.6%), proportion of children with baseline dmf = 0 (14.4%), proportion of children with baseline dmf > 0 and DMF > 0 (94%), and the proportion of children with baseline dmf = 0 and DMF = 0 (83%) (Table
To be sure that my extrapolations were exact, I re-calculated the Relative Risk (RR = 2.57; 95% Confidence Interval, 1.40 - 4.72). Finally, I re-calculated Sensitivity, Specificity, Positive, and Negative Predictive Values to be 0.939, 0.200, 0.445, and 0.827, respectively. While the RR, Sensitivity, and Specificity I calculated were equal to the values reported by Li and Wang, the PPV that I found was lower. If the "true" overall PPV was 44% and not 85%, I suspect that the other PPVs of baseline caries on subsets of primary teeth reported by the authors were also not correct. The numbers of tp, fp, fn, and tn subjects for all the possible combinations tested must be reported and analyzed. If not, the authors' error leads to inverse conclusions.
Even if the reported PPVs were exact, there are other aspects that are not agreeable. First, the predictive power cannot be estimated by the correlation coefficient, RR, or by the PPV without the NPV, as in the paper of Li and Wang. There are many methods to assess predictive power, including the value of Sensitivity plus Specificity (Hausen, 1997). The highest value for this measure deducible from the paper was 119.5, far from the minimum required for an effective test (i.e., 160, with both parameters There is a final important methodological limitation in the paper. Caries prediction is modeled as a high-risk preventive strategy. When caries incidence and the fraction of children at high risk are extremely low, it is possible to save money and resources by identifying children at high risk and offering them efficacious individual protection (Rose, 1992). However, for a high-risk strategy to be justifiable, the fraction of children classified at high risk should not exceed 30% (Hausen, 1997). With the high caries prevalence in primary teeth reported by the authors of the paper, more than 80% of children would be classified as high risk and would receive intensive preventive care, whereas the remaining 20% would not. In this case, the time and the resources required to apply the test would be greater than those saved by leaving such a small fraction of children without preventive care. My opinion is that the paper by Li and Wang demonstrates that caries status of primary teeth is not predictive of caries on permanent teeth in their cohort. Study populations with a high proportion of subjects classified to be at high risk require population-based programs.
REFERENCES
THE AUTHORS REPLY:
1 Department of Basic Science and Craniofacial Biology, New York University College of Dentistry, New York, NY 10010, USA; yihong.li{at}nyu.edu. Our study found that the caries status of primary teeth, especially the primary molars, can be used as a risk indicator for caries development in the permanent dentition of Chinese children. This conclusion was made based not upon the positive predictive value which depends on the prevalence of the disease, but, rather, upon the following findings from the study: (1) a significant correlation between caries in the primary and permanent teeth; (2) a high sensitivity (93.9%) in predicting caries in permanent teeth for children with caries in the primary teeth; (3) a significant relative risk associated with caries in permanent teeth for children who manifested caries compared with children caries-free in the primary teeth; and (4) an increasing pattern of the relative risk associated with caries in permanent teeth as the mean dmfs and mean dmft scores in the primary teeth increase. In fact, our findings were consistent with the work performed by Heller et al. (2000), who used a different research approach. Based on insurance claims data, they reported that primary posterior teeth treatment was significantly associated with future caries treatment in first permanent molars. Their study, in addition to others (Powell, 1998), has suggested that caries experiences in primary teeth should be considered as a risk predictor for future caries. We appreciate Dr. Petti's thoughtful reading of the article and agree that the overall positive predictive value should be lower than that given in the original article, based on the fact that high caries prevalence occurred in the primary teeth and low caries prevalence in the permanent teeth. It should be clarified that the 85.4% was calculated from the subgroup children (23.4% of the total) who had a mean dmft score greater or equal to 7 and who developed caries in their permanent molars. Dr. Petti's second point relates to the relatively low value (119.5) of sensitivity plus specificity used in the caries-risk screening examination. Precisely, the highest combination value was 125.7 in Table 3. By using the combination to evaluate a diagnostic test, one would normally assume that the two values were equally important to a test outcome, which is true for many diseases. Dental caries, however, has unique characteristics. It is infectious but not life-threatening, no single causality has been reported for the disease, a considerable amount of time is required for caries to develop, and preventive measures are available at reasonable cost. Considering these unique characteristics, we strongly believe that a risk-screening test with a high sensitivity is more important than one of high specificity. Thus, we intentionally decided not to use the combination approach, but rather to examine the sensitivity and specificity separately. To date, studies have failed to show the ability of a simple clinical diagnostic test to predict future caries activity accurately in individuals (Powell, 1998). With a multitude of variables and complex statistical models, in which bacterial levels, dietary behavior, salivary factors, and other social variables were included, a combined sensitivity and specificity score could be as high as 173 (Steiner et al., 1992). In our study, we examined only past caries experience; therefore, a low combination value of sensitivity plus specificity was expected.
Dr. Petti also raises an important point regarding caries prediction and caries-preventive strategy. We believe that children having one decayed tooth or caries on only 4 maxillary incisors should not be simply classified as "high-risk individuals". If more than 80% of the children already affected by dental caries in their primary teeth were classified as a high-risk group, it would be pointless to develop and implement a cost-effective preventive measure to prevent dental caries in the permanent dentition, especially in China, which has a severe shortage of dental professionals at all levels (PRCMPH, 1999). Since caries distribution was skewed in this study cohort and in many other populations (Hausen, 1997), various methods have been used to delineate low- and high-risk individuals. For example, Kaste et al. (1992) found that Native American children with a dmft of 5 or more were more likely develop caries in their permanent teeth (RR = 2.4, 95% CI = 1.4-4.3). Heller et al. (2000) demonstrated that primary posterior tooth treatment at ages 4-8 was significantly associated with future caries treatment (RR = 2.5, 95% CI = 2.3-2.7). Bratthall (2000) introduced the Significant Caries Index using DMFT
In our study, we proposed to divide the decayed primary teeth in different risk groups. The main objectives were to make caries prediction more accurate and to introduce a different means of detecting those individuals most in need of enhanced caries prevention. A less significant correlation was found between caries in the primary anterior teeth and that in the permanent teeth. Statistical analyses for the different combinations of caries in the primary molars demonstrated that: (1) the positive predictive values ranged from 47.3% (caries on any of the mandibular molars) to 65.4% (caries on all primary molars); (2) the relative risk values were 1.8 (95% CI = 1.4-2.3) to 3.4 (95% CI = 1.8-6.2); and (3) all of the results were highly significant. Statistical analyses were also performed for children with high risk (dmft As Dr. Petti pointed out, a population-based caries-prevention program would be cost-effective in controlling the high caries prevalence in young Chinese children. Community-based water fluoridation, for example, has been a very successful caries-preventive approach in the United States and is listed as one of the top ten public health achievements in the United States in the 20th century (CDC, 1999). However, it will have to be carefully reconsidered to benefit Chinese children, for several reasons: Chinese food culture and dietary practices are very different from those of the United States, and the optimal level of fluoride that needs to be added to the water supply is yet to be determined (Guo, 2000). One-sixth of the nation (242,885,400 of the entire population) live in areas where excess fluoride has been found in the drinking water, soil, and air (Chen, 1997; Chen, 2000). Lack of infrastructure and professional manpower prevents the implementation of community-based water fluoridation programs. Since China has established an outstanding nationwide primary health-care system and network, we believe that dental public health strategies should utilize the existing system and focus on improving maternal oral health conditions and awareness plus intervention to reduce primary teeth susceptibility to caries and caries incidence in permanent teeth. Further studies of those strategies are critically important. REFERENCES
Journal of Dental Research, Vol. 81, No. 12,
804-806 (2002)
|
|
||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
 |
|
|
 |
|
Sign In to gain access to subscriptions and/or personal tools.
80%).