I I

Comparison of colorimeter SHORT and electrode analysis COMMUNICATIONS of water fluoride Burton L. Edelstein, DDS, MPH David Cottrel, DDS David O’Sullivan, BS Norman Tinanoff, DDS, MS

Introduction Fluoride supplement prescriptions must be based on metric method entailed diluting each sample with the fluoride content of a child’s drinking water.1 Failure SPADNSreagent and comparing transmitted light of to analyze a child’s water source for fluoride increases the sample against transmitted light of a 1.0 ppm water the possibility for either inducing fluorosis or provid- fluoride standard using the Hachcolorimeter that reads ing inadequate cariostatic protection. Twomethods of out fluoride level on an analog scale. The paired find- determining water fluoride content, the ion specific ings were examined for differences that would yield electrode (electrode or ISE) and the SPADNSreagent contradictory fluoride supplement prescriptions based (colorimetric) methods, generally are used. The elec- on currently accepted supplementation schedules. trode method is knownto be more accurate because it is The range of fluoride content in the sample set based less influenced by impurities in the sample2, while the on the electrode method was < 0.01 to 2.6 ppmfluoride. colorimetric method is popular amongdentists because Of the 222 samples, 21%exceeded 0.7 ppm, 12%showed of its low cost and ease of implementationin the office. fluoride content between 0.3 and 0.7 ppm, and 66% Presently, there is conflicting information regarding the contained less than 0.3 ppm. The correlation coefficient usefulness of the colorimetric method. A study compar- for linear association between the two methods was ing the two methocis for water fluoride analysis re- very strong (r = .935). A scatter plot of paired findings ported that prescriptions for dietary fluoride supple- for each sample was overlaid on a graph showing ac- mentation based on the colorimetric analysis would cepted threshold values for prescribing fluoride supple- have been incorrect 45%of the time. 3 Another report ments of various doses (Fig 1). In 5%(N = 11) of suggests that the colorimetric method "may be a valid, paired samples, differences between the two methods reliable, and affordable tool for fluoride analysis in the dental office. ’’4 The purpose of the present study was to re-examine the difference between the electrode and colorimetric techniques to assess the clinical usefulness of colorimetric analysis in a dental practice setting. 1.4- Full Half No Supplen Supplement Supplement Materials and Methods Water samples were obtained from 222 consecutive 1.2- No new patients with private wells presenting to a private pediatric dental practice. All patients reside in New LondonCounty, Connecticut, an area of variable ambi- ent fluoride. Each sample was tested for fluoride con- tent one time each by the electrode and colorimetric methods. The samples analyzed by the electrode method were processed by a technician in a Connecticut-ap- proved public health laboratory that is proficiency tested yearly, while samples analyzed by colorimetry were processed by assistants in the dental practice. The elec- trode method entailed diluting each sample 1:1 with ,...... ~.,.~...~...~.~~..-~..-~.. Full 1,2- cyclohexylene dinitrilotetracetic acid (TISABwith CDTA;Orion Research, Cambridge, MA) to adjust pH and ionic strength. The fluoride concentrations of the ppm F byCol0rimeter samples then were determined using a fluoride elec- ® Fig 1. Scatter ~lot of paired water samplefluoride content trode (Orion 94-09-00 , Orion Research, Cambridge, determinations bg e~ectrodeand colorimetric methods.Shaded MA)connected to a digital readout electrometer (Orion areas show~airs for whichsupplemental fluoride ~rescri~tioas 601A®, Orion Research, Cambridge, MA). The colori- based on both methodswould be ~he same.

PEDIATRICDENTISTRY: JANUARY/FEBRUARY, 1992 ~ VOLUME14, NUMBER1 47 would yield contradictory fluoride supplement pre- Rx Acc~ate?

scriptions. The plot shows p=.22 that the colorimetric sam01e~$ ,3-.7 ~ full no method tends to overstate sarrole is <.3 ~ full yes fluoride content at low fluoride levels compared with the electrode method. Decision analysis 5 was used to calculate the rela- tive accuracy of fluoride prescriptions that would result from each method as well as the relative risk 0=0 of theoretical fluorosis from inaccurate prescrip- tions based on each method. Decision trees were constructed compar- ing electrode assay, colo- rimetric assay, and pre- scribing fluoride supple- mentation with no assay. Results Fig 2. Decisiontree for comparingaccuracy of fluoride supplementprescriptions based on noassay of waterfor fluoride content,assay by colorimeter,and assay by ion specific electrode(ISE). Fig 2 shows a decision Decisionnodes are representedby squares.Chance nodes are representedby circles. Probabilities tree comparing accuracy of on eachbranch are given by P values. Thefar right columnsshow 1 5 possible prescription fluoride supplement pre- outcomesand their associatedaccuracies. scription. The first branch, "no assay," leads to full-strength prescriptions regard- Fig 3 (next page) shows the decision tree comparing less of true sample fluoride content, since all water the theoretical risk of induced fluorosis from fluoride samples are assumed to be fluoride deficient. The true supplementation. Risk of fluorosis was assigned for prevalence of fluoride at various levels, shownto the various fluoride supplementation levels in excess of the right of the probability nodes, will determine the chance currently accepted dose schedule. Decision analysis of the prescription being accurate. In this sample, as shows that the risk of induced fluorosis is 14 times measured by ISE, 66%of wells contained less than 0.3 greater with no assay than with the electrode (28.2 vs. ppm fluoride. Therefore, 66%of the prescriptions based 2.0), and that the colorimetric assay results in slightly on "no assay" were correct, leaving 34%incorrect. The lower fluorosis risk than that of the electrode (1.8 vs. colorimeter branch shows that the "colorimeter" mea- 2.0). sured 24% of the samples to contain > 0.7 ppm, 19% For any given set of samples, the relative prescrip- between 0.3-0.7 ppm, and 57%to be < 0.3 ppm. The next tion accuracy and induced fluorosis risk depend on the node shows the probability of the colorimeter being prevalence of the fluoride content in the ambient water correct at each test level. In all cases in which the supply. The greater the prevalence of ambient fluoride colorimeter was correct, the resultant prescriptions were in area wells, the morelikely that no assay will result in accurate. Thus the weighted average for prescription improper prescriptions. accuracy using the colorimeter was 89.4% [(.94 x .24) (.52 x .19) + (1.00 x .57)]. Since the electrode method Conclusions assumed to be 98%accurate 6, the prescription accuracy The present study confirms previous findings that for the electrode method is assumed to be 98%. Ratios colorimetric assay of fluoride both produces generally between the alternative assay methods are used to de- higher findings than the electrode method3 and closely termine relative accuracy of the three methods. The correlates to electrode findings. 4 Fluoride prescription electrode was 33% more accurate than "no assay" (98 in clinical practice is intended to provide optimal sys- value for electrode vs. 66 for no assay), but only 10% temic fluoride ingestion, but the actual fluoride intake more accurate than the colorimetric assay (98 value for of children is confounded by dietary fluoride, tooth- electrode vs. 89.4 for the colorimeter). paste ingestion, exposure to multiple water sources and

48 PEDIATRICDENTISTRY: JANUARY/FEBRUARY, 1992 - VOLUME14, NUMBER1 variable compliance with prescribed supplements. When differences in pre- P=.22 sarnole ~s .3-.7 ~ 0.500 scriptions resulted from p=.12 the two methods, the pre- p=.66 scribed doses based on sample ~s >.7 ----~ 0.020 colorimetry in this study test says p=.94 p=.24 tended to be low, thereby water sample p=.06 minimizing the risk of sample ~s .3-.7 ~ 0.020 fluorosis. These findings ColOrlmeter _ __ test says .3-.7 ~ p=.52 sample ,S not .3-.7 ~ 0.005 further suggest that the 0=.48 colorimeter should be ac- sample ~S <.3 ~ 0.020 cepted as a clinically use- p= t.O0 ful alternative to the elec- P=O trode method, particularly in areas of high ambient test ~ys >.7 P=.99 sample ~s not >.7 ~ 0.005 fluoride levels.

sample *s .3-.7 ~ 0,020 Dr. Edelsteinis clinical instruc- P=,99 sample *s not .3-.7 ~ 0.005 tor, Mr. O’Sullivanis research 13=.01 assistant, and Dr. Tinanoffis pro- fessor; all are in the Department sample ~S <.3 ~ 0.020 of Pediatric Dentistry, Univer- #=.99 sity of ConnecticutHealth Cen- ter, Farmington,CT. Dr. Cottrel is former resident, Department of Dentistry, Yale NewHaven Fig 3. Decision tree for comparingpresumed fluorosis risk of fluoride supplementprescriptions Hospital. NewHaven, CT. Re- basedon no assayof water for fluoride content, assayby colorimeter, andassay by ion specific print requests shouldbe sent to electrode (ISE). Decision nodesare representedby squares. Chancenodes are represented Dr. Burton L. Edelstein, 190 circles. Probabilities on eachbranch are given by p values. Thefar right columnsshow 1 5 possible HempsteadStreet, NewLondon, prescription outcomesand their associatedrisk for fluorosis. CT06320. 1. AmericanAcademy of Pediatric Dentistry: Oral Health Policies, 4. WeinbergerSJ, Johnston DW,Wright GZ: A comparison of two Protocol for Fluoride Therapy,May, 1985. systems for measuringwater fluoride ion level. Clin Prevent 2. Hach Co: DR100 Colorimeter Manual. Loveland, CO: Hach Co, Dent11: 19-22,1989. 1983. 5. Weinstein MC,Fineberg HV,Elstein AS, Frazier HS, Neuhauser 3. BrossokGE, McTigueDJ, KuthyRA: The use of a colorimeter in D, Neutra RR,McNeil BJ: Clinical Decision Analysis. Philadel- analyzingthe fluoride content of public well water. Pediatr Dent phia: WBSaunders, 1980. 9: 204-7,1987. 6. Orion Research: Instruction Manual,Fluoride Electrodes, Model 94-09-00, FormIM94 96-09/3821, 1983, p 26.

PEDIATRIC DENTISTRY: JANUARY/FEBRUARY, 1992 -- VOLUME 14, NUMBER 1 49