The unpleasant taste may prevent the normal population from using diluted bleach as a mouthwash. A recent pilot study, which has resulted already in two papers (since pooled microbiological samples had been taken, one might expect at least another paper) in the once prestigious Journal of Periodontal Research, might leave scientists even more skeptical. The published RCT does not mention the CONSORT statement. Apparently, no sample size was calculated. Randomization took place and 15 test patients with untreated periodontitis were asked to rinse twice per week with 0.25% sodium hypochlorite while 15 patients were supposed to rinse twice per week with water [sic] for 3 months. At baseline and after 2 weeks, patients received oral hygiene instructions, and pockets were irrigated with either 0.25% sodium hypochlorite or, well, water. The test solution was prepared by the patients, one teaspoonful Chlorox (6% sodium hypochlorite) on one-half glass of water (5 mL on 120 mL).
It is very much concerning that just 12 out of 30 patients (40%) completed the 3-month trial. While awful taste of Chlorox was actually mentioned by two patients who, for that, missed couple of rinses, according to Galvan et al. concerns of delay of proper periodontal treatment and transportation issues were the main reasons for drop outs.
But what is even more disturbing is how the authors analyzed their data. Although Galvan et al. (2014) claim that the subject was the unit for all statistical analyses, Gonzales et al., in an additional analysis of just 7 test and 5 control subjects who completed the 3-month trial write,
“The individual pockets were treated as independent statistical units, based on nonspecific and wide-ranging antimicrobial action of sodium hypochlorite and the observation that pockets with a large range of depths responded positively to the bleach treatment and that residual bleeding on probing sites showed no tendency to cluster in particular patients or around specific teeth.” (Emphasis added.)
Most of spurious evidence of the 1970s and 1980s in periodontics stems from that misconception, be it clinical responses when pockets with different depths were considered independent in a limited number of patients, or observations made in microbiological samples from numerous sites in certain patients. At least since Larry L. Laster’s paper of 1985, periodontists must be aware of inflated p-values and spurious conclusions.
7 Ordered Categorical Response Models
In several chapters so far we have addressed modeling a binary response variable, for instance gingival bleeding upon periodontal probing, which has only two expressions, yes or no (or 1, 0). Many kinds of response variables are ordered categorical. For instance, the Silness & Löe plaque index, which assesses the amount of supragingival plaque at a certain tooth surface adjacent to the gingival margin, comprises clinically defined situations on a scale of scores from 0 to 3 (Table 7.1).
6 January 2015, 11:12 am.
Last modified January 6, 2015.
My MLwiN user manual using periodontal data has been completed but may be supplemented with further models if applicable.
Graphic display of results of a randomized clinical trial may lead to interesting hypotheses which might be tested with more sophisticated statistical methods. In a steady-state plaque environment, the topographical distribution of plaque, as assessed by Silness & Löe’s plaque index, in subjects using triclosan-containing test toothpaste for 6 weeks (after a 4 week preparatory phase) or fluoride control toothpaste is shown in the upper two panels. Results in IL-1 genotype-positive and -negative subjects are differentiated. While plaque distribution was largely comparable, consistent (mean plaque index scores of each examination were displayed on top of each other) and symmetric and followed a distinct pattern; bleeding on probing (a binary response), which is shown in the lower two panels, was not consistent, not symmetric and did not really follow a distinct pattern. Prevalence of bleeding on probing in the two most right panels (Il-1 genotype positive test toothpaste users) seems to differ from the other groups. The procedures described in the following chapter of my manual may indeed allow modeling of site-specific bleeding on probing.
6 Repeated Measures Models for Binary Outcomes
In Chapter 3, we had described simple, and quite complex, repeated measures time series models in which continuous outcomes, for instance, gingival thickness or gingival recession, were modeled over time after the implantation of a bio-resorbable membrane, when it had to be assumed that the responses were non-linear and non-monotonic.
In this chapter we want to model the binary outcome, bleeding on gingival probing, in subjects with mild plaque-induced gingival disease over time. While participants of the 1999 Workshop on Periodontal Diseases and Conditions had realized that most gingival inflammation is indeed dental plaque-induced, there seem to be numerous intrinsic and extrinsic factors which may modify the response. For instance, a common toothpaste compound, Triclosan, seems to dampen gingival inflammation in the presence of dental plaque (Müller et al. 2006). One may also ask whether the so-called interleukin-1 genotype, a combination of two single polymorphisms in the IL-1 gene, i.e. a haplotype, which had been associated with increased susceptibility for destructive periodontal disease (Kornman et al.1997) has a clinically discernable influence on the inflammatory response on dental plaque.
22 December 2014 @ 1:10 pm.
Last modified December 22, 2014.
5 Logistic Models for Binary and Binomial Responses
In the previous chapters, continuous response variables had been considered in various variance components, random intercept, random coefficient, time series, and multivariate time series multilevel models. In this chapter, we want to look at binary or binomial (proportion) responses. We will mainly focus on the logit link function. As usual, we start with a single-level model and extend this to appropriately consider the three-level hierarchical structure. We also explore contextual effects here. Significance testing and model interpretation using odds ratios and variance partition coefficients are discussed.
5.1 Description of the Example Data Set
The data for an example are stored in an EXCEL file (bop_pli01.xlsx). The binary response variable here is presence or absence of bleeding on probing (BOP) at gingival units in 50 students at Kuwait University. All had plaque-induced gingival disease.
29 October 2014 @ 6:17 am.
Last modified Otober 29, 2014.
The (published) picture above had been created based on fixed estimates of a multivariate multilevel time series model of gingival dimensions after the implantation of a bioresorbable membrane for surgical root coverage. One may intuitively compare the multilevel model with a microscope here since raw data or presentation of means and standard deviations would hardly give a similarly elegant impression of what is actually going on after implantation of a membrane for guided tissue regeneration.
The fourth chapter of my new MLwiN manual using own data has been proofread and those who are still interested in “happy multilevel modeling” are encouraged to click here.
4 Multivariate Response Models
In the previous chapter, increasingly complex time series models have been set up in order to model gingival thickness, its width, the position of the mucogingival border relative to the cemento-enamel junction, and gingival recession after surgical implantation of a bio-resorbable membrane for guided tissue regeneration for the treatment of gingival recession. While some of these variables, such as thickness and width of gingiva might be positively related, others are not, for example gingival thickness and recession. Mucosal thickness had been measured at three locations: at the gingival margin, as well as at and below the mucogingival border (Müller et al. 2000d). In order to create general predictions of alterations of gingival dimensions after surgery, one single model would be preferred which might include mucosal thickness as measured at different locations as three different responses.
Multivariate response data are most conveniently incorporated into a multilevel model by creating a lower level below the original level 1 units. This will define the multivariate structure. Here we want to set up a 4-level model with multivariate responses (level 1) measured at different occasions (level 2) nested in higher-level units, i.e. teeth (level 3) and patients (level 4).
3 October 2014 @ 8:47 am.
Last modified October 3, 2014.