A quantile plot for simultaneous representation of clinical and statistical attributes of probing change: application to early identification of the downhill patient

• Published in: Journal of periodontal research Vol. 33; no. 3; pp. 187 - 195
• Main Authors: Cohen, Mark E., Horning, Gregory M.
• Format: Journal Article
• Language: English
• Published: Oxford, UK Blackwell Publishing Ltd 01-04-1998
• Subjects: clinical significance; probing change; quantile plot; statistical significance
• ISSN: 0022-3484; 1600-0765
• DOI: 10.1111/j.1600-0765.1998.tb02310.x

When multiple periodontal sites are observed in patients over time there is an intention to identify those sites where there is important change, typically loss of attachment or increase in probing depth. A change may be declared if it: (a) exceeds a threshold level, and/or (b) is determined to be statistically significant (e.g. regression slope different from zero), perhaps after (c) that significance level has been corrected for multiple testing. These criteria are not often considered when clinical or research decisions are made and there is no universal protocol for their evaluation. A quantile (uniform probability) plot, modified to incorporate additional information, is proposed as a graphical method for the display of changes at multiple sites within a mouth. This plot identifies, for each site, clinical changes beyond a threshold, site‐wise statistical significance and statistical significance adjusted for multiple testing, These alternative criteria for attachment change are, thereby, made explicit, providing a detailed evaluative context. In addition, this methodology permits incorporation of an estimation procedure for the number of sites for which the null hypothesis of no change is false. This statistic can provide evidence of progressive disease even when no site has significant clinical or statistical change and even if the average change is zero. Use of the quantile plot was elucidated by application to simulated data, and to a clinical dataset using a BASIC program to automate the computational process. In the clinical example presented, the approach appeared more effective in detecting periodontal change than traditional clinical and statistical criteria. Pending technical refinement, this graphical approach may represent a new tool for the early identification of the downhill patient.