Risk factors of oral candidosis: A twofold approach of study by fuzzy logic and traditional statistic
Introduction
Oral candidosis (OC) is a common opportunistic infectious disease. The predisposition to oral candidal colonisation and infection is increased with xerostomia,1 immunodeficiency,2, 3, 4, 5, 6 denture wearing7, 8 and tobacco smoking.9 In addition, it has been suggested that high carbohydrate diet10, 11 and diabetes mellitus12, 13, 14 may predispose to OC, although there is conflicting data, particularly on the former.
There is a likelihood that more people will be at risk of OC as a consequence of increasing lifespan, greater use of immunosuppressive therapies and the wider range of drugs that can give rise to hyposalivation/xerostomia.15, 16 For example, the prevalence of diabetes mellitus and hypertension is increasing in the developed countries, and 88% of people over 65 years of age receive multiple medications. In addition, the number of patients receiving immunosuppressive therapy (e.g. following allograft transplants or autoimmune diseases) has risen considerably in the last decade.17 In view of these changes, there is a need to establish more specifically a target of people at risk of OC in order to create for them an appropriate preventive program.
Until now, the association between the several risk factors for OC, above mentioned, and its development and/or chronic maintenance has been widely conducted only by statistical traditional methodology (Aristotelian logic).18, 19, 20, 21, 22, 23 Nevertheless, this approach could have any limits, above all which one related to impossibility to consider intermediate logical value. Consequently the analysis of the non-linear links and complex interactions between multiple variables under study result extremely complicated.24 In order to surpass these restrictions it should be possible to utilise a twofold approach of study by fuzzy logic (FL) and statistical traditional methodology (STM).
FL is a superset of conventional (Boolean) logic introduced by Zadeh in the 1960s as a means to model the uncertainty within natural language.25, 26 Subsequently, fuzzy theory has been applied in several fields, especially in medical applications.27, 28, 29, 30, 31 It is based on theory that the description of a decision in terms of “more…less” is more adequate to real problems than by keen “yes–no” degree and it is not surprising that the first reported using was just in medical investigation.32 Hudson et al., analysing different source of uncertainty in medical decision making, concluded that FL could afford an efficient approach to solve medical problems especially when a complex interactions between the factors under study is present.33 In fact, the important advantage of this analytic method in comparison with classical statistical models, is the capacity to handle, at the same time, a very high number of variables notwithstanding the fact that these are not linearly connected. Therefore, in a framework in which several risk factors operate and need to be analysed simultaneously FL analysis might be a powerful tool for accurately detecting causal relationships, also in the field of infectious diseases.30, 31
The fuzzy neural network (FNN) is one of the most advanced artificial neural networks (ANNs) models, and its most attractive feature is that relationships between input and output variables can be described accurately from the acquired model.26 In the present study the input variables are the predisposing factors for OC, whereas the output variable is OC.
The aim of the present study was to verify the most commonly reported risk factors for OC onset and its chronic status utilising a twofold approach of study by FL and STM.
Section snippets
Study design
A case–control study in a secondary hospital base (Department of Oral Sciences, Section of Oral Medicine, University of Palermo, Palermo, Italy) was undertaken. Between January 1999 and July 2001, 371 patients were referred to the above institution to confirm the clinical diagnosis of OC. The study group was composed of 89/371 patients [mean age 66 years (range 35–88 years), males 31 (34.8%), females 58 (65.2%), smoker 19 (21.3%) and non-smoker 70 (78.7%)] enrolled on the basis of these
Risk of OC
The first step was to evaluate the risk association of the socio-demographical variables (A, G, and S) with OC (binary outcome: present/absent) in the study group vs. controls. This stepwise was performed by means of FL and STM. After examination of the inference rules elaborated by FL, the OC risk was found to be higher in females than males, in both categories it increased with A (≥54 years) regardless of tobacco smoking (RMSE = 0.363, Accuracy = 0.80, Sensitivity = 0.79, Specificity = 0.81) (Fig. 2).
Discussion and conclusions
Oral candidosis is the most common opportunistic oral infection in human beings. Several predisposing factors have been recognised as potentially involved into the development and, above all, into the chronicity and/or relapse of this infection (age, tobacco smoking, hyposalivation/xerostomia, denture wearing, antibiotic and corticosteroid therapies, endocrine disorders, immunodeficiency and contemporary malignancy) and the analysis of their relationship has been widely conducted only by
Acknowledgements
The authors would like to thank Maria E. Milici (Department of Hygiene and Microbiology, University of Palermo, Italy) for the use of laboratory facilities and assistance in microbiological evaluation.
This work was supported in part by the Italian Ministry of University and Research (MIUR), Rome, Italy.
References (56)
- et al.
Fungal load and candidiasis in Sjogren's syndrome
Oral Surg Oral Med Oral Pathol Oral Radiol Endod
(2003) - et al.
Superficial fungal infections in 102 renal transplant recipients: a case–control study
J Am Acad Dermatol
(2003) - et al.
Candidal carriage in the oral cavity of human immunodeficiency virus-infected subjects
Oral Surg Oral Med Oral Pathol Oral Radiol Endod
(2002) - et al.
Cytotoxic drugs, radiotherapy and oral candidiasis
Oral Oncol
(2004) - et al.
Biofilm formation of Candida albicans is variably affected by saliva and dietary sugars
Arch Oral Biol
(2004) Improving dental treatment for the medically complicated patient
Oral Surg Oral Med Oral Pathol Oral Radiol Endod
(2005)A note on prototype theory and fuzzy sets
Cognition
(1982)- et al.
Generation of an intelligent medical system, using a real database, to diagnose bacterial infection in hospitalized patients
Int J Med Inform
(2001) - et al.
Using fuzzy sets to analyze putative correlates between age, blood type, gender and/or race with bacterial infection
Artif Intell Med
(2001) - et al.
Fuzzy modeling in symptomatic HIV virus infected population
Bull Math Biol
(2004)