The role of amputative and non-amputative foot deformities severity in the risk for diabetic ulceration classification systems building: a cross-sectional and case-control pilot investigation

Introduction Foot deformities and amputations are parameters that have been studied as risk factors for diabetic foot ulceration (DFU). However, inclusion of “foot deformities” and “amputations” in a single, broad variable and with reference to the severity of these deformities, may better characterize subjects who are prone to develop DFU. Methods The objective of the study was the examination of amputative and non-amputative foot deformities severity as risk factor for DFU in relation with the other established risk factors. A cross-sectional and case-control study was conducted from October 2005 to November 2016. One hundred and thirty-four subjects with type 1 and 2 diabetes, with and without active foot ulcers, participated. A structured quantitative interview guide was used. Univariate logistic regression analysis for the literature's established risk factors was performed, as well as for two versions of the “amputative and non-amputative foot deformities severity” variable. Subsequently, multivariate logistic regression analysis (MLRA) for three models and receiver operating characteristic (ROC) curve analysis were carried out. Results From the MLRA, only PAD (peripheral arterial disease) was significant (OR 3.56, 95% CI 1.17-10.82, P=0.025 and OR 3.33, 95% CI 1.02-10.08, P=0.033). Concerning the ROC curve analysis of the models, the one with the three categories amputative and non-amputative foot deformities severity variable, had the greatest area under the ROC curve (0.763, P<0.001). Conclusion A united variable for lower extremity amputations and other foot deformities with reference to their severity, could be more helpful to the clinicians in identifying patients with diabetes at risk for foot ulceration.


Introduction
The development of ulcers, of catalytic etiology either intrinsic (e.g resulting from high plantar pressures due to prominent metatarsal heads) or extrinsic (e.g. resulting from a pebble during walking shoeless) [1], at feet of persons with diabetes (diabetic foot ulceration, DFU), can bring about serious complications both individually (amputation-related disability and increased mortality) and socially (economic burden of the health systems) [2]. According to the epidemiologic studies, the DFU risk factors that predominantly have been identified include peripheral neuropathy, peripheral arterial disease (PAD), structural foot deformities (hammer toes, claw toes, etc) and the history of amputation and/or previous ulceration [1,2].
The terms amputative and non-amputative, concerning the separation of amputations from the rest foot deformities in people with diabetes, are more accurate in relation to a potential use of the terms "extrinsic" for amputations and "intrinsic" for deformities such as claw toes or prominent metatarsal heads. A non-amputative deformity could have a cause outside of diabetic neuropathy, which is an intrinsic factor (e.g. hammer toes can be a result of trauma or inappropriate shoes) [14]. Foot deformities and their severity are parameters that have been studied in the past as risk factors for ulceration development in patients with diabetes [6,9]. Although, the terms foot deformities and amputations are confusing in the literature with glaring example the recent IWGDF definitions and risk classification system of 2015 [5], in which amputations once is included in the term foot deformity (IWGDF definitions, p. 17), while another time is not (Table 1, p. 18). Since amputations are also deformities, the administration of foot deformities as a broad variable, including both amputative and non-amputative ones, is more precise.
Severity of foot deformities only recently has been studied, precisely and with breadth, as a united variable including both amputative and non-amputative ones [13].
No study yet has examined the amputative and non-amputative foot deformities severity as risk factor for DFU in association with the established risk factors (peripheral neuropathy, PAD, history of previous ulceration).
The aim of this study was the examination of amputative and nonamputative foot deformities severity as risk factor for DFU in relation with the other established risk factors, as well as of the participants' sociodemographic and clinical characteristics.

Study design
The study was a cross-sectional, case-control research.

Setting
The research came about at three diabetic foot clinics of general hospitals and one wound unit of a special hospital in a large capital city. Ethics approvals were granted by the hospitals' scientific committees.

Subjects
The study participants were individuals with type 1 and 2 diabetes and with or without foot ulcers. Patients with cognitive disturbances were excluded from the study.

Recruitment
One hundred and thirty-four patients were conveniently approached by the head investigators during their scheduled first or subsequent visit to the healthcare facilities, from October 2005 to November 2016. The sample size was calculated implementing approximately the Garson's [15] rule of thumb whereby the number of cases in the smaller of the two binary outcomes in binary logistic regression divided by the number of predictor variables should be at least 20 [15]. All participants were enrolled after providing written informed consent.

Data collection
For the collection of the data, a structured quantitative interview guide with closed-ended questions was used. The principal researchers interviewed one-on-one each patient gathering and recording demographic and clinical data.

Measurements
The parameters that were measured were related to:  [20,21].

Data analysis
At first, because there were only two observations for the severe category of the Waaijman et al. [13,19] variable from the small pilot sample and of the fact that the recommended smallest of the classes of the depended variable in a regression model is at least 10 events per parameter [15], the amputative and non-amputative foot deformities severity parameter from a four categories variable (none, mild, moderate, severe), yielding high logistic coefficients [15], was altered to a three classes one (none, mild and moderate/severe) with the last two categories combined and following this to a two classes one (none/mild and moderate/severe) with the first and last two categories combined. For the multivariate regression analyses, the "enter"variable selection method was used and 5% probability criterion was set for the variables to enter the models. After the multivariate regression investigation of the aforesaid variables, and considering that the research purpose was prediction [15], a ROC (receiver operating characteristic) curve analysis for compering the yielded models took place.  Table 1.

Inferential
The univariate logistic regression analysis, in terms of the variables that were involved in the three models ( Table 3, Table 4).
In regards to the model 1 multivariate logistic regression analysis (MLRA), none significant variable was yielded (  (Figure 1, Table 5).

Discussion
Even though this was a small-sized pilot study, given the fact that coping with a problem as common as diabetic foot ulceration necessitates a larger cohort, for testing the feasibility of the methodology that was chosen [22], it managed to bring in useful results.
The most important finding of the study was the fact that both models 2 and 3, with the three and two categories amputative and non- By the univariate logistic regression analysis, the parameters of peripheral neuropathy, PAD, amputative foot deformities and history of previous ulceration, in concordance with the literature [8,9,[23][24][25][26] were discovered to be significantly associated with the presence of active foot ulceration.
In terms of the sociodemographic and the clinical characteristics that were not examined in the context of inferential analysis, by the descriptive analysis, the prevalence of wearing appropriate footwear (53.3%) was in consonance with the literature [27][28][29][30], in which the prevalence in question was calculated to be 52% [21].

Competing interests
The authors declare no competing interests.

Authors' contributions
Aristomenis Kossioris collected and analyzed the data, as well as he wrote the bulk of the article. Nicholas Tentolouris helped with the data collection, the writing of the abstract and he consulted regarding the methodology and statistical analysis. Chariclia V Loupa helped regarding the data collection and she contributed to tables writing.
Minos Tyllianakis contributed to writing the Methods section, to formatting the manuscript, as well as he had the overall supervision of the project. All authors read and approved the final manuscript.