STRENGTHS AND LIMITATIONS OF THIS STUDY
-
The study employed an advanced statistical model for analysis, which allowed for a more rigorous examination of the data.
-
The fact that the study was a follow-up study has the benefit of being able to demonstrate the temporal relationships.
-
The follow-up duration of the study was long, with a maximum duration of 10 years. This extended period of observation helps provide a more comprehensive understanding.
-
Due to the retrospective nature of the study, it was not possible to investigate the impact of certain sociodemographic, behavioural and clinical factors.
-
Because of incomplete data, risk factors like body mass index could not be considered in the analysis.
Introduction
Diabetes mellitus (DM) is an important public health issue that is approaching epidemic proportions due to lifestyle changes and living situations globally.1 DM affected about 537 million adults globally in 2021, and this figure is estimated to climb to 643 million by 2030 and 783 million by 2045, with low-income and middle-income countries bearing the brunt of the predicted increase in prevalence. In Africa, 24 million, or 1 in 22 adults, had DM in 2021. It is expected that it will be about 55 million by 2045. Ethiopia is one of the top five African countries in terms of the number of diabetics and has an estimated 1.9 million adults with DM.2 3
Diabetes is marked by several long-term implications that affect nearly all systems in the body, but none are more severe than those affecting the foot. Diabetic foot ulcer (DFU) is the most common and lethal complication of DM.4 The International Working Group on the Diabetic Foot defines DFU as ‘a break of the skin of the foot that includes minimally the epidermis and part of the dermis in a person with DM’.5 It is any ulceration, necrosis, gangrene or full-thickness skin defect that develops below the ankle in a diabetic person, regardless of duration.6
Globally, the prevalence of DFU ranges from 3% to 13%, with a pooled global prevalence of 6.3%.7 It affects between 9.1 and 26.1 million patients worldwide each year, with a lifetime risk ranging from 19% to 34%.8 9 DFUs typically heal slowly, necessitating protracted hospital stays or they may never heal, raising the risk of infection, tissue necrosis and gangrene.10 Unhealed ulcers can result in the amputation of toes, sections of the foot or the lower leg.11 DM complications account for 40%–60% of all non-traumatic lower limb amputations globally, with DFU accounting for 80% of these amputations.12 A lower limb or part of a lower limb is amputated due to DFU every 30 s. These silent problems are now becoming a significant cause of disability globally.13 14
DFUs pose substantial health and socioeconomic challenges, harming patients’ quality of life and imposing a significant economic burden on patients, their families, the healthcare system and health economics.15 16 DFUs account for 24.4% of all diabetic healthcare costs and half of all diabetic hospitalisations.17 18
Diabetes foot ulcers affect both developed and developing countries, although the lifetime burden is highest in low-income and middle-income countries due to late diagnosis, lack of awareness and ineffective health services. The burden of DFU is projected to increase further as DM incidence grows across Africa. In sub-Saharan Africa, DFU is a major public health problem that contributes to extended hospitalisation and mortality.7 19 20 In Africa, the pooled prevalence of DFU was 7.2% in 2017 and 13% in 2018, with 12.6%, 16.4%, 16.7%, 4.6% and 11.9% in Northern, Western, Central, Southern and Eastern Africa, respectively. Studies also showed a rising trend in DFU, with a particularly noticeable trend emerging after 2001.7 20 The pooled prevalence of DFU in Ethiopia was 12.98%21 with a cumulative incidence ranging from 12.1% to 17.86%.22–24 In recent decades, Ethiopia has shown a shift in its population’s lifestyle towards urbanisation, and non-communicable diseases such as DM have emerged due to these rapid changes. Alongside this diabetes-related complications, including DFUs, are becoming more common.25
Previous studies showed that various factors such as sex, age, place of residence, type of DM, treatment type, lipid profiles, fasting blood sugar (FBS) level, proteinuria, blood pressure levels, family history of DM, duration of DM and different comorbidities influence the development of DFUs in patients with diabetes.7 20–23 26–31
The prognosis is good if DFU is detected early and appropriate therapy is initiated. However, delays in treatment may have detrimental effects and potentially necessitate the amputation of the foot.11 32 33 Numerous studies conducted on DFU in Africa, including Ethiopia, have been restricted to estimating prevalence from cross-sectional studies and identifying factors through case–control studies. Although there are a limited number of published follow-up studies,22–24 34 35 they are unable to pinpoint the time to development and predictors of DFU. Therefore, this study aimed to assess the incidence of DFU and identify its predictors among patients with DM at a diabetic follow-up clinic in Central Ethiopia.
Methods
Study design, setting and period
A hospital-based retrospective follow-up study was conducted by reviewing the medical records of patients with DM who were on follow-up from 1 January 2012 to 31 December 2022, at Adama Hospital Medical College (AHMC), which is located in Adama City, Oromia regional state, 99 km southeast of Addis Ababa. The hospital has more than 1300 staff, a yearly outpatient flow of 226 000, 500 beds and an admittance rate of 173 patients per week. It serves catchment areas with over five million people and acts as a referral hub for zones and regions nearby.
Population, sample size and sampling procedure
All newly diagnosed patients with DM who were enrolled from 1 January 2012 to 31 December 2022, and had a follow-up at AHMC were included in the study while patients who had DFU at the time of diagnosis for DM, patients whose date of DM diagnosis was not recorded and patients with an unknown date of DFU diagnosis were excluded from the study.
The sample size was determined using STATA V.14 statistical software based on the power and sample size analysis approach by considering covariates significantly associated with the incidence of DFU from previous studies under the following assumptions: Cox proportional hazard model, 95% confidence level, 80% power and 10% withdrawal probability. Therefore, the largest sample size of 416 was found using residence as a predictor {power Cox, hratio (2.3) failprob (0.121) wdprob (0.1)}22 and taking a 10% to compensate for incomplete data, the minimum required sample size was found to be 462.
After identifying the patients who met the inclusion criteria, the sampling frame was created by gathering the patients’ medical registration numbers from the registration book. Then, study participants were chosen by a simple random sampling technique using computer-generated random numbers.
Operational definitions
Diabetic foot ulcer
Non-traumatic skin lesions (partial or full thickness) below the ankle of a person with DM and documented in the patient’s follow-up card.23 36
Event
The occurrence of DFU within the follow-up period.
Censored
Those who did not experience DFU by the end of the study period, who died before experiencing DFU during the study period, or who lost follow-up before experiencing DFU for reasons unrelated to DFU, were considered censored.
Data collection tools and procedure
A structured data extraction checklist was developed in English, as all records were kept in this language. The checklist was developed based on existing DM patients’ medical records and by reviewing other previously conducted similar studies. The medical registration number was used to identify individual patient records, and data were extracted by reviewing follow-up charts and patient cards. Three trained nurses were assigned as data collectors under the supervision of two public health professionals.
Data quality control
To ensure data quality, training was given to all data collectors and supervisors on the tools, all the objectives of the study and the data collection procedure. The data extraction checklist was pretested on 5% of the sample size (n=23) of records before January 2012, and the content validity and adequacy of the checklist were evaluated. During the data retrieval process, close day-to-day supervision was made by both the principal investigator and the designated supervisors to verify the completeness, accuracy and consistency of the data and to ensure ethical standards were upheld.
Data processing and statistical analysis
The data were examined for inconsistencies, coding errors, completeness, clarity and missing values before entry. For data entry, Epi Info V.7.2 was used, and then the data were exported and further analysed by using STATA V.14 software. Descriptive measures such as median with interquartile range (IQR) for continuous variables and frequency and proportion for variables of a categorical nature were used to characterise the study population. The missing data were managed with multiple imputations for creatinine, haemoglobin A1C (HbA1c), triglyceride and total cholesterol levels. The study participants’ outcomes were dichotomised as (code ‘1’) an event (developing DFU) and (code ‘0’) censored, and some continuous variables were categorised for ease of analysis and otherwise utilised as continuous.
The Kaplan-Meier method was employed to estimate the survival time and to compare survival experiences between different exposure groups across time, and survival differences were assessed using a log-rank test at a 5% significance level. The incidence rate of DFUs was calculated for the entire cohort by dividing the total number of incident cases by the total person-months and person-years of follow-up. Because the greatest observed analysis time was censored and the survival curve did not dip below 0.5, the median survival time was not able to be determined. Hence, the restricted mean survival time (RMST) was computed in this study. RMST is regarded as the area under the survival curve up to a particular time horizon and is often a more precise estimate than mean or median survival times, offering a more thorough understanding of survival. It is proposed as a novel substitute measure in survival analysis and could be useful when the proportional hazards assumption cannot be made or when the event rate is small. It provides an alternative to the usual metric of median survival time, and the resultant value indicates the average survival time up to a specific time point.37 38
The proportional hazard assumptions were assessed using graphical methods, the Schoenfeld residual test (both global and scaled), and the presence of time-dependent variables. Cox proportional hazard regression was fitted to identify the predictors of DFU development. The variance inflation factor (VIF) and tolerance were computed to check the existence of multicollinearity, and the goodness of fit of the model was tested using Cox-Snell residuals against Nelson-Aalen’s cumulative hazard function. The predictors with a p<0.25 in the bivariable analysis were fitted in the multivariable survival model. Statistical significance was declared at a level of significance of 5%, and an adjusted hazard ratio (AHR) with a 95% confidence interval (CI) was used to present estimates of the strength of the association.
Patient and public involvement
Patients and/or the public were not involved in the design, conduct, reporting or dissemination plans of this research.
Results
Sociodemographic characteristics
A total of 418 DM patient records were reviewed and included in the final analysis. The median age of the study participants was 47 (IQR=35–58) years. Of the total, 223 participants (53.35%) were male, and the majority, comprising 307 participants (73.44%), resided in Adama (table 1).
Sociodemographic and clinical characteristics of patients with diabetes mellitus followed from January 2012 to December 2022 (n=418)
Baseline clinical characteristics
At baseline, the median FBS level was 224 mg/dL (IQR 190.8–288.5) while the median HbA1c and creatinine levels were 10% (IQR 8.0–11.4) and 0.92 mg/dL (IQR 0.79–1.01), respectively. Of the participants, 342 (81.82%) were patients with type 2 DM, 172 (41.15%) were on oral hypoglycaemic agent (OHA) treatment, 251 (60.05%) had systolic blood pressure (SBP) levels of 140 mmHg and above, and 305 (72.90%) haddiastolic blood pressure (DBP) levels under 90 mm Hg (table 1).
Baseline comorbidity-related status
In this study, a total of 230 patients (55.02%) had no known comorbidities. Meanwhile, 119 patients (28.47%) had 1 comorbidity, 57 patients (13.64%) had 2 and 12 patients (2.87%) had 3 comorbidities. Specifically, out of these comorbidities, 141 patients (33.73%) had hypertension, 44 patients (10.53%) had diabetic neuropathy, 12 patients (2.87%) had experienced a stroke and 5 patients (1.20%) had diabetic nephropathy (figure 1).
Summary of comorbidities among patients with diabetes mellitus followed from January 2012 to December 2022 (n=418). CHD, chronic heart disease; CKD, chronic kidney disease; PAD, peripheral arterial disease.
Incidence of DFU
During the study’s follow-up period, 26 (6.2%) (95% CI 4.1% to 8.6%) patients developed DFU, and 392 (93.78%) patients were censored (online supplemental figure 1).
Supplemental material
Participants in this study were followed for a minimum of 3 months and a maximum of 120 months, with a median follow-up time of 45 months (IQR 21–73). The total person-time of observation was 20 604 person-months or 1716.48 person-years. The overall incidence rate of DFU was 1.26 (95% CI 0.86 to 1.85) per 1000 person-months or 1.51 (95% CI 1.03 to 2.22) per 100 person-years of observation. Since the median survival time was undetermined due to the censoring of the longest observed analysis time, the survivor function fails to get to zero; the restricted mean is the best estimate of survival time in this instance. So, in this study, the RMST was 110.7 (95% CI 107.4 to 114.1) months (figure 2).
Overall Kaplan-Meier survival curve estimate of patients with diabetes followed from January 2012 to December 2022 (n=418).
Comparison of survival experience
The Kaplan-Meier method was applied to compare survival experiences between different exposure groups across time, and the log-rank statistical approach was used to determine if there was a statistically significant variation in survival time. Accordingly, there was a significant difference in survival experiences tailored to the type of DM, the type of treatment, the DBP level and the presence of peripheral arterial disease (PAD).
Significantly longer survival was observed among patients with type 1 DM compared with patients with type 2 DM in the survival experience to develop DFUs. Likewise, patients who were treated with insulin had a longer survival probability to develop a DFU than those who were treated with oral and mixed medication while patients who were treated with mixed medication had a shorter survival probability than those who were treated with insulin and oral medication. A shorter survival probability was also observed among patients with DBP levels of 90 mm Hg or higher and patients with PAD (figure 3).
Kaplan-Meier survival curve showing survival difference between different exposure groups among patients with diabetes, January 2012–December 2022 (n=418). DBP, diastolic blood pressure; DM, diabetes mellitus; OHA, oral hypoglycaemic agent; PAD, peripheral arterial disease.
Predictors for the incidence of DFU
Before fitting the model, the necessary assumptions of the Cox proportional hazard regression were checked using the Schoenfeld residual test, and it was found that all covariates and the full model satisfy the proportional hazard assumption (0.0697–0.9733, with a global test p value of 0.8092. The plot of -ln (-In (survival probability) versus ln (survival time)) was also done for different categories of predictor variables, and the lines were nearly parallel, indicating the proportional hazard assumption was valid. Also, the interaction of each of the covariates with time was assessed, and all of the time-dependent interactions were found to be non-significant.
In the bivariable Cox proportional hazard regression model, type of treatment, DBP level, SBP level, HbA1c, hypertension, neuropathy and PAD were predictors that had crude associations with the development of DFU at a p<0.25. Before fitting the multivariable Cox proportional hazard regression model, the VIF and tolerance were computed to check for multicollinearity. The maximum VIF was 2.38, with a mean VIF of 1.42, and the minimum tolerance value was 0.42. Hence, there was no multicollinearity among the variables. After adjusting for the potential effect of other variables, in the multivariable Cox regression analysis, DBP level, type of treatment and PAD were found to be statistically significant predictors of development of DFU at p<0.05. Accordingly, patients with DM with a DBP level of 90 mm Hg or higher had a 2.91 times greater hazard of developing DFU than those with a DBP level below 90 mm Hg (AHR 2.91, 95% CI 1.25 to 6.77). The hazard of experiencing DFU was 3.24 times higher for patients with DM taking both insulin and OHA as compared with patients taking only OHA (AHR 3.24, 95% CI 1.14 to 9.19), and compared with patients with DM without PAD, those with PAD had a 5.26 times increased hazard of experiencing DFU (AHR 5.26, 95% CI 1.61 to 17.18) (table 2).
Bivariable and multivariable Cox regression analysis for predictors of diabetic foot ulcer among patients with diabetes mellitus, Central Ethiopia (n=418)
After fitting the final model, the goodness of fitness of the model was assessed using Cox-Snell residuals. The graph of Cox-Snell residuals against Nelson-Aalen’s cumulative hazard function closely follows the 45° line, indicating that the model fits the data well (online supplemental figure 2).
Supplemental material
Discussion
DFUs pose a substantial and mounting global healthcare challenge, especially in low-income and middle-income countries where access to comprehensive diabetic care is frequently constrained.39 40 As a potentially lethal complication of DM, DFUs commonly precipitate prolonged hospitalisation, lower limb amputation, and heightened morbidity and mortality rates.1 32 33 41 Thus, this study sought to assess the incidence and predictors of DFU among patients with DM in a diabetic follow-up clinic in Central Ethiopia.
This study found that the incidence rate of DFU was 1.51 per 100 person-years. The cumulative incidence was 6.2%. The RMST was 110.7 (95% CI 107.4 to 114.1) months, implying that the average time for patients with DM to live without developing DFU was 110.7 months. Additionally, DBP level, type of treatment and having PAD were statistically significant predictors of DFU development.
The overall incidence rate of DFU was 1.26 (95% CI 0.86 to 1.85) per 1000 person-months or 1.51 (95% CI 1.03 to 2.22) per 100 person-years of observation. This finding was comparable with a study conducted in referral hospitals in Northwest Ethiopia.22 However, it was higher than studies done in Japan42 and Denmark,43 but it was lower than studies conducted in Seattle, USA44; Wuhan, China30; and Felege Hiwot Referral Hospital, Ethiopia.23 Differences in the duration of the follow-up time and recruiting criteria could be possible reasons for the variations in the incidence rate of DFU. Moreover, discrepancies in the findings could be explained by differences in the availability and efficacy of diabetes care and management programmes, as well as preventive and early detection approaches. It might also be due to sociocultural differences, economic disparities, methodological differences and sample size variations.
During the follow-up period, the proportion of patients with diabetes who developed DFU was 6.2% (95% CI 4.1% to 8.6%). This is in agreement with a global systematic review and meta-analysis of 67 peer-reviewed articles (6.3%), a systematic review conducted in Arab countries (Egypt 4.2%, Jordan 4.65%, Bahrain 5.9%) and a study done in Tanzania (6.8%).7 35 45 But it was higher than a study done in Spain (0.42%)46 and a systematic review conducted in Iraq (2.7%)45 and lower than a study done in Italy (35.5%),47 a systematic review and meta-analysis of 55 articles in Africa (13%),20 a systematic review and meta-analysis of 11 articles in Ethiopia (12.98%),21 and other recent studies conducted in Ethiopia.22 23 The dissimilarity may be attributable to disparities in healthcare systems that aid in the early detection of diabetes and its complications, as well as differences in the quality of care provided to patients with diabetes and variations in sociodemographic characteristics.
In this study, in comparison to patients with DM with a DBP level below 90 mm Hg, those patients with a DBP level of 90 mm Hg or higher had a greater hazard of developing DFU. This finding is similar to a study conducted in Saudi Arabia.48 This can be explained by the fact that higher DBP frequently accompanies overall higher blood pressure and when coupled with the weight-bearing aspect of standing and walking, increased pressure on the foot can cause excessive mechanical stress and damage to the skin and tissues. Elevated DBP can also result in endothelial dysfunction or a reduction in the function of the cells that line the blood vessels. This malfunction reduces the synthesis of nitric oxide, a compound vital to optimal blood vessel dilatation and function. Consequently, blood vessel constriction occurs, limiting blood flow to the feet and raising the risk of foot ulcers. Furthermore, higher DBP is linked to chronic low-grade inflammation, which can cause tissue damage and impede the natural healing process. So, in patients with DM, elevated DBP coupled with diabetes-related inflammation can further aggravate the risk of foot ulcers.49 50
The type of treatment was another independent predictor of DFU development. In this study, the hazard of developing DFU among patients taking a combination of insulin and oral medication was 3.24 times greater compared with patients taking OHA alone. This finding is consistent with studies from Pakistan,51 Iraq52 and North India.53 This may be due to patients taking multiple medications may encounter shifts in blood sugar levels as a result of the synergistic effects of such therapies. These fluctuations in blood sugar levels can harm blood vessels and hinder wound healing, increasing the risk of DFU.54 55 Further, managing multiple medications and treatment regimens might make it challenging to maintain medication compliance and practice good diabetic self-care. Inconsistency in medication usage or poor adherence to foot care practices could exacerbate the risk of foot ulcers.56 57
PAD is a condition characterised by the narrowing or blockage of the blood vessels that supply the extremities, notably the legs and feet. It is an important cause of non-healing ulcers, lower limb amputation and mortality, especially among diabetics.58 In concordance with findings from studies done to assess the association between PAD and DFU in patients with DM in Peru,31 Pakistan,59 Saudi Arabia,28 Egypt60 and a retrospective cohort study in Ethiopia,22 this study showed patients with DM with PAD had a 4.83 times higher hazard of experiencing DFU compared with those without PAD. The justification could be that patients with PAD have constricted blood vessels, reducing blood flow to the legs or lower extremities. Ischaemia, caused by reduced blood flow, can induce nerve and other tissue damage, predisposing patients with DM to DFU. Furthermore, decreased blood flow to the foot jeopardises the supply of oxygen and nutrients required for tissue health and repair. In diabetics who already have prolonged wound healing owing to elevated blood sugar levels, PAD aggravates the condition and lengthens the healing period, increasing the risk of DFU.61 62
When interpreting our study’s findings, it is important to consider certain limitations. Owing to the retrospective nature of the study, we could not explore the influence of certain sociodemographic, behavioural and clinical factors. Moreover, the presence of incomplete data hindered the inclusion of crucial risk factors such as body mass index in our analysis.
Conclusions
In this study, the incidence rate of DFU was relatively high compared with previous studies. DBP level, type of treatment and PAD were independent predictors of DFU development. Healthcare providers should give greater attention to patients with diabetes with elevated DBP levels, patients taking combined medications and patients with comorbidities during their entire follow-up period. Due attention should also be given to advising patients on high blood pressure, combined medications, and the comorbidities of DM at the initiation of DM treatment.
Data availability statement
Data are available on reasonable request. All data and materials are available from the corresponding author without undue reservation.
Ethics statements
Patient consent for publication
Ethics approval
Ethical approval was secured from the Institutional Ethical Review Board (IERB) of Adama Hospital Medical College (Ref no. AHMC/MPH/27/6/2015). Since it is a retrospective study of a medical record review, individual consent was not required. The IERB waived the informed consent requirement. Confidentiality and anonymity were maintained during all phases of the study. Data were held on a secured, password-protected system and all the procedures in the study were conducted based on the principles of the Declaration of Helsinki.
Acknowledgments
The authors like to acknowledge Adama Hospital Medical College for providing all necessary assistance. We also are thankful for the data collectors and supervisors.
This post was originally published on https://bmjopen.bmj.com