The study took place in the Department of Internal Medicine of the Yalgado OUEDRAOGO Teaching Hospital. The internal medicine department within the hospital is a multidisciplinary department that mainly deals with metabolic and endocrine pathologies and Human Immunodeficiency Virus (HIV) infection.

This was a descriptive and analytical cross-sectional study combining retrospective data collection from previous years on patients’ follow-up and prospective data collection of missing information in their respective medical records. The study population consisted of a cohort of diabetes mellitus patients, at least 18 years of age, followed and treated in the Department of Internal Medicine of the Yalgado OUEDRAOGO Teaching Hospital between January 1, 2010 and December 31, 2018.

Included in this study were T2DM patients followed and treated for at least one year with at least two HbA1c measurements within the last year of follow-up. Not included in the study were patients who had made a change in diabetes therapy within the last 3 months and also those who did not respond to the call to complete the missing information in the medical record.

This was a systematic random sampling from all the patients who were followed in the internal medicine department between January 1, 2010 and December 31, 2018, following a draw step of 10.

The questionnaire was developed by the authors and their peers in the Department of Internal Medicine of the Yalgado OUEDRAOGO Teaching Hospital (Appendix 1). We also used Morisky’s assessment questionnaire [26] in order to appreciate the quality of the treatment’s compliance (Appendix 2).

Prolonged poor control of diabetes mellitus was reported in all patients with at least one of the outcomes of the various HbA1c measures in the last year of follow-up having a threshold above 7%. Alternatively, fructosamine was used to assess poor glycemic control when at least one of the outcomes of the various fructosamine measurements in the last year of follow-up had a threshold above 290 micromoles/L. Adherence was measured using Morisky’s questionnaire [26].

Lipid fraction abnormalities have been defined by adopting the 2017 standards of the French National Authority for Health on lipid assessment [27] (Appendix 3). Microangiopathy was defined by the presence of either neuropathy and/or nephropathy and/or retinopathy. Neuropathy was defined by an abnormality of the filament test and/or an abnormality of the tuning fork test and/or reflex abnormalities. Nephropathy was defined by measurement of 24-hour proteinuria. The cut-off value for 24-hour proteinuria was 300 mg/24 h. Retinopathy was retained based on the results of fundus examinations with an ophthalmoscope. Macroangiopathy was defined by the presence of coronary artery disease, cerebrovascular disease and arteriopathy of the lower limbs. These findings were systematically searched for in the patients’ follow-up medical records. Arteriopathy was defined both on clinical criteria (intermittent claudication, decubitus pain, absence of both foot pulses, gangrene, ischemic ulcer) and imaging criteria (systolic pressure index less than 0.9, systolic pressure of the big toe greater than or equal to 1.3 accompanied by medicalcosis, calcifications, aneurysm and a perivascular environment appreciates hemodynamics for therapeutic purposes). The monthly gain consisted of the patient’s total income plus donations. Monthly expenses (direct cost of management) for diabetes mellitus included expenses for: travel, consultation, testing, and purchase of medication and capillary blood glucose strips.

Data collection covered the period from February 15, 2020 to May 15, 2020. The collection tool was an anonymous questionnaire. Data collection was done in conjunction with follow-up consultations by a physician based on a survey form. The use of medical records made it possible to extract clinical data (the conditions leading to the identification of diabetes mellitus, time since diabetes mellitus diagnoses, presence of micro- or macro-angiopathy) and paraclinical data (blood glucose and HbA1c results) of previous years follow-up. The individual interview allowed the completion of certain missing information, in particular drug compliance based on items from Morisky’s assessment questionnaire [26], certain sociodemographic and anthropometric characteristics and the financial status of the patients. Sociodemographic data (age, gender, level of education, average monthly income, marital status and residence), anthropometric data (weight, height, waist circumference and body mass index [weight divided by height squared]), and behavioral factors (alcohol and tobacco consumption) were collected.

Confidentiality during the interview and respect for the anonymity of the questionnaire user were observed throughout this study. All personnel directly or indirectly involved in the data collection were health workers who were bound by strict professional secrecy. Free and informed verbal consent was obtained from patients prior to their inclusion in the study. Patients’ refusal to participate in the study did not preclude their diabetes management and follow-up at the center. Personal information about each patient was coded with a number that did not identify the patient when the study results were published.

Statistical analyses were performed using Statistical Product and Service Solutions (SPSS) version 20. Quantitative variables were presented as means or medians with indicators of dispersion (standard deviation and percentile). Qualitative variables were described by the frequencies of the different modalities.

To determine the risk factors associated with poor glycemic control, we used a logistic regression model. A bivariate analysis was performed to analyze the relationship between each variable and poor glycemic control. A final model was then determined by multivariate logistic regression. All variables associated with poor glycemic control in bivariate logistic regression with a p-value less than 0.20 were included in the full model. Then, we used a stepwise descending method to obtain the final model, which was then tested by a Receiver Operating Characteristic (ROC) curve. The significance level was set at 5%. Raw and adjusted Odds-Ratio (OR) and 95% confidence interval (CI) were presented.

A total of 270 patients were included in this study. Poor glycemic control of diabetes mellitus was observed in 73.70% or approximately two-thirds of the patients in the cohort. The mean age of patients was 55.97 years (standard deviation [SD]: ±11.52 years) and the sex ratio (male/female) was 0.6 in favor of female. More than half of the population had a formal education (59.26%). One out of two patients (55.92%) was overweight. More than one out of two patients (67.04%) had at least one underlying health condition recorded in their medical history. History of arterial hypertension was the most represented (41.85%) condition. More than two thirds (85.19%) of participants were supported by their families in the management of their diabetes mellitus. The average expenditure for diabetes mellitus care was 55.82 USD (SD: 28.25) per month. Table 1 shows the distribution of patients in the study according to their general characteristics.

The average time period since diagnostic of diabetes mellitus was 5.85 years (SD: ±5.15 years). At least one complication of diabetes mellitus was found in more than half (68.15%) of the population. Almost all (94.81%) the patients in the study were receiving a pharmacological treatment for T2DM. The mean number of tablets taken was 2.94 (SD: ±1.56) per patient. Table 2 shows the distribution of study patients by their clinical status of diabetes mellitus.

In multivariate analysis with bivariate logistic regression, six factors were independently associated with prolonged poor diabetes mellitus control. These were no formal education (OR = 8.34, 95% CI [1.97 - 35.22]; p < 0.01); family support for diabetes mellitus management (OR = 0.65, 95% CI [0.45 - 0.94]; p = 0.02); presence of abdominal obesity (OR = 2.27, 95% CI [1.08 - 4.77]; p = 0.03); a history of hospitalization (OR = 7.39, 95% CI [2.97 - 18.39]; p < 0.01); poor adherence to antidiabetic treatment (OR = 2.97, 95% CI [1.42 - 6.18]; p < 0.01); presence of microangiopathy (OR = 5.05, 95% CI [2.36 - 10.81]; p < 0.01).

Table 3 and Table 4 show respectively the relationship between the general characteristics of the study population and prolonged poor diabetes mellitus control and the relationship between diabetes mellitus characteristics and prolonged poor control.

The model specification was verified using the Roc statistical method. Figure 1 shows sensitivity and specificity of factors associated with prolonged poor control of T2DM.

The area under the curve was 0.88, which reflects a very good sensitivity and specificity of these factors associated with poor long-term control of T2DM.

Approximately two-thirds (73.70%) of the patients in our study had prolonged poor diabetes mellitus control. This proportion was similar to that reported by Kibirige in Uganda (73.5%) [28] and Omar in Sudan (71.9%) [29]. It was lower in the studies by Tapsoba in Burkina Faso (57.5%) [23], Souliotis in Greece (57.1%) [30] and Cai in China (31.5%) [31]. In contrast to studies by Tekalegn in Ethiopia (80%) [32], Hai in Pakistan (81.6%) [15] and Rahman in Bangladesh (82%) [33] which found a higher prevalence.

Abbreviations: OR: odds ratio; CI: confidence interval; HIV: Human Immunodeficiency Virus; USD: United States Dollar.

Abbreviations: OR: odds ratio; CI: confidence interval; T2DM: type 2 diabetes mellitus.

Several factors could explain these differences in proportion. For example, in Cai’s study in China, where the prevalence of poor diabetes mellitus control was lower than in our patients, the assessment of diabetes mellitus control was done in patients at the beginning of their therapeutic education [31]. It has been documented that the profile of diabetes mellitus control tends to deteriorate overtime [8] [13] [15] [18] [33].

Similarly, in the Da Vico study, where the proportion of poor diabetes mellitus control was 36.7% at 12 months after the diagnostic of diabetes mellitus, all patients had received therapeutic education [34]. Our high proportion of poor diabetes mellitus control could be explained by the low completion rate of therapeutic education in the patient population, 24.3% [35] versus 100% [34]. Indeed, therapeutics education creates conditions for improved patient compliance and follow-up.

Compared to the study that was conducted in Burkina Faso [23], the endpoints for diabetes control differed: mean HbA1c versus normal HbA1c on a continuous basis. The greater difficulty in maintaining normal HbA1c on a continuous basis may explain why there were fewer patients in control than in the previous study in which HbA1c could vary over time.

Smoking was not a factor associated with prolonged poor diabetes mellitus control in our study (OR = 2.08, 95% CI [0.69 - 6.26]; p = 0.71). Some studies found contrary results [19] [43] [45] [46]. According to Ohkuma, HbA1c levels increase proportionally with the daily amount and duration of tobacco exposure [47].

In our study, alcohol consumption was not a factor associated with prolonged poor diabetes mellitus control (OR = 0.89, 95% CI [0.51 - 1.58]; p = 0.71). Achila found opposite results [10]. Wiss in his study reported that alcohol reduced the HbA1c level, which may indicate a beneficial effect on glycemic control [48]. This study concluded that moderate and/or heavy alcohol consumption led to a significant reduction in the HbA1c level, highlighting the correlation between the amount of alcohol and the HbA1c level [31].

In our study, poor medication adherence was associated with prolonged poor diabetes mellitus control (OR = 2.97, 95% CI [1.42 - 6.18]; p < 0.01). This observation was found in many other studies [18] [33].

A study conducted in Burkina Faso reported that the main reasons for non-compliance with diabetes mellitus medication treatment were oversight and the geographical and financial inaccessibility of patients [49]. To these, we can add:

­ Frequent interruptions in therapy due to denial of the chronicity of the illness,

­ Work overload leading to a communication problem and having a negative impact on the doctor-patient relationship and trust. One meta-analysis found that physician consultation time in developing countries was less than 5 minutes in 18 of the 67 countries included in the study [50]. The patient’s needs and concerns cannot be addressed in this amount of consultation time.

Physical activity practice was not a factor associated with prolonged poor diabetes mellitus control in our study (OR = 0.67, 95% CI [0.38 - 1.17]; p = 0.16). Studies have found similar results [16] [18]. Others, however, found contrary results [33] [39] [43] [51].

This result seems unexpected. Indeed, several meta-analyses and systematic reviews have clearly demonstrated the benefits of physical activity practice in terms of glycemic control.

In our study, neither the possession of a glucometer (OR = 0.86, IC 95% [0.5 - 1.5]; p = 0.61), nor the practice of self-monitoring of blood glucose (OR = 0.93, 95% CI [0.54 - 1.60]; p = 0.81) were factors associated with diabetes mellitus imbalance. These data were contrary to those in the literature [19] [33].

Our result could be explained by the lack of optimization of the self-monitoring of blood glucose device (difficult access to strips, inefficient use, less reactivity or inability of our patients to self-manage) in our context.

Patients who were supported by their families were protected from prolonged poor diabetes mellitus control in our study (OR = 0.65, 95% CI [0.45 - 0.94]; p = 0.02). Similar results have been reported by some authors [15] [19].

This support from the family would help the patient to have a less contemptuous view of the disease. The family could also be a source of motivation for the patient through a personalized and often negotiated self-monitoring of blood glucose.

Time since diabetes mellitus diagnostic was not associated with prolonged poor diabetes mellitus control in our study (OR = 1.05, 95% CI [0.52 - 2.15]; p = 0.88). Some authors found similar results [23] [30]. Other studies found opposite results [8] [13] [15] [19] [33] [40] [43] [50] [52].

From a pathophysiological perspective, one would expect poor control of diabetes mellitus due to the continuous decline in β cell functioning characteristic of T2DM and the difficulty in balancing diabetes mellitus over time. However, data modeling of large intervention studies has not been able to hypothesize a threshold value for the age of diabetes mellitus [1] [2] [3] [4].

Patients with a hospitalization history for an acute diabetes mellitus-related event had a higher risk of prolonged poor diabetes control (OR = 14.79, 95% CI [6.44 - 33.96]; p < 0.01). In fact, according to one study, these hospitalizations occurred in patients who were generally unobservant, with already complicated diabetes mellitus showing HbA1c levels greater than 10.61% vs. 9.07% respectively before and after hospitalization [53].

Microangiopathy (OR = 8.85, 95% CI [4.64 - 16.90]; p < 0.01) or macroangiopathy (OR = 2.47, 95% CI [1.40 - 4.34]; p < 0.01) were factors associated with prolonged poor diabetes mellitus control. Authors found similar results [1] [2] [3] [4]. Demoz, on the other hand, reported opposite results [18]. In the literature, poor control of diabetes mellitus is thought to be associated with the progression of the disease to complications. However, the association of microangiopathy with poor glycemic control in our study seems unexpected. Perhaps it was necessary to correlate fasting blood glucose measurements to better appreciate this association rather than normal HbA1c on a continuous basis.

The use of oral antidiabetic drugs alone was not associated with prolonged poor diabetes mellitus control in our study (OR = 0.93, 95% CI [0.5 - 1.71]; p = 0.82). Our results were contrary to those of a study from Algerian [13]. On the other hand, the use of combination oral antidiabetic drugs was associated with prolonged poor diabetes mellitus control (OR = 11.33, 95% CI [2.89 - 44.33]; p < 0.01). Our results were similar to those of the Algerian study [13]. The explanation may stem from therapeutic inertia related to the patient and/or the practitioner. Indeed, in practice, many dual therapy patients remain unbalanced but are reluctant to prescribe insulin; oral triple therapy is an option that is very often inaccessible in our context.

It is important to note that the use of combination of oral antidiabetic drugs is associated with poorer compliance due to the number of tablets and the number of doses to be taken. This could lead to a move towards prescribing a single antidiabetic drug.

Insulin use alone (OR = 24.29, 95% CI [5.54 - 106]; p < 0.01) or in combination with oral antidiabetic drugs (OR = 14.66, 95% CI [2.43 - 88.48]; p < 0.01) was a factor associated with prolonged poor diabetes mellitus control in our study. These results were similar to those of other studies [13] [16] [19] [30].

This result could be explained by the insulin recovery, in a context of prolonged diabetes mellitus condition that is therefore difficult to balance. In addition to this would be the problem of adherence to insulin therapy, its cost and the difficulties of conservation (disruptions of the cold chain). It is also known that the balance of diabetes mellitus tends to deteriorate with the duration of follow-up [8] [13] [15] [18].

Patients on dietary intake measures alone were protected against prolonged poor diabetes mellitus control in our study (OR = 0.08, 95% CI [0.02 - 0.3]; p < 0.01). Such an observation was made by Tapsoba in Burkina Faso [23]. On the other hand, our results were contrary to those of an Algerian study [13].

Our result can be explained by the following treatment approach: when a patient in his last year of follow-up is still on hygienic dietary measures alone, this diabetes mellitus condition is most often balanced and there would be no need to prescribe medication.