Diagnostic and Evolutionary Aspects of Inaugural Diabetic Ketoacidosis in the Dakar Hospital Setting: A Descriptive and Analytical Cross-Sectional Study in the Endocrinology-Metabolism-Nutrition Department of CHN de Pikine ()
1. Introduction
Diabetic ketoacidosis (DKA) is the most common acute metabolic complication of diabetes [1], marked by ketone buildup in the blood and often accompanied by acidosis [2]. It is a critical metabolic emergency with a mortality rate below 5% in recent decades, which varies depending on the patient’s condition and the timeliness of therapeutic intervention [3] [4]. DKA can be the initial sign of diabetes in 15% - 30% of adults and up to 40% of children with type 1 diabetes [5] [6]. Although it is historically linked to type 1 diabetes, it is increasingly observed in adults with type 2 diabetes, especially in sub-Saharan Africa [7] [8]. This change has led to the inclusion of “ketosis-prone diabetes” (KPD) syndromes in the updated 2018 classification [6] [9]. Recognizing “atypical diabetes” is crucial, as its management and progression differ from classical diabetes [7]. While ketosis-prone diabetes appears in various populations, detailed reports from sub-Saharan Africa remain limited [7]. The short-term progression and prognosis of diabetic ketoacidosis vary across centers, affecting durations of ketosis resolution and hospital stay [10]. This study aims to identify the epidemiological profile, diagnostic features, and progression of initial diabetic ketoacidosis in a Dakar hospital.
2. Patients and Methods
This cross-sectional study, conducted from January 1, 2020, to January 31, 2021, included all patients with diabetic ketoacidosis hospitalized in the Endocrinology-Metabolism-Nutrition department at the National Hospital Center of Pikine. The diagnosis of initial diabetic ketoacidosis was based on specific criteria:
Capillary hyperglycemia over 2.5 g/L (measured with an Accu-Chek glucometer).
Ketonuria exceeding two crosses (determined by Ketodiastix strips) and/or pH < 7.
Patients not previously diagnosed with diabetes.
Pregnant women were excluded.
For each patient, the following parameters were studied:
Epidemiological Data: This includes age, gender, and lifestyle.
Clinical Data: Patient data included medical and surgical history, family history of diabetes, physical condition, presence of Kussmaul breathing, consciousness disturbances, cardinal symptoms, time to consultation before admission, vital signs, body measurements, and other clinical findings.
Paraclinical Data: First-line tests included anti-GAD and/or anti-IA2 antibodies, alongside HbA1c, blood count, inflammatory markers, lipid profile, and blood ionogram.
Therapeutic and Evolutionary Data: Key metrics included ketosis correction time, total rapid insulin doses used, metformin prescription during or soon after hospitalization, mortality, complications (e.g., hypokalemia, venous thromboembolism, infections), hospital stay length, and six-month glycemic status, including HbA1c, average glucose levels, and hypoglycemia incidence.
Diabetes Typing: Diabetes classification relied on anti-GAD and/or anti-IA2 antibody tests or clinical and evolutionary factors (e.g. patient’s age, morphotype, symptom onset, family history, treatment response, and exogenous insulin needs). Diabetes classifications included:
Type 1 Diabetes (T1D): Age under 40, thin or normal BMI, no family history.
Type 1b/Atypical/Ketosis-Prone Diabetes (KPD): Remission within 14 weeks post-hospitalization, insulin cessation due to hypoglycemia, HbA1c ≤ 6.5% for three months with oral medication or diet, no complications at diagnosis.
Type 2 Diabetes (T2D): Age over 40, family history of diabetes, history of macrosomia in women, complications at diagnosis.
Other Diabetes Types: Based on ADA guidelines.
Data Collection and Analysis
Data were recorded on a pre-established form and analyzed using SPSS version 18.
Descriptive Analysis: Calculated frequencies and proportions for qualitative variables and means with standard deviations for quantitative variables.
Analytical Analysis: Cross-tabulations, Pearson Chi-square, or Fisher’s exact test was used for frequency comparisons. Variance analysis tested means, with significance set at p < 0.05. Grouped parameter analysis compared findings with existing literature in the discussion.
The text processing was done using Word XP Professional.
3. Results
Descriptive Study
Sociodemographic Data
In this study, 54 patients were hospitalized for inaugural diabetic ketoacidosis (DKA). The prevalence of inaugural DKA among diabetic patients admitted during this period was 17.1%. The mean age of patients was 38.54 years (range: 16–60 years). There was a male predominance, with a sex ratio of 1.16. Smoking was noted in 6 patients, and 24 patients did not engage in physical activity. Hypertension was reported in seven patients, five of whom had poor control, and two were on treatment. Two patients were undergoing corticosteroid therapy. Psychiatric follow-up was reported in two patients: one receiving haloperidol and anticholinergic therapy, and the other receiving haloperidol and chlorpromazine. Two female patients reported a history of macrosomia. A family history of diabetes was found in 35 patients, with fathers being the most commonly affected (50%), followed by mothers (44.11%).
Among clinical signs of DKA, severe acidosis with altered consciousness was present in 14.9% of cases (13% with confusion and 1.9% in a coma). Kussmaul breathing was observed in 13% of patients upon admission. The cardinal symptoms of diabetes were present in 92.4% of cases. The average time from the onset of cardinal symptoms to consultation was 14.2 days. Gastrointestinal symptoms were present in 18 patients (31.5%), with 11 patients (20.4%) reporting vomiting associated with abdominal pain. The average capillary blood glucose level was 3.43 g/L. Glycosuria (++ level) was observed in 57.4% of cases, while ketonuria (++ level) was noted in 40.7%.
Additional clinical signs were observed in 24 patients. Infections, identified as the precipitating factor in 50% of cases, were mostly soft tissue-related, followed by urinary tract, foot infections, and pneumonia. No precipitating factors were found in the remaining cases, and no cardiovascular events were reported (Table 1).
50 patients did not report fever at admission, and 27 presented with tachycardia. The mean BMI was 25.27 kg/m2 (range: 17.2 - 39.5 kg/m2), and half of the patients had abdominal obesity. The mean waist circumference was 86.9 cm (range: 56 - 152 cm) (Table 2).
Table 1. Distribution of patients based on urine test results.
Urine Test |
Cases (N = 54) |
Percentage (%) |
Glycosuria |
++ |
31 |
57.4 |
+++ |
17 |
31.5 |
++++ |
6 |
11.1 |
Ketonuria |
++ |
22 |
40.7 |
+++ |
20 |
37.0 |
++++ |
12 |
22.2 |
Table 2. Distribution of patients by infection site and identified pathogens.
Infection Site |
Cases (N=27) |
Percentage (%) |
Soft tissue |
7 |
26.0 |
Urinary tract |
6 |
22.2 |
Lungs |
5 |
18.5 |
Diabetic foot |
4 |
14.8 |
Blood |
2 |
7.4 |
Meninges |
2 |
7.4 |
Gastroenteritis |
1 |
3.7 |
Among non-specific findings, anemia occurred in 10 patients (18.5%). Leukocytosis with neutrophilia was noted in 24 patients (44.4%) and linked to elevated CRP in 19 cases. Acute kidney injury, resolved by rehydration, was seen in 21 patients (38.9%). At admission, 25 patients (46.3%) had dysnatremia, 12 (22.3%) had dyskalemia, and 6 showed hyperosmolarity with DKA.
In microbiology, urinary cultures were positive in 6 cases, blood cultures in 1, malaria detected in 1, pus cultures in 2, and COVID-19 PCR positive in 1 patient.
For glycemic control, the average HbA1c was 12.2% (range: 7.8% - 16.9%). Average LDL was 1.21 g/L, HDL was 0.40 g/L, and triglycerides were 1.61 g/L. Mixed dyslipidemia appeared in 14 patients (25.9%). Troponin tests were negative in all patients with repolarization abnormalities. Radiographic and ECG anomalies were found in 4 patients; the ECG showed 3 with repolarization disorders and 1 with necrosis evidence.
Patients with severe acidosis and altered consciousness were managed in intensive care units using insulin therapy via an infusion pump. Regular insulin was administered, starting at a rate of 10. The average total insulin dose was 442.3 IU (range: 80 - 1800). Twenty-nine patients (53.7%) received total doses between 100 and 400 IU of insulin. The average time for ketosis compensation was 34.6 hours (range: 4 - 172), with twenty patients (37%) compensating between 10 and 40 hours.
Rehydration was performed in all patients. The volume and type of solution varied depending on hydration status and electrolytic parameters. Rehydration began with a sodium chloride solution at a dose of 1 liter in the first hour, then was reduced in the following hours. Once the blood glucose level fell below 250 mg/dl, the sodium chloride solution was replaced with a 5% glucose solution.
Antibiotics were given to 38 patients, with 16 (42.1%) receiving ceftriaxone and 15 (39.4%) amoxicillin-clavulanic acid. Metformin was prescribed to 23 patients (43.6%) during hospitalization. Prophylactic anticoagulation was started in 38 patients (70.3%), and statins were given to those with mixed dyslipidemia. The COVID-19 positive patient was moved to an epidemic treatment center. All patients received therapeutic education.
The average length of hospital stay was 7.6 days (range: 3 - 29). Regarding complications, two patients developed hypokalemia after insulin therapy and received oral potassium supplementation. One patient experienced worsening consciousness disturbances, requiring transfer to the intensive care unit, and subsequently died. Among the patients with diabetic foot, three had poor outcomes. One diabetic foot, classified as 2D according to the Texas classification, progressed to necrosis of the 4th toe, leading to an amputation. Another diabetic foot, also classified as 2D, progressed to gangrene of the posterior and lateral aspects of the forefoot, necessitating debridement. The amputation site developed wound suppuration with necrosis of the stump, requiring a reoperation with leg amputation.
The average total insulin dose at discharge was 53 IU (range: 24 - 104). Regarding insulin therapy continuation at discharge, 51 patients received insulin therapy. Regular premixed insulin was prescribed for 41 patients, and premixed insulin analogue for 6 patients. Two patients were placed on a basal-bolus regimen with long-acting analogue insulin (glargine) and rapid-acting analogue insulin (aspart). Two patients were placed on a basal-bolus regimen with regular premixed insulin and regular rapid-acting insulin. Two patients refused insulin therapy and were given a combination of oral antidiabetic medications.
Regarding glycemic status at 6 months, 17 patients (31%) were lost to follow-up. Insulin therapy was discontinued in 22 patients who switched to oral antidiabetic medications.
Fifteen patients underwent antibody testing. Three tested positive, and twelve tested negative. According to WHO classification, the patients were categorized as follows (Figure 1):
Figure 1. Diabetes typing according to the WHO classification.
Type 1 diabetes was diagnosed in 16.7% of patients.
Ketosis Prone-diabetes was diagnosed in 22.2% of patients.
Type 2 diabetes was diagnosed in 57.4% of patients.
Corticosteroid induced diabetes was diagnosed in 3.7% of patients.
Table 3. Cross-tabulation between consciousness disorders and patient characteristics.
Patient Characteristics |
No Consciousness Disorder |
Obnubilation |
Coma |
P-value |
Average Ketoneuria (cross) |
2.76 ± 0.77 |
3.00 ± 0.82 |
4.00 |
0.234 |
Average Glucosuria (cross) |
2.37 ± 0.80 |
2.50 ± 1.12 |
3 |
0.719 |
HbA1c (%) |
12.12 ± 1.81 |
12.89 ± 1.97 |
15.10 |
0.187 |
The average ketonuria was higher in coma patients (4.00) than in obnibulated (3.00 ± 0.82) and conscious patients (2.76 ± 0.77). The difference was not statistically significant (P = 0.234).
The average glucosuria and ketonuria were higher in coma patients (3) than in obnibulated (2.50 ± 1.12) and conscious patients (2.37 ± 0.80). The difference was not statistically significant (P = 0.719).
The average HbA1c was higher in coma patients upon admission (15.10) than in obnibulated (12.89 ± 1.97) and conscious patients (12.12 ± 1.81), with a non-significant difference (P = 0.187) (Table 3).
Table 4. Cross-tabulation between kussmaul’s dyspnea and patient characteristics.
Patient Characteristics |
Presence of Kussmaul’s Dyspnea |
Absence of Kussmaul’s Dyspnea |
P-value |
Average Blood Glucose (g/L) |
3.50 ± 1.19 |
3.41 ± 1.02 |
0.845 |
Average Ketoneuria (cross) |
2.86 ± 0.90 |
2.81 ± 0.77 |
0.879 |
Average glycosuria (cross) |
2.26 ± 1.11 |
2.40 ± 0.80 |
0.89 |
HbA1c (%) |
13.83 ± 1.93 |
12.05 ± 1.75 |
0.016 |
The average blood glucose levels were similar in patients with Kussmaul’s dyspnea at admission (3.50 ± 1.19) and those without (3.41 ± 1.02), with no significant difference (P = 0.845).
The average ketonuria levels were also similar in patients with Kussmaul’s dyspnea at admission (2.86 ± 0.90) and those without (2.81 ± 0.77), with no significant difference (P = 0.879).
However, the average HbA1c was significantly higher (P = 0.016) in patients with Kussmaul’s dyspnea at admission (13.83 ± 1.93) compared to those without (12.05 ± 1.75) (Table 4).
Table 5. Characteristics of the Three Main Types of Diabetes Found (Type 1 Diabetes - Type 1b Diabetes - Type 2 Diabetes).
Patient Characteristics |
Type 1 Diabetes (9 patients) |
Type 1b Diabetes (12 patients) |
Type 2 Diabetes (31 patients) |
P-value |
Age (years) |
24.67 ± 6.46 |
35 ± 6.15 |
43.71 ± 9.20 |
0.000 |
Average Blood Glucose (g/L) |
3.64 ± 0.93 |
3.36 ± 0.99 |
2.81 ± 0.79 |
0.764 |
Average Body Mass Index (kg/m2) |
21.73 ± 3.44 |
26.65 ± 6.63 |
25.96 ± 4.82 |
0.069 |
Normal BMI (24 patients) |
8 (33.3%) |
4 (16.7%) |
12 (50%) |
0.017 |
Infections |
5/9 (55.5%) |
2/12 (16.7%) |
19/31 (61.3%) |
0.03 |
Unspecified Causes |
4/9 (44.4%) |
10/12 (83.3%) |
12/31 (38.7%) |
0.03 |
Total Insulin Dose at Discharge (IU) |
50.44 ± 11.78 |
63.83 ± 14.73 |
49.61 ± 23.82 |
0.124 (NS) |
Insulin Discontinuation (patients) |
0 |
12/12 (100%) |
10/31 (32.2%) |
0.0001 |
The average age of diagnosis for Type 1 diabetic patients (24.67 ± 6.46) was younger compared to patients with ketosis-prone diabetes (35 ± 6.15) and Type 2 diabetic patients (43.71 ± 9.20). The body mass index (BMI) was significantly lower (P = 0.069) in Type 1 diabetic patients (21.73 ± 3.44) than in patients with ketosis-prone diabetes (26.65 ± 6.63) and Type 2 diabetic patients (25.96 ± 4.82). Infections were found in 61.3% of Type 2 diabetic patients (P = 0.03), while unspecified causes of diabetic ketoacidosis were found in 83.3% of patients with ketosis-prone diabetes, with significant differences (P = 0.03). Insulin therapy was discontinued in all patients with ketosis-prone diabetes and in 32.2% of Type 2 diabetic patients, with a significant difference (P = 0.0001) (Table 5).
4. Discussion
We conducted a descriptive, cross-sectional, analytical study over 12 months. An advantage of our study was the infrequency of missing data. It included 54 patients and revealed a hospital prevalence of 17.1%. The short-term progression and prognosis of diabetic ketoacidosis (DKA) vary widely between centers, with delays in ketosis compensation and hospital stay durations being highly variable.
However, our study has some limitations:
Epidemiological Aspects
The prevalence of inaugural DKA ranged from 15% to 30% in adults and up to approximately 30% to 40% in children with type 1 diabetes (T1D) [1] [5] [11]. In Africa, it ranged from 30% to 40%, according to different series [12] [13]. In our study, the prevalence was 17.1%.
It is well known that diabetic ketoacidosis (DKA) most commonly occurs in children and adolescents with type 1 diabetes, but it is increasingly observed in adults as well [8]. The majority of patients in our study were over 40 years old. The average age was 38.54 years, consistent with some studies: slightly higher than Mee Kyoung K’s study in Korea (34.24 years) [14], and close to Thewjitcharoen Y’s study in Thailand, where the average age was 37.7 years [15]. A male predominance was observed in several African and Asian studies, though the reason for this sex difference remains unclear [15]. In our study, males predominated with a sex ratio of 1.16.
According to the literature, smoking increases the risk of developing type 2 diabetes by approximately 44% [16] and induces chronic inflammation [16]. Nicotine also has a direct toxic effect on the pancreas and insulin receptors, contributing to insulin resistance [16]. In our study, 11.1% of patients were smokers. Recent studies have demonstrated that chlorpromazine (CPZ) induces hyperglycemia and glucose intolerance in humans [17]. One of our patients was treated by chlorpromazine.
Diagnostic Aspects
Consciousness disturbances were reported in 14.9% of our patients, with 1.9% in a coma. Most patients had glycosuria and ketonuria at two crosses and were admitted to a pre-coma ketoacidotic state. This aligns with the literature, where only 10% of patients are comatose upon admission [13]. The average capillary blood glucose was 3.43 g/l, lower than the 6.40 g/l reported by Thewjitcharoen Y in Tunisia [15]. We attribute this lower average to the majority of patients being referred from other facilities where intravenous insulin therapy had already begun. Average glycosuria and ketonuria were higher in patients with acidotic syndrome, though the difference was not significant. The average HbA1c was higher in patients who presented with coma or Kussmaul’s breathing, also with no significant difference. This is consistent with the literature, where glycosuria and ketonuria are elevated in the stage of acidotic coma [15]. The average HbA1c in our study was 12.2%, similar to that reported by Taieb A in Tunisia (12.17%) [2] but higher than that reported by El Jadi (10.7%) [18].
In our study, infection was identified as a decompensating factor in 50% of cases. This aligns with other series, including Taieb A. (37.1%) and Chihaoui M. in Tunisia (41.3%) [2] [19]. We found a predominance of urinary tract infections, soft tissue infections, pneumonia, and infected diabetic foot ulcers [20]. These results were consistent with those of Riden D. [20]. The predominance of infectious etiology in our regions may be correlated with several factors: ignorance of diabetes, poverty, the endemicity of certain diseases such as malaria, and delays in consultation. For 50% of the patients, no decompensating factor could be identified. Thewjitcharoen Y. found no triggering factors in 52% of their patients [15].
Nine patients (16.6%) were classified as type 1 diabetics (T1D). This proportion was lower than those reported by H. X. Liu in China and Riden D., which were 30.1% and 33.3%, respectively [20] [21]. DKA is a classic mode of discovery for T1D. The average age of T1D in our series was 24.67 years; T1D generally occurs in younger individuals. The average blood glucose at the time of diabetes discovery was higher in T1D (3.64 ± 0.93 g/l) compared to those with KPD (3.36 ± 0.99 g/l) and those with T2D (2.81 ± 0.73 g/l). This is consistent with the literature, where patients with KPD present with an initial discovery similar to that of T1D [22]. Another form of T1D occurring in older individuals is LADA. One of our patients was diagnosed with T1D at the age of 35.
Twelve patients were classified as type 1b diabetics, also known as ketosis-prone diabetes (KPD). This subtype of diabetes is situated between T1D and T2D [23]. The lack of consensus regarding its classification poses an obstacle to conducting studies on KPD. The WHO and ADA classify it as “idiopathic type 1 diabetes” or “type 1B” [23]. A study conducted over four months revealed a high frequency of KPD (28.3%) in Cameroon, suggesting it may be more widespread than previously described. The average age of onset of KPD in our study (35 ± 6.15) is similar to other studies, typically between 35 and 45 years [24]. It is higher than that of T1D (24.67 ± 6.46) and lower than that of T2D (43.71 ± 9.20), as reported in the literature. A family history of diabetes was found in 75% of patients with KPD. This frequency was higher than that reported by Brunel G. and Chihaoui M., who reported 37.5% and 57.3%, respectively [19] [24]. The difference in frequency between the studies is difficult to explain, as patients come from different socio-economic backgrounds and countries [24]. In our study, men were more frequently affected by KPD, with a sex ratio of 2, consistent with other series (58% to 76%) [24]. The average BMI at KPD discovery was 26.65 ± 6.63 kg/m2, similar to French studies (25 - 26 kg/m2) but differing from American studies (29 - 37 kg/m2). An undifferentiated cause was most common in KPD patients (83.33%), showing a significant difference (p = 0.03), aligning with the literature. In Brazzaville, no decompensating factor was found in 65.4% of KPD patients [22].
Thirty-one patients were classified as type 2 diabetics. T2D typically occurs in patients with a family history of diabetes and may be associated with dyslipidemia. In our study, 61.8% of patients had a family history of diabetes, and 54.8% had dyslipidemia. The higher rate of ketosis in T2D compared to T1D can be explained by the study’s focus on hospitalized patients in an adult endocrinology department. Some teams argue that the diagnosis of T2D during inaugural DKA may constitute a distinct entity from classic type 2 diabetes. This has led to the development of the modified French ADA classification, which has been adopted in some studies, such as Chihaoui’s [19]. In this classification, diabetics discovered in an acidotic state with positive autoantibodies are classified as “type 1,” while those without autoantibodies are classified as “ketosis-prone type 2 diabetes” (equivalent to KPD). However, T2D is a latent disease that may be asymptomatic for years, especially in contexts where early screening is less common than in Western countries. Moreover, T2D can present with ketosis in the presence of a triggering factor, even though ketosis is not a classic initial presentation of the disease. It will, therefore present as an acute initial case with severe hyperglycemia and ketosis, indicating acute insulin deficiency, similar to T1D. Furthermore, infection was found more frequently in T2D (61.3%) with a significant difference (p = 0.03). T2D can present with ketosis in the presence of a precipitating factor. Hence, we chose to classify our patients according to the WHO classification. Other studies have also preferred to adopt the ABeta classification, as in the study by Taieb [2]. Thus, the search for pancreatic autoantibodies and the assessment of beta cell secretion capacity through C-peptide measurement may be helpful. Unfortunately, these biological parameters are not routinely available in our country, so we could not adopt this classification.
Two patients developed diabetes due to corticosteroid therapy. Among patients hospitalized in internal medicine, approximately 11% receive high doses of corticosteroids (>40 mg prednisone/day for at least 48 hours), with 64% experiencing at least one episode of hyperglycemia and nearly half developing diabetes [18]. One patient had lupus and the other had vasculitis. They were on 40 mg prednisone for 5 months and 60 mg prednisone for 4 months, respectively, without dose reduction as they were lost to follow-up.
The HbA1c level at diagnosis was not discriminative in our study and was nearly equal across the three patient groups. This is consistent with the results of Brunel [24].
Therapeutic and Evolutionary Aspects
Intravenous insulin therapy and rehydration were implemented for all our patients. The average length of hospitalization was 7.6 days. Riden D. [20] reported an average hospitalization duration of 8.5 days. The main complication observed was hypokalemia, affecting 13% of our patients. DKA is associated with a potassium deficit, which can initially be masked by hyperkalemia due to acidosis, proteolysis, and insulin deficiency [25]. The correction of these disorders through hydration and insulin therapy can reveal this deficit. Ach Taieb and his team reported fewer cases of hypokalemia (2.8% compared to 13% in our study) [2]. One patient experienced worsening consciousness disturbances, resulting in death. The mortality rate reported in the literature is less than 1%.
At the time of DKA discovery, patients with KPD typically require insulin therapy [26]. Subsequently, hypoglycemic episodes are observed in most patients. After the acute episode, the clinical course resembles classic T2D and may not require insulin. Most of these patients remain in near-normoglycemic remission without insulin for several months or years. The remission phase is defined by insulin withdrawal with an HbA1c level below 6.5% for more than 3 months. In the literature, this remission typically begins between 9 and 14 weeks after the discovery of diabetes and can last for several years [26]. In our study, patients experienced hypoglycemic episodes that required insulin discontinuation within the first three months, with HbA1c remaining below 6.5% for more than 3 months.
5. Conclusion
In conclusion, inaugural diabetic ketoacidosis is common among overweight adults with Type 2 diabetes and often occurs in Ketosis-Prone Diabetes (Type 1b). Proper typing is crucial for guiding treatment and follow-up. Distinguishing Type 1 from Ketosis-Prone and Type 2 diabetes requires antibody testing, which is unavailable to all patients. However, rapid remission within 14 weeks post-hospitalization helps diagnose ketosis-prone diabetes retrospectively. Under specialized care, these patients generally show favorable outcomes. Since undiagnosed diabetes increases the risk of DKA and diabetes prevalence is high in our area, we recommend more mass diabetes screenings.