Acute promyelocytic leukaemia: A central South African experience

mortality rate was 18.5 % , and the 30-day mortality rate was 33.3 % . Sanz and modified Sanz scores were significantly associated with 7-day mortality but not 30-day mortality. Creatinine ≥ 105 µmol/L was significantly associated with both 7-and 30-day mortalities. Patients who died within the first 30 days of admission had significantly higher median white cell counts and partial thromboplastin times. Hypogranular APL was identified in 55.6 % of patients. Conclusion: The short-term mortality of APL at UAH is in keeping with findings at other treatment centres in middle-income countries. Despite being considered rare, hypogranular APL was the predominant type in this cohort. Contribution: This study highlights the need for practices pertaining to peripheral smear utility and interpretation to be reviewed outside of tertiary centres.


Introduction
of granulocyte precursors (promyelocytes) in the peripheral blood and bone marrow. 3 These abnormal promyelocytes can evoke life-threatening haemorrhage resulting from disseminated intravascular coagulation (DIC) and hyperfibrinolysis. 8,9,10,11 Bleeding accounts for 32% -64% of early deaths and is the major cause of treatment failure during induction. 9,12,13,14,15,16 Coagulopathy, necessitating prompt treatment and intensive management with blood products and referral to experienced treatment centres, has become the hallmark of APL. 17,18 The increased availability of targeted therapies such as alltrans retinoic acid (ATRA) and arsenic trioxide (ATO), combined with anthracycline chemotherapy, has improved APL outcomes. Complete remission rates of 90% -95%, long-term survival exceeding 75% and early death (defined as death during induction therapy) below 10% have been reported in large multi-centre trials. 9,12,13,19 Population-based studies have contradictory findings, with early death rates ranging between 17.3% and 29% in upper-middle to highincome countries. These studies defined early deaths as death occurring either within 14 or 30 days of diagnosis, respectively. 14,15,17,18,20 Similar findings in several institutional studies in low-and middle-income countries further highlight discrepancies and the need for ongoing populationbased studies. 21,22,23 The primary objective of the study was to describe the demographic variables, clinical characteristics and laboratory features that might have been associated with short-term (7-day and 30-day) mortality of patients with APL treated at the Division of Clinical Haematology at Universitas Academic Hospital (UAH), Bloemfontein, South Africa, during the period 2015-2019. The secondary objectives were to describe the treatment provided and the supportive care required during the first 30 days.

Study design
A retrospective, quantitative, cross-sectional study was performed.

Setting
The Division of Clinical Haematology is a sub-speciality public facility that treats patients from the Free State, Northern Cape and parts of the Eastern Cape provinces in South Africa, and Lesotho, a neighbouring country.

Patients
Patients were identified from the UAH institutional APL database and laboratory databases and from the electronic pharmacy records for patients who received ATRA at UAH. Consecutive files of all patients older than 13 years of age with APL, who had the first presentation during the study period (01 January 2015 -31 December 2019), were included. The diagnosis of APL was based on morphologic features and molecular tests, including fluorescence in situ hybridisation (FISH) and polymerase chain reaction (PCR). Patients were excluded if they presented for follow-up, had relapsed disease, had variant translocations (other than t (15;17)), or if clinical and laboratory data could not be traced.

Data sources
The data collected were demographic variables, clinical and laboratory findings during the first 30 days of admission, and pharmacy and transfusion records. Demographic and clinical data were obtained from the patients' files and the MEDITECH electronic patient file system. Laboratory data were retrieved from TrakCare, the National Health Laboratory Service (NHLS) electronic database. No personal identifiers were captured.
Study data were collected and managed using REDCap (Research Electronic Data Capture), 24,25 hosted by the University of the Free State (UFS).

Quantitative variables
Full blood count (FBC), coagulation data and CD4 counts were collected as continuous variables. Age, various dates, creatinine clearance, baseline left ventricular ejection fraction (LVEF), number of units transfused (red cell concentrate, platelets, cryoprecipitate and fresh frozen plasma independently) and dose of dexamethasone were collected as discrete variables.
Sanz, 26 modified Sanz 27 and DIC scores 28,29,30 were calculated based on the results of collected laboratory variables. A positive FISH result was defined by the presence of the translocation t (15;17). A positive quantitative polymerase chain reaction (qPCR) was defined by the presence and quantity of t(15;17), PML-RARA fusion transcript copies. The 7-and 30-day mortalities were based on the first presentation to UAH.

Categorical variables
Patients were classified according to the peripheral smear morphological characteristics described by the WHO as having either hypogranular or hypergranular APL. 3 The granular type was assigned by the haematopathologist at the time of diagnosis and confirmed by a second haematopathologist. Gender, nationality, clinical features at presentation (bleeding sites, infection sites, comorbidities, HIV status and HIV viral load) and outcomes (neutropenic sepsis, differentiation syndrome, ICU admission, outcome and cause of death) were collected as categorical variables.

Statistical analysis
Data analysis was performed using Statistical Analysis Software (SAS), Version 9.4 (SAS Institute Inc.; Cary, NC, United States [US]). Numerical variables were expressed as medians (interquartile ranges [IQR]) and compared using the Kruskal-Wallis test. Categorical data were expressed as percentages and compared using the chi-square or Fisher's exact test. The level of statistical significance was set at p < 0.05.

Ethical considerations
Approval to conduct this study was obtained from the UFS Health Sciences Research Ethics Committee (HSREC) (ethics approval number UFS-HSD2020/0448/3006) and the Free State Province Department of Health (approval number FS_202005_034). Signed informed consent was not required because of the retrospective nature of the study.

Number of patient files
Fifty-three files were screened for eligibility, of which 27 were included in the study. Figure 1 summarises the patient files screened for eligibility and reasons for exclusions.

Demographic variables
The majority (n = 22; 81.5%) of the patients were South African citizens from the Free State (n = 13; 59.1%) and Northern Cape (n = 8; 36.4%) provinces. One South African patient (4.5%) was referred from the Eastern Cape, while the remainder were referred from Lesotho. A slight male predominance of 55.5% (male-to-female ratio of 1:0.8) was observed. The median age of the patients was 31.8 years (range 15-65 years).

Clinical features
Comorbidities were documented in over a third (n = 10; 37.0%) of the study population. Hypertension was most frequently reported, followed by human immunodeficiency virus (HIV) infection (n = 6 and n = 3, respectively). Other comorbidities (all n = 1) included previous pulmonary tuberculosis, diabetes mellitus, asthma, mitral valve prolapse and renal failure. Three patients had more than one comorbidity.
All patients were symptomatic on presentation to UAH, and slightly more than three-quarters (n = 21; 77.8%) had more than one symptom at presentation. The most frequently reported clinical feature was symptomatic anaemia (n = 20; 74.1%) followed by bleeding (n = 18; 66.7%) and infection (n = 9; 33.3%). Both bleeding and thrombotic complications were observed at presentation. Table 1 summarises the clinical features of the patients at presentation to UAH.

Laboratory, molecular and genetic characteristics of patients Full blood count
All patients had an FBC performed on admission. Table 2 summarises the baseline FBC findings.

Morphological findings
Hypogranular APL was the predominant morphological type among patients (n = 15; 55.6%), hypergranular APL occurred in 11 (40.7%) patients and the morphological type of one patient was unknown.

Coagulation studies
Coagulopathy screening was performed on all patients on admission.

Modified Sanz score
When the modified Sanz 27 score was applied, approximately half of the patients (n = 14; 41.9%) were classified as high-risk patients and a quarter of the patients were classified as ultra-high-risk patients (n = 7; 25.9%). Intermediate-risk (n = 5; 18.5%) and low-risk (n = 1; 3.7%) patients accounted for less than a quarter of the total study population.

Genetic features
A diagnosis of APL was confirmed by FISH for t(15;17) in 26 (96.3%) cases and by PML-RARA reverse transcriptionpolymerase chain reaction (RT-PCR) in a single case. The median percentage of positive t(15;17) cells was 65% (IQR: 57% -75%). A baseline qPCR for PML-RARA was requested for six (22.2%) patients, with a result being obtained in five (18.5%) cases. In one patient, the total ribonucleic acid (RNA) yield was insufficient to produce a result. Among those with a qPCR result, the median number of PML-RARA copies was 3277 (IQR: 1229-4273 copies).

Management of patients with acute promyelocytic leukaemia The first suspicion of acute promyelocytic leukaemia and time to referral to Universitas Academic Hospital
Approximately three-quarters (n = 20; 74.1%) of the patients were referred to UAH with a clinical and laboratory suspicion of APL, while the remaining patients (n = 7; 25.9%) were referred with alternate diagnoses. All patients from Lesotho were referred with a presumed diagnosis of APL based on peripheral smear results.
The time and date of the first suspicion of APL, based on morphologically suggestive smear results, were known for all 22 South African patients, but unknown for the patients from Lesotho. In 15 (68.2%) South African patients, the morphological diagnosis was made prior to referral to UAH.  Figure 2). A delay of 46 h 12 min was reported for one patient, with delayed transportation documented as the cause. Figure 2 illustrates the various time intervals from the first suspicion of APL to referral to UAH.

Index presentation: Referral complications
In the context of this article, 'index presentation' refers to the date and time of the first FBC performed on a patient at the referring healthcare facility. In three patients, the index presentation was unknown, and for these cases, the documented date and time of the first presentation to a healthcare facility was used.
The median time (days) between index presentation and transfer to UAH was three days (IQR: 1-6 days). However, when a peripheral blood smear result accompanied the FBC at index presentation and APL was reported on the smear, a significantly shorter median time between presentation and referral to UAH was observed (p = 0.01). Table 3    had pancytopenia, one patient had leucocytosis (white cell count 57.59 × 10 9 /L) and one had both anaemia (haemoglobin 4.6 g/dL) and thrombocytopenia (platelet count 11 × 10 9 /L). All seven patients subsequently had peripheral blood smear results suggestive of APL.

Management of patients at Universitas Academic Hospital
One patient's treatment file could not be traced, and another patient died prior to the commencement of treatment. The results pertaining to treatment in particular are reported for the remaining 25 patients. All other analyses included these two patients' information. In addition to ATRA, all patients received chemotherapy with daunorubicin, an anthracycline. The median time (days) that chemotherapy commenced after the first presentation to UAH was one day (IQR: 1-2 days). The LVEF was quantified at baseline in most patients (n = 22; 88.0%). One patient had a low LVEF of 42% at baseline, requiring a daunorubicin dose reduction.
Corticosteroids (dexamethasone) were prescribed to most patients (n = 20; 80.0%). The indications for dexamethasone were prophylaxis to prevent and treat differentiation syndrome (n = 20; 80.0% and n = 4; 20.0%, respectively). All patients with a white cell count above 10 × 10 9 /L received dexamethasone according to treatment guidelines. 31,32,33 Five (20.0%) patients required admission to the intensive care unit (ICU) within the first 30 days of hospitalisation. All patients required blood products. Table 4 describes the supportive treatment of patients in terms of blood products received during the first 30 days of admission.
Patients with a Sanz score in the high-risk category required more supportive care in terms of blood products (309 vs. 194 units) and ICU admissions (4 vs. 1). In contrast, the median number of red cell transfusions per patient was significantly higher (P = 0.01) in the intermediate-/low-risk Sanz score group than that in the high-risk group. There were no statistically significant differences between the highrisk and intermediate-/low-risk groups regarding the number of units of platelets, fresh frozen plasma and cryoprecipitate transfused (Table 5).
No statistically significant difference was observed between the groups' ICU admission rates and transfusion requirements using the modified Sanz score.

Short-term mortality
Two-thirds (n = 18; 66.7%) of patients were alive at 30 days, of which seven (38.9%) had a hospital stay of more than 30 days. The median admission time at UAH for patients that survived the first 30 days was 27 days (IQR: 24-51 days). A 100% (n = 5) mortality rate was observed among patients admitted to the ICU. The causes of death included intracranial haemorrhage (n = 2), ischemic stroke (n = 1), ischemic stroke with haemorrhagic transformation (n = 1), sepsis (n = 4), differentiation syndrome with sepsis (n = 1) and an unidentified cause (n = 1).
More than half (n = 5/9; 55.6%) of deaths occurred within the first seven days of admission to UAH. Ischemic and haemorrhagic complications were the cause of death in patients who died in the first seven days of admission (n = 4/9; 44.4% and one unknown). This finding was contrary to those who died after the first seven days, where sepsis and differentiation syndrome were documented as the cause of death (n = 4; 44.4%). Tables 6a and 6b summarise the association between various parameters and the 7-and 30day mortality rates.

Discussion
Early mortality in APL remains a considerable challenge at UAH. As a referral and specialised haematology treatment centre, we would have expected our APL outcomes to be in line with other academic centres. 15,16 However, our mortality even exceeds the mortality of the Canadian Cancer Registry, which included all patients with APL managed at various centres and not only those managed at specialist referral centres. 15 This may be because of a delay in timely diagnosis and referral, although contrary to what has been shown in other South African centres. 16 A third of the patients in the cohort died within the first 30 days of presentation to UAH, with more than half of the deaths occurring within the first week. This observation was in keeping with the findings of institutional studies in other lower-middle and upper-middle-income countries, 21,22,23 although contrary to the results of large-scale clinical trials and population-based studies in high-income countries. 9,13,14,17,18,19,34 Nevertheless, a recent study from another centre in South Africa demonstrated a low early mortality rate of 13%, 16 which is comparable to findings at centres in Europe 17 and North America. 19 A slight male predominance was observed in our study, contrary to findings reported in several other institutional studies. 16,17,21,22,23 Most patients received ATRA soon after morphological diagnosis. However, the under-utility of the initial peripheral blood smear at referring centres delayed the diagnosis and transfer to the treatment centre (P = 0.01). Equally concerning was that almost a fifth of patients did not have APL reported on the blood smear performed at index presentation, which might be a consequence of inexperienced technologists and the absence of an on-site pathologist at peripheral centres. Conversely, despite all patients from Lesotho having a morphological diagnosis of APL on index presentation, delays in transfer occurred in most cases. Suspicious peripheral smears identified by laboratory professionals in Lesotho are transferred across an international border to a private laboratory in South Africa, but these delays could not be measured. In most cases, initiation of treatment was not dependent on confirmatory diagnosis by FISH or PCR.
Baseline PCR for t(15;17) was not requested for most patients before induction therapy. The utility of qualitative PCR t(15;17) has become increasingly important, with several studies highlighting the prognostic implications between the variant PML-RARA breakpoints. 21,34,35 Additionally, quantitative PCR t(15;17) at baseline remains important in later assessments for minimal residual disease. 33 A unique feature among the study population was the observation of hypogranular APL being the predominant type, which, to the researchers' knowledge, is the first time this has been reported in an institutional cohort. 16,17,18,19,28,30,31 Despite this finding, morphological variance alone was not significantly associated with either the 7-day or the 30-day mortality.
Bleeding complications were a prominent finding among patients. All patients met laboratory criteria for DIC, threequarters of whom had supportive clinical features of DIC at presentation. Death because of bleeding and thrombotic complications occurred within the first seven days of presentation, while infection and differentiation syndrome accounted for deaths thereafter. This might be explained in part by a delay in diagnosis and transfer of patients, which was observed in two cases (n = 2/5; 40.0%) who died within the first week of presentation. Importantly, patients were classified as high risk and ultra-high risk by the Sanz and The marked occurrence of coagulopathy at presentation emphasises the need for adequate transfusion support during induction therapy. This is challenging in a resource-limited setting, both financially and regarding the availability of blood products. Despite similar numbers, patients in the Sanz highrisk group received significantly (p = 0.01) fewer red cell transfusions than the non-high-risk group -likely a consequence of the longer duration of hospitalisation in the latter group, none of whom died within the first week of presentation.
While the availability of ATO has increased globally, access to ATO as first-line therapy remains a significant challenge at UAH. This resulted in all patients being initiated on an anthracycline-based chemotherapy regimen, which was not the current international standard of care in the 12 non-high-risk patients. 33 Although most patients were classified as high risk and ultra-high risk upon the first presentation, neither the Sanz nor the modified Sanz scores were significant predictors of

Risk stratification
Sanz score -  30-day mortality. This was contrasting to the 7-day mortality in which both risk groups were associated with death in the first seven days of presentation (p = 0.04 and p = 0.03, respectively). Additionally, a white cell count of > 10 × 10 9 /L was associated with 7-day mortality (p = 0.03). Patients who died within the first seven days of admission were found to have a significantly higher median white cell (p = 0.01) and neutrophil count (p = 0.01). Similar to previous studies, patients who died within the first 30 days of admission were found to have significantly higher median white cell counts (p = 0.03) 22,27,34 and partial thromboplastin times (p = 0.02). 34 While the clinical haematology unit at UAH uses the Sanz classification for risk stratification, if these patients were to be risk stratified according to the updated recommendations from the European LeukemiaNet (ELN), they would fall into the high-risk patient category (WCC > 10 × 10 9 /L). 33 Considering the high mortality rate in patients with WCC > 10 × 10 9 /L and the recommendation to use ATO upfront in high-risk patients according to the ELN guidelines 33 , 7-day mortality may be improved by the addition of ATO in this risk group. This finding may be useful to motivate the availability of ATO at UAH.
A creatinine level of ≥ 105 µmol/L was associated with both 7-day and 30-day mortalities (p = 0.003 and p = 0.03, respectively). Conflicting findings regarding elevated creatinine and early death have been reported. During the development of the modified Sanz score, Lou et al. 27 found no association between creatinine levels of ≥ 105 µmol/L and death before or during the first 30 days of induction among a cohort of 426 newly diagnosed APL patients. More recently, a single-centre study comprising 91 patients between 2 and 21 years of age demonstrated an association between creatinine levels of ≥ 0.7 mg/dL (61.9 µmol/L) and death within 30 days of diagnosis of APL. 36 These findings were similar to De la Serna et al., 37 where a creatinine level of ≥ 1.4 mg/dL (123.8 µmol/L) was associated with death during induction. The proposed mechanism behind the elevated creatinine levels in both studies was microthrombi of the renal vasculature. Among the three patients with a creatinine level of ≥ 105 µmol/L in our study, the possibility of this proposed aetiology was supported by leukocytosis (white cell count range 41-92 × 10 9 /L), overt DIC at presentation and thrombosis at other organ sites. The Cairo-Bishop definition for tumour lysis syndrome 38 was not met in any of the patients.

Study limitations
Although this study represents all adult patients treated at UAH, the small sample size may limit generalisability of findings. The availability of ATRA outside the tertiary setting is challenging in an upper-middle-income country, and data regarding pre-referral patient management (including transfusion of blood products and ATRA administration) were unknown. The consequence of this situation was of particular concern for patients who died within the first week of presentation.
This study aimed to provide insight into the short-term outcome of APL without considering long-term mortality at UAH, which is still unknown. A selection bias is acknowledged, as only patients who were referred and alive at the time of arrival were included in the study. This bias is important, considering high-risk patients and the 7-day mortality.

Recommendations
The most crucial obstacle to treatment initiation currently remains early recognition and referral to a specialist centre. Cost-saving measures at peripheral healthcare facilities often limit extensive investigations. However, given the high mortality of APL at UAH, combined with the significant impact on resources such as blood products, intensive care facilities and the prolonged hospital stay that patients experience, we recommend that every patient with an abnormal FBC, especially with cytopenia or leukocytosis, have an accompanying peripheral blood smear performed.
The lack of pathologists outside the tertiary setting in our country necessitates education and outreach programmes to peripheral centres that may contribute to the prompt diagnosis of APL. It is pivotal to avoid treatment delays while awaiting the transfer of high-risk patients. Blood products and ATRA need to be available and accessible at both regional and district levels. At a tertiary level, quantitative PCR remains underutilised, and it is recommended at baseline and for treatment monitoring purposes.
Potential future studies to determine the long-term survival of patients with APL and a larger study cohort may add to the pool of knowledge on this rare condition.

Conclusion
Globally, this is the first study that shows a predominance of the hypogranular type of APL, although it was not associated with short-term mortality. Despite the high 30-day mortality, the Sanz and modified Sanz scores were not predictors of outcome at 30 days in this study but were instead associated with 7-day mortality. A creatinine level of ≥ 105 µmol/L was found to predict both 7-day and 30-day mortalities, while a WCC of > 10 × 10 9 /L was a predictor of 7-day mortality. Patients who died within the first seven days had a significantly higher median neutrophil and WCC, whereas those who died within the first 30 days had a significantly higher median WCC and partial thromboplastin time. The short-term mortality of APL at UAH was more in keeping with findings at other treatment centres in upper-middleincome countries, with factors such as transport inadequacies, availability of treatment and limited staff and resources being additional impediments in the management of APL.