Confidentiality was assured by giving each patient a unique number. The study was started after getting ethical approval from the appropriate body of the medical faculty of the university. Patients excluded from the study were: patients age < 13 years, patients who were not admitted, who had wrong diagnosis, or there were insufficient data for diagnosis.

Clinical staging of HIV/AIDS was made based on World Health Organization (WHO) staging system for HIV infection and disease [14]. The most recent CD4 count was taken into consideration. Poor outcome at hospital discharge was defined as patients who remained in the same clinical condition or deteriorated without death. Improved patients were patients who showed partial or complete recovery. The presumptive diagnosis of the neurological manifestations of HIV/AIDS was made as follows:

Twenty-five percent (25%) came from outside Addis Ababa and the mean duration of symptoms was 30.6 days (range 1-240). The mean duration of symptoms for the major neurologic disorders identified was cerebral toxoplasmosis (25 days, median 14 days), tuberculous meningitis (34 days), cryptococcal meningitis (26 days) and bacterial meningitis (10 days). The diagnosis of HIV was made in 117 patients before current hospital admission and there was no statistically significant difference in case-fatality between patients who were diagnosed before and after admission (p = 0.315). The mean duration between diagnosis of HIV and current hospital admission was known for108 patients, and ranged between 1 and 120 months (mean 14.8 months). History of treatment with antiretroviral therapy (ART) was identified in 54(15.6%) patients with mean duration of treatment of 7.2 months (median 5 months, range1-29). ART was discontinued before hospital admission in 5 cases (They went to holy water). Co-trimoxazole prophylaxis was given to 15 cases before current admission and 2(13.3%) of these developed cerebral toxoplasmosis (p-value = 0.230).

The mean hospital stay was 21.9 days (range1-120). The mean hospital stay in days for improved and deceased patients was 26.4 (range 7-120) and 16.9 (range 1-91), respectively. The clinical manifestation of patients with neurological disorders is shown in Table ​Table11 and the neurologic manifestation of HIV/AIDS in relation to outcome status is shown in Table ​Table2.2. Herpes Zoster scar was identified in 67(19.3%) of cases. Co-morbidities identified at admission were pulmonary tuberculosis 18(5.2%) and pneumonia 18(5.2%). Some patients (n = 7, 2%) had more than one neurologic complications of HIV/AIDS: cryptococcal meningitis and neurosyphilis (n = 3), cryptococcal meningitis and GBS (n = 1), neurosyphilis and GBS (n = 1), tuberculous meningitis and tuberculoma (n = 1), cerebral toxoplasmosis and neurosyphilis (n = 1). The Glasgow coma scale (GCS) was recorded in 85(24.5%) of cases and 35(41.2%) had score of ≤ 8. The Case-fatality of patients with GCS score of ≤ 8 and > 8 was not different (p-value = 0.936). The WHO clinical staging of HIV/AIDS patients showed that majority (n = 332, 95.7%) were stage IV and the remaining were stage III (n = 15, 4.3%) cases.

The occurrence of seizure in cerebral toxoplasmosis in other studies [2,3] was 25% and 29%, respectively, which was lower compared to our finding. In another study [12], seizure occurred in 40% of patients with cerebral toxoplasmosis which was similar to our study. In a study [12] the occurrence of seizure in tuberculous meningitis (28%) and cryptococcal meningitis (50%) was lower and higher, respectively, compared with findings in our study. The common neurologic disorders identified in our cases were: cerebral toxoplasmosis (36.6%), tuberculous meningitis (22.5%) and cryptococcal meningitis (22.2%), and the frequencies were similar to other studies [5,6,11,12,15]. In contrary to our findings, tuberculous meningitis and bacterial meningitis were rare findings in patients with neurologic manifestations of HIV/AIDS in the United States [2]. In a study from India [12], cerebral toxoplasmosis was detected in 8.8%(5/57) which is very low compared to our finding. The number of patients with presumed diagnosis of primary CNS lymphoma was very low in our cases probably due to failure to do brain biopsy when empiric therapy for toxoplasmosis failed.

Stoke was identified in 2.3% of our patients which is low compared to a finding in another study [2] which was 12%. The low number of patients (2%) with multiple neurologic disorders compared to 13.5% in another study [2] probably shows the limited investigations available in our set up. Unlike other studies [2,16,17] HIV-encephalopathy was diagnosed in only one of our patients (0.28%).

Majority of our patients (89.7%) had a CD4 count below 200/mm3 which is consistent with another study [12]. Note that adult healthy Ethiopians have a mean absolute CD4 T-cell count of 775 cell/μL(range 366-1,235) in one study [18]. Neuroimaging studies in our cases showed focal lesions in 56.8% which is higher when compared to 25.5% obtained in another study [2]. The commonest (80%) cause of mass lesion on imaging studies in our patients was cerebral toxoplasmosis which is similar to other studies [1-3,5-7] but contrary to a study done in South Africa [9] where tuberculosis predominates as a cause of focal brain lesion(69%). In our study, normal neuroimaging finding and cerebral atrophy were detected in22.7% and 3.9%, respectively which is lower compared to another study [2]. These might be explained by the limited access to neuroimaging studies and the low incidence of HIV-encephalopathy, which is the main cause of brain atrophy in HIV/AIDS patients, in our cases. A study [19] on HIV-related neuropathology showed that the incidence of HIV-encephalopathy rises as patients survive longer after initiation of antiretroviral therapy and is the most common neurologic complications of HIV/AIDS in such set up [2]. In a study from Nigeria [10] which included ART naive patients only, HIV-encephalopathy was found in 12%(19/154) of cases which is higher when compared to our study. These show that history of treatment with ART alone does not explain the low incidence of HIV-encephalopathy in our patients. Our patients were highly selected group of patients with advanced disease stage in whom the neurologic symptoms might be overshadowed by overwhelming systemic illness. In addition, lack of adequate screening techniques and readily available imaging studies might contribute.

Amphotericin B was not readily available and treatment of cryptococcal meningitis was either Amphotericin B (n = 32) or oral fluconazole only (n = 45) [There was no significant difference in mortality, p-value = 0.260]. The limited availability of standard therapy might explain partly the higher case-fatality of cryptococcal meningitis (48.1%) in our cases compared to 32.4% in another study [2]. Sulfadiazine-pyrimethamine was not available and patients with cerebral toxoplasmosis were treated with pyrimethamin-sulfadoxin. This might partly explain the higher mortality in our set up (44.1%) compared to 30.2% in patients who were treated with standard treatment [2].

The overall hospital mortality was 45% which is high compared to other studies. In a study done in Nigeria [10] which included outpatients and inpatients, an overall 6-month mortality of 45% was found. In another study from India [11], a mortality of 34% was recorded. The in-hospital mortality of patients with neurologic manifestations of AIDS in Brazil [7] was 32%. The case-fatality for the major neurologic disorders in our study was: tuberculous meningitis (53.8%), cryptococcal meningitis (48.1%), cerebral toxoplasmosiss (44.1%) and bacterial meningitis (33.3%). In a study done in Cambodia [15], the case-fatality rates of tuberculous meningitis and cryptococcal meningitis was similar to our study which were 57% and 49%, respectively. In studies done in South Africa [20] and Nigeria [10] the case-fatality of HIV associated tuberculous meningitis was 71.8% and 77.5%, respectively, which is higher compared to our finding.

Both univariate and multivariate logistic regression analysis showed that address from Addis Ababa, altered sensorium and seizure were predictors of mortality. The unexpected relatively lower mortality of patients from outside the city of Addis Ababa might be partly explained by selection bias of cases who were relatively stable and able to travel to the city for better health care services. History of ART at admission was not found to improve mortality significantly.

This unexpected finding might be due to the relatively short period of treatment (mean 7 months, median 5 months), noncompliance (5 patients discontinued ART and of these 3 died) and only 15.6% of our cases were on ART. In patients with history of ART the common neurologic complications identified were: cerebral toxoplasmosis (n = 13), cyptococcal meningitis (n = 12), tuberculous meningitis (n = 10), bacterial meningitis (n = 9), stroke (n = 4) and others (n = 10). Further study is recommended to identify other possible reasons for the failure of ART in our patients.

There were several limitations to this study. Patients who were treated as outpatient only were not included and therefore the whole spectrum of neurologic manifestation might not be addressed. The absence of certain neurologic disorders in our study (myelopathy, peripheral neuropathy other than inflammatory demyelinating polyneurpathy, myopathy, etc.) might be partly explained by inclusion of inpatients only. Our hospital is a tertiary hospital and the chance of admitting selectively critically sick patients is high due to referral from other hospitals. The retrospective nature of the study might have caused unrecoverable missing data in some cases. There was no autopsy, biopsy, viral load and polymerase chain reaction done. CSF glucose and protein determination was done in minority of patients.

To improve care of patients with neurologic manifestations of HIV/AIDS, early diagnosis of HIV infection, patient education, starting prophylaxis when needed (e.g. INH and cotrimoxazole prophylaxis, BCG vaccination to decrease incidence of tuberculous meningitis), starting ART timely, making standard therapy for opportunistic infections readily available(e.g. sulfadiazine-pyremethamin with leucovorin and Amphotericin B), improving availability and quality of laboratory investigations and neuroimaging studies are very important.