Classification of HIV Infection and Disease in Women from Rwanda: Evaluation of the World Health Organization HIV Staging System and Recommended Modifications

Methods

Study Cohort

Details of the sampling design and participant recruitment have been published [17-21]. Briefly, during 1986 and 1987, a consecutive sample of 3702 women aged 18 to 35 years was recruited from pediatric and perinatal care outpatient clinics in Kigali, Rwanda, and tested for HIV antibody [17, 18]. From this group, a random sample stratified by HIV status was selected for prospective cohort studies; this sample included 460 HIV-positive and 998 HIV-negative women [19-21]. As in previous studies [19], the follow-up rate after 4 years was approximately 90%.

Evaluation

Clinical evaluation of these women included a semi-annual medical history and weight measurement and an annual physical examination. Body mass index was evaluated using the Quetelet index (weight in kilograms/height in meters²). Testing for HIV was initially done using an enzyme-linked immunosorbent assay (Wellcome Diagnostics, Research Triangle Park, North Carolina), with confirmation by immunofluorescence (Virion, Lucerne, Switzerland) or Western blot (DuPont, Wilmington, Delaware). Other laboratory testing included an annual complete blood count with differential and a Westergren erythrocyte sedimentation rate.

Outpatient medical care was provided at the project site between scheduled visits. Inpatient care was provided at the Kigali Central Hospital, the city's only community hospital, which is located 500 meters from the project site. Information from outpatient and hospital records was incorporated into the clinical data base. For women who died and those for whom these data were not available, family members were interviewed about symptoms before death.

Statistical Analysis

To evaluate the prognostic significance of a staging system, women were classified at baseline into mutually exclusive stages. We evaluated time to death using the Kaplan-Meier survival analysis. Survival curves were compared using the Wilcoxon test of equality over strata; this test is more sensitive than the log-rank test to losses that occur early in the survival distribution [22]. We defined the second follow-up visit in this cohort study (which occurred 12 months after enrollment) as the “baseline date”; this allowed us to include in our “baseline” data all medical conditions that had been diagnosed within the preceding year. Thirty-six months of follow-up were available after this new baseline date.

We did four types of analyses. First, we evaluated survival using the current WHO HIV staging system. Second, we did a series of exploratory analyses to evaluate the prognostic significance of various clinical characteristics and laboratory measurements; these were mostly chosen on the basis of previous analyses of morbidity and mortality in this cohort [19]. We also evaluated prognosis associated with chronic or acute oral or genital ulcers, even if the cause was unknown. Third, we evaluated the subset of women who died during the follow-up period and described the time from first report of certain clinical conditions until the date of death. Fourth, we proposed some revisions to the WHO staging system and evaluated the effect of these modifications.

For evaluation of hematocrit, women were categorized as having values above or below the lower quartile (the lower 25% of results); for erythrocyte sedimentation rate, women were categorized as having values above or below the upper quartile (the upper 25% of results). We chose three categories for body mass index because the WHO staging system has three “weight loss” categories; cut-points were selected using whole numbers that would reflect the distribution of body mass index results and allow sufficient numbers in each category for analysis.

Operational Modifications in the WHO Staging System

For purposes of analysis and because of the methods of data collection, we made several operational adaptations to the WHO clinical staging system. We did not include a formal performance scale, as suggested in the revised WHO staging system [8]. We also did not specifically evaluate women in this cohort for HIV-related encephalopathy. The remaining key features of the WHO system are summarized in the Appendix. The wasting syndrome (for clinical stage IV) was defined as documented weight loss (during any 6- to 12-month interval) of greater than 10% of body weight plus either chronic diarrhea or chronic fever and asthenia. For clinical stage III, pneumonia, meningitis, and pyomyositis were considered severe bacterial infections.

Clinical stage II was adapted in four ways: 1) The WHO proposal includes persons with unintentional weight loss less than 10% of body weight. Because this could theoretically include persons with weight loss of 1% to 2% of body weight [within the range of normal weight fluctuations or variability in measurement], we included in stage II persons with weight loss of 5% to 10% of body weight. 2) Minor mucocutaneous manifestations included oral or genital ulcers of any duration and dermatitis [including pruritic dermatitis]. 3) Because of the way questions were asked, we included women who had had herpes zoster within the preceding 6 years [rather than within the preceding 5 years as suggested by WHO]. 4) Recurrent upper respiratory tract infections were defined as recurrent sinusitis noted on two consecutive medical histories that were obtained 6 months apart; we did not include minor illnesses such as colds.

In the WHO laboratory axis, stage A includes a CD4+ lymphocyte count greater than 500 × 10⁶/L^or a total lymphocyte count greater than 2000 × 10⁶/L. Stage B includes a CD4+ count of 200 to 500 × 10⁶/L or a lymphocyte count of 1000 to 2000 × 10⁶/L. Stage C includes a CD4+ count less than 200 × 10⁶/L^or a lymphocyte count less than 1000 × 10⁶/L[7]. Because CD4+ counts were not available, we based the laboratory staging solely on the total lymphocyte count.

Results

Study Population

We included in our analysis 412 women who were HIV seropositive at baseline; the median age was 28 years (range, 19 to 38 years). Most were involved in a stable relationship, either a legal (26%) or a common-law marriage (43%). Forty-four percent reported only one lifetime sexual partner, and another 26% reported only two lifetime partners. Ten percent of women were pregnant. Ninety-six women (23%) died after a median of 27 months (range, 3 to 51 months) from baseline evaluation. Among women who did not die, the median length of follow-up was 36 months.

Survival Based on the WHO Staging System

At the time of baseline evaluation, 35% of women were in WHO clinical stage I, 39% were in clinical stage II, 19% were in stage III, and 8% were in stage IV. Of the stage IV diseases listed in the Appendix, the only conditions diagnosed among women in our analysis were the HIV wasting syndrome, extrapulmonary cryptococcosis, esophageal candidiasis, extrapulmonary tuberculosis, and Kaposi sarcoma. Although stages I and II were associated with the best prognosis, stages III and IV overlapped, with stage III associated with a worse prognosis at 12 months (Figure 1). The 12-month cumulative mortality was 1% for women in WHO clinical stage I, 3% for those in clinical stage II, 16% for those in stage III, and 6% for those in stage IV; the 36-month cumulative mortality was 9% for women in stage I, 15% for those in stage II, 25% for those in stage III, and 27% for those in stage IV (P = 0.001).

In 391 women (95%), total lymphocyte counts were measured at baseline (median lymphocyte count, 2604 × 10⁶/L; lower and upper quartiles, 1968 × 10⁶/L and 3591 × 10⁶/L, respectively). Seventy-three percent of women were in WHO laboratory stage A, 24% were in stage B, and 3% were in stage C. Although stage C was associated with the worst prognosis, stages A and B overlapped, particularly during the first 18 months of follow-up (Figure 2). The 12-month cumulative mortality was 5% for women in WHO laboratory stage A, 4% for those in stage B, and 18% for those in stage C; the 36-month cumulative mortality was 15% for women in stage A, 19% for those in stage B, and 36% for those in stage C (P = 0.088).

Exploratory Analyses

Clinical Measurements

On the basis of previous studies of predictors of mortality [19], we selected four conditions for evaluation. Although only 3% of women had microscopically confirmed oral candidiasis in the year before baseline evaluation, this condition was associated with a significantly worse prognosis; the 36-month cumulative mortality was 50% for those with oral candidiasis and 15% for those without oral candidiasis (Table 1). Women with a history of herpes zoster in the preceding 6 years had a worse survival, although differences between women with and without such a history were not significant.

To study variables that might reflect the wasting syndrome, we evaluated both weight loss and body mass index. In 402 women, weight was measured at least twice in 1 year; these women were categorized as having a weight loss greater than 10% of body weight, a weight loss of 5% to 10% of body weight, or a weight loss of 5% or less of body weight. Differences in survival among these categories were not significant (Table 1). In contrast, body mass index was a strong predictor of mortality. In 390 women, both height and weight were measured at the baseline evaluation; the median body mass index was 21.5 kg/m² (lower and upper quartiles, 19.9 kg/m² and 23.4 kg/m², respectively). These women were categorized as having an index of 19 kg/m² or less, an index greater than 19 kg/m² but ≤ 21 kg/m², and an index greater than 21 kg/m². The 36-month cumulative mortality was 29%, 19%, and 12%, respectively; these differences in survival were statistically significant.

Although chronic mucocutaneous herpes simplex virus infection is considered a clinical stage IV condition, microbiological diagnosis of ulcerative lesions was not possible. We therefore evaluated a history of chronic (lasting at least 1 month) oral or genital ulcers, regardless of cause; acute (lasting less than 1 month) ulcers; or no ulcers. A history of an ulcerative lesion was associated with decreased survival; chronic ulcers were associated with the worst prognosis. The 36-month cumulative mortality was 32% for women with chronic ulcers, 21% for those with acute ulcers, and 14% for those with no ulcers (Table 1).

Laboratory Measurements

We evaluated four laboratory measurements. In 396 (96%) women, total leukocyte counts were measured at baseline evaluation (median count, 6125 × 10⁶/L; lower and upper quartiles, 4750 × 10⁶/L and 7775 × 10⁶/L, respectively). Mortality in women with leukocyte counts greater than and less than 5000 × 10⁶/L^did not differ significantly (Table 1). As noted above, the stratification of women into three stages on the basis of total lymphocyte count was not significantly associated with prognosis. Stratifying women into groups with lymphocyte counts greater than or less than 2000 × 10⁶/L^also did not result in significant differences.

In 395 women (96%), hematocrit was measured at the baseline evaluation (median hematocrit, 0.41; lower and upper quartiles, 0.38 and 0.43, respectively). The survival of women with measurements above and below the lower quartile significantly differed; the 36-month cumulative mortality was 42% for those with a hematocrit less than 0.38 and 11% for those with a hematocrit of 0.38 or more (Table 1).

In 390 women (95%), the erythrocyte sedimentation rate was measured at baseline (median rate, 39.5 mm/h; lower and upper quartiles, 15 mm/h and 65 mm/h, respectively). Women with rates above and below the upper quartile significantly differed in survival; the 36-month cumulative mortality was 32% for women with an erythrocyte sedimentation rate greater than 65 mm/h and 11% for those with an erythrocyte sedimentation rate of 65 mm/h or less (Table 1).

Selected Opportunistic Infections and Malignancies in Women Who Died

Among the 96 women who died, pulmonary or extrapulmonary tuberculosis developed after enrollment in 12 (median time to death after this diagnosis, 10.2 months). Kaposi sarcoma was diagnosed after enrollment in 6 women who died (median time to death, 4.6 months). Four women who died were reported to have cryptococcal meningitis (median time to death, 4.1 months), and 3 were reported to have esophageal candidiasis (median time to death, 0.9 months).

Revised Staging System

Clinical Staging

A revised clinical staging system was developed on the basis of modifications to the WHO proposal (see Appendix). We expanded the criteria for stage IV to include oral candidiasis, all chronic (lasting at least 1 month) oral or genital ulcerative lesions, and all cases of tuberculosis (whether pulmonary or extrapulmonary). Acute oral or genital ulcers were retained as stage II conditions (minor mucocutaneous manifestations). We also replaced the criterion of weight loss greater than 10% of body weight (in clinical stages III and IV) with that of a body mass index of 19 kg/m (²) or less and replaced the criterion of a weight loss less than 10% of body weight (in clinical stage II) with that of a body mass index between 19 kg/m² and 21 kg/m².

At baseline evaluation, 30% of women were in modified clinical stage I, 41% were in stage II, 18% were in stage III, and 12% were in stage IV (Figure 3). According to this system, women were differentiated in terms of survival, with no overlap. The 12-month cumulative mortality was 1% for women in modified clinical stage I, 3% for those in stage II, 7% for those in stage III, and 17% for those in stage IV; the 36-month cumulative mortality was 7% for women in modified stage I, 15% for those in stage II, 19% for those in stage III, and 36% for those in stage IV (P < 0.001).

Laboratory Staging

In contrast to the total lymphocyte count, an elevated erythrocyte sedimentation rate and a low hematocrit were both strongly associated with mortality. In our two-stage laboratory system, laboratory stage A was defined as an erythrocyte sedimentation rate of 65 mm/h or less and a hematocrit of 0.38 or more, and laboratory stage B was defined as an erythrocyte sedimentation rate greater than 65 mm/h or a hematocrit less than 0.38. On the basis of laboratory data, 392 women (95%) could be classified into one of two revised laboratory stages: 273 women (70%) were in modified laboratory stage A, and 119 (30%) were in modified laboratory stage B (on the basis of at least one abnormal laboratory value). The prognoses in the two stages differed for all time periods, and the differences were significant (Figure 4). The 12-month cumulative mortality was 1% for women in modified laboratory stage A and 13% for women in stage B; the 36-month cumulative mortality was 10% for women in modified stage A and 33% for women in stage B (P < 0.001).

Among the 390 women in whom both laboratory assays were done, the prognosis was worst for 47 women with a hematocrit less than 0.38 and an erythrocyte sedimentation rate greater than 65 mm/h (24-month cumulative mortality, 37%), followed by 23 women with a hematocrit less than 0.38 and an erythrocyte sedimentation rate of 65 mm/h or less (24-month mortality, 27%), 47 women with a hematocrit of 0.38 or more and an erythrocyte sedimentation rate greater than 65 mm/h (24-month mortality, 11%), and 273 women with a hematocrit of 0.38 or more and an erythrocyte sedimentation rate of 65 mm/h or less (24-month mortality, 4%) (P < 0.001).

Interaction of Clinical and Laboratory Stages

We next evaluated the prognostic significance of our modified laboratory staging system within each modified clinical stage. The survival for women stratified by both clinical and laboratory status at baseline evaluation is shown in Table 2. Because of small numbers and wide confidence intervals at 36 months, results are shown only through 24 months of follow-up. Within modified clinical stage I, laboratory stage did not significantly affect prognosis. Within modified clinical stages II through IV, however, laboratory stage B was associated with a significantly worse prognosis. Among women in laboratory stage A, mortality rates were similar in stages II through IV.

Combined Clinical and Laboratory Staging System

Because clinical status and laboratory results both have prognostic significance, we developed a single staging system that combines both clinical and laboratory criteria. We defined Kigali stage I as modified clinical stage I (asymptomatic infection), regardless of laboratory results. Kigali stage II was defined as modified clinical stage II or III [“mild” or “moderate” disease] and modified laboratory stage A. Kigali stage III was defined as either 1) modified clinical stage II or III and modified laboratory stage B or 2) modified clinical stage IV (“severe” disease) and modified laboratory stage A. Kigali stage IV was defined as modified clinical stage IV (“severe” disease) and modified laboratory stage B.

Four hundred two women (98%) could be classified into one of these four stages on the basis of clinical and laboratory criteria; 123 (31%) were in stage I, 160 (40%) were in stage II, 98 (24%) were in stage III, and 21 (5%) were in stage IV. These stages were significantly associated with prognosis (Figure 5). The 12-month cumulative mortality was 1% for women in combined stage I, 2% for those in stage II, 8% for those in stage III, and 33% for those in stage IV; the 36-month cumulative mortality was 7% for women in combined stage I, 10% for those in stage II, 29% for those in stage III, and 62% for those in stage IV (P < 0.001).

Discussion

On the basis of our analysis of data from a cohort of Rwandan women, we proposed a revised HIV staging system for use in sub-Saharan Africa. With respect to the clinical criteria, we suggested four major modifications to the WHO clinical staging system. First, we considered oral candidiasis to be a stage IV condition. The prevalence of oral candidiasis at any time was low, but this may partially reflect limited survival after this condition develops. Our findings are consistent with those of studies from industrialized countries, where oral candidiasis is a strong predictor of progression of and death from HIV-related disease [23-25].

Second, we replaced the weight loss criterion with that of body mass index, which was a better predictor of subsequent mortality. Our measure of body mass index, the Quetelet index, has been used as an indicator of body adiposity [26] and of adult chronic energy deficiency [27]. Use of this index requires only height and weight, thus allowing clinicians and researchers to determine status at a single point in time rather than relying on weight change over an unspecified interval. We selected our lower cut-point of 19.0 kg/m² on the basis of the body mass index distribution in this cohort; other investigators have suggested different index values as indicating chronic energy deficiency in developing countries [27]. Developing an appropriate measure of undernutrition is important for HIV staging systems in sub-Saharan Africa because the wasting syndrome is a major cause of HIV-related deaths [19].

Third, we defined any chronic oral or genital ulcer as a stage IV condition. This is in part a recognition that microbiological resources used to diagnose specific infectious agents such as herpes simplex virus are limited in many developing countries, particularly in settings outside urban or tertiary hospitals. In addition, if HIV-induced immune suppression impairs healing or response to standard therapy for other genital ulcer diseases such as chancroid [28], a chronic ulcer may still reflect advanced HIV disease. We retained acute ulcers as a stage II condition, which is consistent with the inclusion of minor mucocutaneous manifestations; of interest, a history of acute ulcers was associated with a prognosis somewhat worse than that for women with no ulcers.

Fourth, we included pulmonary tuberculosis as a stage IV condition. This is consistent with the current CDC and WHO classifications of pulmonary tuberculosis as an AIDS-defining disease in adults and adolescents [1, 3]. Tuberculosis is a common cause of pulmonary disease among HIV-infected persons in Rwanda [21, 29, 30]. Tuberculosis has developed at different degrees of immune suppression among HIV-infected African patients [31], causing some investigators to suggest that pulmonary tuberculosis may be of limited use as a marker of HIV disease stage. Although tuberculosis may develop at an earlier stage of immune suppression, it may also be associated with a worse survival once it occurs. According to previous data from this cohort, 5 (42%) of 12 HIV-infected women with pulmonary tuberculosis and no known extrapulmonary disease died within 15 months [21]. Investigators of a clinical trial of different regimens in Ugandan HIV-infected patients with pulmonary tuberculosis (excluding those who were severely debilitated) reported a 1-year survival rate of 65% to 72% [32]. In addition to death directly caused by tuberculosis, progression of HIV disease may be accelerated if tuberculosis adversely affects the immune system or enhances HIV expression [33, 34]. If so, this accelerated progression may lead to an increased risk for death from other immunodeficiency-associated illnesses.

Women with a history of herpes zoster had a somewhat worse prognosis than those who did not, although this difference was not significant. Previous studies from Rwanda have suggested that herpes zoster may be an early manifestation of HIV-induced immunosuppression that occurs before other HIV-related signs and symptoms develop [35].

This analysis was based on a cohort of African women, a group in whom the natural history of HIV disease is not completely understood. Data on gynecologic manifestations in these women are also being evaluated. On the basis of these and other analyses, clinical staging criteria may be expanded to include cervical neoplasia or other abnormalities.

With respect to laboratory criteria, CDC and WHO staging systems both include CD4+ lymphocyte counts [3, 7]. Although the CD4+ count provides important prognostic information, determinations of lymphocyte subsets require resources and technical expertise not routinely available in many settings in developing countries. In the WHO staging system, the total lymphocyte count is used as an alternative to the CD4+ count [7]. Studies from the United States indicating that the total lymphocyte count is a useful predictor of low CD4+ counts [36] suggest that this substitution may be reasonable in certain settings. However, in our cohort, classifications based on total lymphocyte count were not significantly associated with mortality. Our analysis does not support the use of a staging system that is based on total lymphocyte count in sub-Saharan Africa.

In contrast, both the erythrocyte sedimentation rate and hematocrit strongly predicted survival. Other studies have also shown that a low hematocrit or hemoglobin level and an elevated erythrocyte sedimentation rate are associated with progression of HIV disease or a decrease in CD4+ counts [37-40]. When these two measurements were combined, prognoses among women with symptomatic HIV infection could be differentiated, even within each clinical stage. One caution about the use of laboratory measures such as erythrocyte sedimentation rate and hematocrit in an HIV staging system is that they are relatively nonspecific and may be affected by various underlying illnesses, including conditions (such as malaria) that are not directly related to advanced HIV disease. However, our data suggest that laboratory results abnormal for any reason were associated with a significantly worse survival; therefore, these measures provide important prognostic information. Many of the clinical criteria proposed for use in developing countries (such as fever, diarrhea, and weight loss) are also fairly nonspecific and may be caused by various diseases. Because the laboratory tests we propose are inexpensive, routinely available in many clinical settings in developing countries, and provide prognostic information among symptomatic HIV-infected women, we believe they are useful additions to an HIV staging system. Additional researchers evaluating this system may wish to examine the effect of excluding persons with abnormal laboratory values that can be explained by diseases not related to HIV infection. As HIV staging systems become further refined, researchers may also wish to consider the utility of other prognostic laboratory assays such as β₂-microglobulin [41].

Use of any laboratory assay may be associated with both measurement and individual variability. For this reason, some investigators have recommended that all abnormal laboratory values at a certain cutoff level be confirmed by retesting a second specimen; this has been suggested for CD4+ counts [42]. To improve the specificity of our staging system, we suggest that in persons tested for erythrocyte sedimentation rate and hematocrit, at least two consecutive specimens should be obtained at least 1 month apart; for a given assay, both specimens would need to be abnormal.

Current CDC and WHO staging systems have separate clinical and laboratory axes, resulting in 9 possible combinations in the CDC system [3] and 12 in the WHO system [7]. This creates a potentially confusing situation as clinicians compare, for example, the relative prognosis of patients in category 3B with that of patients in category 4A. Our modified system consists of four clinical and two laboratory stages. We have also proposed a four-stage system that combines clinical and laboratory criteria and reflects the worse prognosis associated with both clinically severe disease and abnormal laboratory values. Having fewer stages may also help increase acceptance of a staging system.

Our staging system was developed on the basis of predictors of mortality in a cohort of HIV-infected women in Rwanda. This cohort was drawn from outpatient community-based clinics rather than from inpatient hospitals or specialized settings. We specifically developed our system for use in sub-Saharan Africa, which has been heavily affected by the HIV pandemic. This system should be considered a proposal for additional discussion and evaluation. In determining the applicability of this system, it is important that it be independently validated in different groups of HIV-infected persons in sub-Saharan Africa, including those persons enrolled in prospective cohort studies. On the basis of additional analyses and considerations, this initial proposal may be further modified and refined.

Our revised HIV staging system reflects the range of clinical outcomes from asymptomatic infection through serious illness, includes inexpensive and available laboratory assays, is related to prognosis, and can be used in sub-Saharan Africa. We recommend that additional investigators evaluate this staging system for use in both clinical management and research in developing countries, including trials of vaccines and therapies to prevent the progression of HIV disease.

Appendix

Key Features of the World Health Organization and of the Modified Clinical HIV Staging Systems

WHO clinical stage I: 1) asymptomatic infection; 2) persistent generalized lymphadenopathy; 3) acute retroviral syndrome; and 4) performance scale level at which patients can perform normal activities (not included in this analysis).

Modified clinical stage I: Same as WHO clinical stage I (adaptations as noted in the text).

WHO clinical stage II [“mild” disease]: 1) unintentional weight loss less than 10% of body weight; 2) minor mucocutaneous manifestations; 3) herpes zoster developing within the previous 5 years; 4) recurrent upper respiratory tract infections [for example, bacterial sinusitis]; and 5) performance scale level at which symptoms are present but patients are almost fully ambulatory (not included in this analysis).

Modified clinical stage II: Same as WHO clinical stage II (adaptations noted in the text), but body mass index between 19 kg/m² and 21 kg/m² is substituted for weight loss. Minor mucocutaneous manifestations include acute oral or genital ulcers.

WHO clinical stage III [“moderate” disease]: 1) unintentional weight loss greater than 10% of body weight; 2) chronic [lasting >1 month] diarrhea; 3) prolonged (lasting >1 month) fever [intermittent or constant]; 4) oral candidiasis; 5) oral hairy leukoplakia; 6) pulmonary tuberculosis developing within the previous year; 7) severe bacterial infections [for example, pneumonia or pyomyositis]; 8) vulvovaginal candidiasis that is chronic [lasting >1 month] or poorly responsive to therapy; and 9) performance scale level at which patients remain in bed less than 50% of daytime but more than normal (not included in this analysis).

Modified clinical stage III: Same as WHO clinical stage III [adaptations noted in text] but with the following changes: 1) Oral candidiasis is not included; 2) body mass index of 19 kg/m² or less is substituted for weight loss; and 3) pulmonary tuberculosis is not included.

WHO clinical stage IV [“severe” disease]: 1) the HIV wasting syndrome, defined as unexplained weight loss greater than 10% and either chronic diarrhea or chronic weakness and unexplained fever; 2) Pneumocystis carinii pneumonia; 3) toxoplasmosis of the brain; 4) cryptosporidiosis with diarrhea lasting longer than 1 month; 5) isosporiasis with diarrhea lasting longer than 1 month; 6) extrapulmonary cryptococcosis; 7) cytomegalovirus disease affecting organs other than the liver, spleen, or lymph nodes; 8) visceral or chronic [lasting >1 month] mucocutaneous herpes simplex virus infection; 9) progressive multifocal leukoencephalopathy; 10] any disseminated endemic mycosis; 11) candidiasis of the esophagus, trachea, bronchi, or lungs; 12] disseminated atypical Mycobacterium infection; 13) nontyphoidal Salmonella septicemia; 14] extrapulmonary tuberculosis; 15) lymphoma; 16] Kaposi sarcoma; 17) HIV-related encephalopathy [not included in this analysis]; and 18) performance scale level at which patients remain in bed more than 50% of daytime (not included in this analysis).

Modified clinical stage IV: Same as WHO clinical stage IV [adaptations noted in text] but with the following changes: 1) addition of oral candidiasis; 2) substitution of body mass index of 19 kg/m² or less for weight loss; 3) addition of chronic [lasting ≥ 1 month] oral or genital ulcer; and 4) addition of pulmonary tuberculosis.