Since 1985, the Johns Hopkins Hospital and University have operated a comprehensive primary and subspecialty clinical practice for treating patients with HIV infection. In this practice, patient management is individualized according to the needs and preferences of patients and their clinicians, but prophylaxis for opportunistic illness generally adheres to accepted treatment guidelines [21]. This practice allows the opportunity to assess the development over time of opportunistic disease as a complication of HIV infection, the distribution of the CD4⁺ counts at which opportunistic disease occurs, the duration of survival after opportunistic disease develops, and the effect of preventive interventions introduced in the past several years on the occurrence of opportunistic infections and cancers in clinical practice. The contemporary natural history of opportunistic disease in an urban population may be particularly relevant given the recent trends toward higher rates of HIV infection and the acquired immunodeficiency syndrome (AIDS) in women, members of racial and ethnic minority groups, and injection drug users [22, 23].

Methods

Top Methods Results Discussion Author & Article Info References

Study Sample

The Johns Hopkins AIDS Service provides longitudinal primary and subspecialty care for many HIV-infected patients in the Baltimore metropolitan area. To be registered in the clinic, patients must have a previous diagnosis of HIV infection. Each new patient has an initial comprehensive medical and psychosocial evaluation by a physician or physician's assistant, a nurse, and a social worker. In this evaluation, standardized instruments are used to collect extensive demographic, clinical, laboratory, pharmaceutical, and psychosocial characteristics [24]. Thereafter, similar data are collected every 6 months by review of all ambulatory and inpatient medical records, supplemented by automated reviews of Johns Hopkins Health System databases and reviews of all available records from facilities other than Johns Hopkins. Trained medical records technicians abstract these data onto standard data collection instruments. To ensure a high rate of longitudinal follow-up, medical records from other institutions in which the patients may have received care are routinely sought. Information on death is obtained from patient charts and from a separate death registry maintained by the clinic that receives reports from families, funeral homes, other medical institutions, and local coroners. In addition, the names of patients whose vital status is unknown for more than 12 months are searched for in the death records of the Maryland Bureau of Vital Records and the National Death Index. Nurses and physicians associated with the HIV clinic check the validity of a 10% sample of all collected data. If systematic error is identified, personnel are retrained, and the affected variable is reabstracted on all patients. The entry dates of our study population currently extend from July 1989 to December 1994. Follow-up extends to April 1995. Only about 12% of patients in the study group were censored before dying.

Definitions

We focused on opportunistic diseases that are included in Category C of the AIDS surveillance criteria of the Centers for Disease Control and Prevention [25]. The following diagnoses and criteria were used in our analyses:

1. Pneumocystis carinii pneumonia: a) microbiological identification using induced sputum or bronchoalveolar lavage fluid or b) symptomatic presentation with compatible chest radiologic study and clinical response to an appropriate therapeutic regimen.

2. Mycobacterium avium complex bacteremia: isolation from blood culture required but respiratory isolation alone not included.

3. Cytomegalovirus infection: retinitis diagnosed by clinically compatible examination by an ophthalmologist; colitis and esophagitis diagnosed by histopathologic confirmation of cytomegalovirus inclusions.

4. Mycobacterium tuberculosis disease: isolation in culture from pulmonary specimen, blood, or other tissue.

5. Toxoplasma gondii encephalitis: symptomatic clinical presentation with compatible computed tomographic scan or magnetic resonance image of the brain and response to appropriate therapy; nonresponse to therapy shown by histopathologic evidence of the organism; neurologist-confirmed diagnosis.

6. Candida esophagitis: symptomatic presentation of dysphagia or difficulty in swallowing with endoscopic evidence of invasive fungi or with clinical response to appropriate therapy.

7. Cryptococcal meningitis: evidence or culture of the organism in the cerebrospinal fluid.

8. Disseminated herpes simplex viral disease: identification of organism in tissue; includes chorioretinitis, visceral organ infection, central nervous system infection, or widespread cutaneous involvement.

9. Cryptosporidiosis: symptomatic clinical presentation with identification in stool specimen or intestinal biopsy specimen.

10. Kaposi sarcoma: typical clinical appearance or histopathologic confirmation.

11. Non-Hodgkin lymphoma: histopathologic confirmation of high-grade cancer.

12. Progressive multifocal leukoencephalopathy: symptomatic neurologic presentation with compatible radiologic imaging study of the brain and confirmation with or without biopsy by a neurologist.

13. The wasting syndrome: involuntary loss of more than 10% of the baseline body weight with chronic diarrhea or chronic weakness and enigmatic fever; no specific cause despite appropriate evaluation.

14. AIDS dementia complex: disabling cognitive and neurologic motor dysfunction interfering with activities of daily living; neurologist-confirmed diagnosis.

15. Herpes zoster: clinical presentation with pain and compatible dermatologic or disseminated rash.

Analysis

Our analyses focused on opportunistic disease that was diagnosed after enrollment into the Johns Hopkins HIV Clinic. We used the date at which each of these infections or cancers was diagnosed for the first time in the patient, with one exception: We analyzed both the first diagnosis of P. carinii pneumonia (primary P. carinii pneumonia) and the second diagnosis of P. carinii pneumonia (secondary P. carinii pneumonia). We limited the study group to patients whose CD4⁺ lymphocyte count at enrollment was less than 300 cells/mm³, because these infections or cancers seldom occur at higher CD4⁺ counts.

We calculated the incidence rate of each diagnosis by using the number of new diagnoses of the opportunistic disease that occurred after entry into the clinic as the numerator and the total number of person-years in patients who had not yet been assigned that specific diagnosis as the denominator. We calculated Poisson 95% CIs using standard formulas [26]. We further compared the incidence rates between the earlier period (1989 through 1992) and later period (1993 through 1995) using Poisson regression [27]. This definition of earlier and later periods provided an approximately equal amount of person-time in each comparison group. We also assessed the association between preventive drug therapy and the incidence of selected opportunistic infections using multivariate Poisson regression with adjustment for CD4⁺ lymphocyte count. These preventive therapies included prophylaxis for primary and secondary P. carinii pneumonia (trimethoprim-sulfamethoxazole, aerosolized pentamidine), toxoplasmosis (trimethoprim-sulfamethoxazole), M. avium complex bacteremia (rifabutin), and candidiasis and cryptococcal meningitis (fluconazole). Dapsone could not be analyzed because it was used infrequently. We defined these drugs as being used for preventive therapy if they were given regularly (usually once daily; trimethoprim-sulfamethoxazole, however, could be given three times weekly). Patients developing the opportunistic infection of interest were considered to have preventive therapy failure only if therapy had first been used at least 90 days before diagnosis. This decision was made to minimize the chance that these drugs were started as therapy in patients already suspected of having the infection.

We next determined the nearest CD4⁺ lymphocyte count within a maximum of 6 months before the diagnosis of each of the opportunistic illnesses. For each opportunistic disease, we examined the distribution of CD4⁺ counts, including the mean, median, and first and third quartiles (that is, the 25th and 75th percentiles). We then calculated the probability of developing each of the conditions over time using the Kaplan-Meier product-limit method [28]. These estimates were plotted for each diagnosis over 3 years.

Finally, we calculated Kaplan-Meier product-limit estimates for survival after diagnosis of each opportunistic infection. All-cause mortality was used for this analysis of survival.

We did all analyses using the SAS (SAS Institute, Cary, North Carolina) or EGRET (Statistics and Epidemiology Research, Seattle, Washington) software program.

Results

Top Methods Results Discussion Author & Article Info References

The demographic and clinical characteristics of the 1246 patients who had CD4⁺ lymphocyte counts of 300 cells/mm³ or less are shown in Table 1. The mean patient age was 35 years, 75% of patients were men, 76% were black, and 51% had injection drug use as a risk behavior for HIV transmission. The mean CD4⁺ count was 117 cells/mm³, and the mean duration of follow-up was 441 days. The most common opportunistic disease occurring before clinic enrollment was P. carinii pneumonia (239 patients [19%]). All other opportunistic complications occurred in no more than 10% of patients.

The overall incidence rate for each opportunistic disease is shown in Table 2. Incidence rates range from a low of 0.4 per 100 person-years for M. tuberculosis disease to a high of 13.3 per 100 person-years for Candida esophagitis. Between the earlier period (1989-1992) and the later period (1993-1995), the incidences of several opportunistic infections decreased (comparison done using Poisson regression): cryptococcal meningitis (earlier period, 3.0 per 100 person-years [95% CI, 1.9 to 4.4 per 100 person-years]; later period, 1.4 per 100 person-years [CI, 0.7 to 2.7 per 100 person-years]; P = 0.04); secondary P. carinii pneumonia (earlier period, 6.2 per 100 person-years [CI, 3.1 to 5.6 per 100 person-years]; later period, 3.7 per 100 person-years [CI, 2.5 to 5.5 per 100 person-years]; P = 0.01); and herpes zoster (earlier period, 4.2 per 100 person-years [CI, 2.8 to 6.0 per 100 person-years]; later period, 2.4 per 100 person-years [CI, 1.4 to 3.9 per 100 person-years]; P = 0.05). The change in incidence rates for all other opportunistic diseases had a 95% CI that included 1.0 and a P value greater than 0.10.

Table 2 also shows the mean, median, and first and third quartiles of the CD4⁺ count associated with each opportunistic disease. Figure 1 shows plots of the CD4⁺ counts (mean, median, and first and third quartiles). Only herpes simplex virus infection, M. tuberculosis disease, and herpes zoster occurred at median and mean CD4⁺ counts greater than 100 cells/mm³. The remaining infections occurred at a mean CD4⁺ count less than 100 cells/mm³ and a median CD4⁺ count less than 50 cells/mm³. Toxoplasmosis, progressive multifocal leukoencephalopathy, the wasting syndrome, secondary P. carinii pneumonia, cytomegalovirus infection, and M. avium complex bacteremia also occurred at a mean CD4⁺ count less than 50 cells/mm³.

View larger version (16K): [in this window] [in a new window]   Figure 1. Boxplot of the median (line inside the box), first quartile (bottom of the box), third quartile (top of the box), and mean (asterisk) CD4+ lymphocyte count at the time of the development of opportunistic disease. Can = Candida esophagitis; CMV = cytomegalovirus infection; Crp = cryptosporidiosis; Cry = cryptococcal meningitis; DEM = acquired immunodeficiency virus dementia complex; HSV = herpes simplex virus infection; HZos = herpes zoster; KS = Kaposi sarcoma; MAC = Mycobacterium avium complex bacteremia; NHL = non-Hodgkin lymphoma; PCP = primary Pneumocystis carinii pneumonia; PCP2 = secondary Pneumocystis carinii pneumonia; PML = progressive multifocal leukoencephalopathy; Tox = Toxoplasma gondii encephalitis; WS = the wasting syndrome.

We used Poisson regression to assess the association between the use of preventive therapy and selected opportunistic infections (Table 3). Fluconazole use was associated with an adjusted relative rate of 0.49 (P = 0.06) for cryptococcal meningitis and an adjusted relative rate of 0.61 (P = 0.005) for candidiasis. Rifabutin use was associated with a CD4⁺-adjusted relative rate of 0.37 (P = 0.002) for M. avium complex bacteremia. Use of trimethoprim-sulfamethoxazole was associated with a CD4⁺-adjusted relative rate of 0.33 (P = 0.02) for secondary P. carinii pneumonia and an adjusted relative rate of 0.66 (P = 0.08) for primary P. carinii pneumonia.

Figure 2 shows Kaplan-Meier estimates of the probability of opportunistic infection and cancer over time for disease with a 3-year probability greater than 0.10; Figure 3 shows the estimates for disease with a 3-year probability less than 0.10. The probability of Candida esophagitis is almost 0.30 by 3 years of follow-up, with the probability of P. carinii pneumonia (primary or secondary), cytomegalovirus, M. avium complex bacteremia, and AIDS dementia complex ranging between 0.10 and 0.20 by 3 years (Figure 1). Toxoplasmosis, herpes zoster, the wasting syndrome, and cryptococcal meningitis are the next most common disorders, with the 3-year probabilities ranging from 0.05 to 0.10 (Figure 2). Disseminated herpes simplex virus infection, cryptosporidiosis, progressive multifocal leukoencephalopathy, M. tuberculosis disease, Kaposi sarcoma, and non-Hodgkin lymphoma occurred with a probability less than 0.05.

View larger version (14K): [in this window] [in a new window]   Figure 2. Kaplan-Meier product-limit estimates of the probability of developing opportunistic disease over 3 years (probability greater than 0.10). CAN equals Candida esophagitis; CMV equals cytomegalovirus infection; DEM equals acquired immunodeficiency virus dementia complex; MAC equals Mycobacterium avium complex bacteremia; PCP equals primary Pneumocystis carinii pneumonia; PCP2 equals secondary Pneumocystis carinii pneumonia.

View larger version (17K): [in this window] [in a new window]   Figure 3. Kaplan-Meier product-limit estimates of the probability of developing opportunistic disease over 3 years (probability less than 0.10). CRP equals cryptosporidiosis; CRY equals cryptococcal meningitis; HSV equals herpes simplex virus infection; HZ equals herpes zoster; KS equals Kaposi sarcoma; MTB equals Mycobacterium tuberculosis disease; NHL equals non-Hodgkin lymphoma; PML equals progressive multifocal leukoencephalopathy; TOX equals Toxoplasma gondii encephalitis; WS equals the wasting syndrome.

Finally, Kaplan-Meier estimates of the probability of surviving after diagnosis of each opportunistic disease is shown in Table 4. The median estimates range from 35 days for non-Hodgkin lymphoma to 680 days for herpes zoster. In comparison, the median survival from an entry CD4⁺ count of 50 cells/mm³ is 580 days in a patient who did not develop any of these opportunistic diseases, with a 1-year probability of 0.71 (CI, 0.66 to 0.76).

Discussion

Top Methods Results Discussion Author & Article Info References

We analyzed the incidence and Kaplan-Meier probability estimates of the development of opportunistic infection and cancer in the past 5 years in an urban clinical practice of HIV-infected patients. Incidence rates ranged from a low of 0.4 per 100 person-years for M. tuberculosis infection to a high of 13.3 per 100 person-years for Candida esophagitis. The 3-year Kaplan-Meier probability of the occurrence of these two infections ranged from 0.01 to 0.30. Between 1989-1992 and 1993-1995, the incidence rates of cryptococcal meningitis, secondary P. carinii pneumonia, and herpes zoster decreased, but rates of the other opportunistic diseases did not change. After adjustment for CD4⁺ count by multivariate Poisson regression, fluconazole use was associated with a decreased rate of cryptococcal meningitis and candidiasis; rifabutin use was associated with a decreased rate of M. avium complex bacteremia; and trimethoprim-sulfamethoxazole use was associated with a decreased rate of secondary P. carinii pneumonia. In general, the CD4⁺ count at which these opportunistic diseases first occurred was relatively low. Only herpes zoster, M. tuberculosis disease, and herpes simplex virus disease first occurred at a median CD4⁺ count greater than 100 cells/mm³. All other opportunistic infections and cancers first occurred at a median CD4⁺ count less than 50 cells/mm³. The median duration of survival after the development of opportunistic disease ranged from a low of 35 days for non-Hodgkin lymphoma to a high of 680 days for herpes zoster.

The frequency of the occurrence of many of these opportunistic complications is lower than previously reported frequencies. A prominent example of this decreased rate is P. carinii pneumonia, which previously occurred in as many as 40% to 60% of patients with CD4⁺ count less than 200 cells/mm³ [8, 9, 29]. Although P. carinii pneumonia is still one of the five most common opportunistic infections, the probability of its occurrence is greater than 20% over 3 years of follow-up. The incidence rate is 9 per 100 person-years, and the median CD4⁺ count at the time of the development of P. carinii pneumonia is less than 50 cells/mm³. This decreased incidence of primary P. carinii pneumonia is consistent with the introduction of effective preventive therapy for P. carinii pneumonia in the late 1980s. The lack of a reduction in the incidence of primary P. carinii pneumonia over the 5 years of our analysis suggests that more effective prophylactic therapies are still needed to further reduce the incidence of this important opportunistic infection, which now develops in patients with later stages of immunosuppression. In our patient sample, the incidence of secondary P. carinii pneumonia decreased between 1989-1992 and 1993-1995, possibly because trimethoprim-sulfamethoxazole, relative to aerosolized pentamidine, was used more in the later period than in the earlier period [30, 31]. The incidence of cryptococcal meningitis also decreased between the earlier and later years. The increasing use of fluconazole since 1990, particularly in patients with more severe immunosuppression, may be responsible for this decrease [18, 32]. The incidence of the most common opportunistic infection in our patients, Candida esophagitis, did not decrease during these 5 years, but the relative risk for this complication in fluconazole recipients did decrease. Oral azole therapy is not recommended for routine prophylaxis because of the risk for the development of azole-resistant fungal infection. Whether the high incidence of this complication will decrease in subsequent years is unclear.

The incidence of herpes zoster also decreased in the later years. Whether this reflects greater use of antiviral therapy [33] or other factors is unknown. The incidence of herpes simplex virus and cytomegalovirus infections did not change, nor did that of Kaposi sarcoma or non-Hodgkin lymphoma. Both of these cancers have been associated with herpes-virus-like DNA sequences [34-36]. Notably, the previously predicted high incidence of non-Hodgkin lymphoma has not occurred [37]. Instead, the incidence of approximately 2 per 100 person-years remains similar to that previously reported in the late 1980s in another large cohort [12]. Kaposi sarcoma is less common in our patients than was described in earlier studies [38-40]. This finding may reflect the demographic composition of our primarily non-white population, in which injection drug use, rather than sexual transmission, is the predominant risk factor for HIV transmission.

Cytomegalovirus infection, M. avium complex bacteremia, and AIDS dementia complex continue to be relatively common opportunistic diseases, with incidences similar to that of P. carinii pneumonia. Until recently, no preventive therapy for cytomegalovirus infection has been effective. With the recently shown efficacy of oral ganciclovir, the incidence of cytomegalovirus infection may decrease [41]. The incidence of M. avium complex bacteremia did not diminish between the earlier period and the later period. Rifabutin, available since 1992, provides effective preventive therapy for M. avium complex bacteremia [17]. However, this drug has been limitedly used in our patient population. Among patients who did use rifabutin, however, the likelihood of developing M. avium complex bacteremia was reduced. The rate of toxoplasmosis has also not declined between the earlier period and the later period, despite the greater use of trimethoprim-sulfamethoxazole [20, 42-44]. As patients with AIDS survive longer, the potential benefit of some preventive therapies may be offset by an increased risk for opportunistic infection that is due to more severe and prolonged immunosuppression. The cause of AIDS dementia complex is currently unknown, but HIV type 1 may play the primary role [45]. Although AIDS dementia complex probably responds to higher-dose zidovudine [46], the effect of conventional doses of zidovudine on the incidence of this disease is unclear.

The rarest opportunistic infections among those we studied were progressive multifocal leukoencephalopathy, M. tuberculosis disease, and cryptosporidiosis. Our patients receive routine screening with tuberculin skin testing, and patients with a positive test result routinely receive preventive isoniazid therapy. Although HIV-related tuberculosis is not uncommon in Baltimore [47], the overall incidence of tuberculosis in Baltimore is substantially lower than that in other cities of similar size as a result of aggressive tuberculosis control efforts [48]. Our patients with chronic diarrhea have their stools routinely examined for cryptosporidia. Although this organism may not always be identified in a single stool specimen, the chronicity of cryptosporidiosis in advanced HIV disease makes underascertainment an unlikely cause of this low incidence. Progressive multifocal leukoencephalopathy, caused by the JC virus, has historically been a relatively uncommon opportunistic infection in HIV disease. It occurs in 2% to 3% of patients with AIDS [49], and our study does not suggest that this incidence has increased.

Comparing earlier published data with the CD4⁺ lymphocyte counts at which opportunistic disease develops in our patients is instructive [6]. In each disease, the median CD4⁺ count is now lower at the time of diagnosis that it was in the late 1980s. Given the greater awareness of these diagnoses and the clinical surveillance of our patients, we have no reason to believe that this is due to a diagnostic bias from later detection of disease. Instead, we believe that this result reflects the fact that all of these diseases are occurring later in the course of HIV infection, only after more severe immunosuppression has occurred. Another more recent study from 10 U.S. cities also indicated that the CD4⁺ counts at which AIDS-defining diseases were occurring in the early 1990s was less than 100 cells/mm³ [50]. The use of antiretroviral therapy, targeted preventive therapy for opportunistic disease, and more comprehensive clinical management have probably enabled patients to remain free of severe and life-threatening complications of HIV infection for longer periods. Even with a relatively low CD4⁺ count of 50 cells/mm³, patients who do not develop any of these opportunistic infections or cancers have a substantially longer duration of survival than those who develop almost any of these opportunistic diseases.

There are several potential caveats about the inferences that can be drawn from our data. First, the sample size for some of these opportunistic complications is relatively small. This is reflected particularly by the relatively wide 95% CIs around some of the Kaplan-Meier point estimates for survival beyond 2 years after onset of infection. Second, the men and women we studied are predominantly from an inner-city population characterized by a racial minority that has relatively low socioeconomic status and relatively high use of injection drugs. Thus, our results may be most generalizable to such a population. Although the medical management of these patients is individualized according to the needs and preferences of the patients and their clinicians, use of prophylaxis for opportunistic illness generally adheres to accepted contemporary treatment guidelines and standards. We have previously shown that receipt of medical care and the clinical outcomes of that care do not vary in any patient group by the major sociodemographic or behavioral characteristics of the patients [51]. Therefore, our results may be most generalizable to HIV-infected patients from urban populations who have reasonably good access to comprehensive HIV care. Finally, reported use of prophylactic therapy is based on prescription by the clinicians and does not necessarily capture patient adherence to the drug prescribed. If some of the patients classified as using a prophylactic drug were not taking the drug, the true effect of the drug in reducing the relative rate of disease may have been even greater.

Given these considerations, our analysis shows that opportunistic infections and cancers complicating HIV infection continue to occur commonly but now occur almost uniformly at more advanced stages of disease than in previous years. Our results also provide evidence of the effectiveness of several prophylactic medications in clinical practice. Nevertheless, opportunistic infection continues to be a relatively common complication of HIV infection, particularly when immunosuppression is advanced. Although our study suggests that fluconazole, rifabutin, and trimethoprim-sulfamethoxazole effectively reduce the risk for certain opportunistic infections, they do not eliminate the risk; other opportunistic illnesses for which no prophylactic therapies are effective continue to occur. By using prophylaxis for opportunistic disease as recommended by current guidelines [52], the clinician can expect the incidence and probability of opportunistic disease over 3 years of follow-up to be similar to those found in our patients. The expected duration of survival will vary depending on the complication.

In summary, our analysis of an urban clinical population indicates that opportunistic infections and cancer complicating HIV infection are first occurring at more advanced immunosuppression than in previous years. The incidences of P. carinii pneumonia, the most common opportunistic infection, and Kaposi sarcoma, the most common cancer, appear to have decreased compared with historical estimates from the late 1980s. The incidences of cryptococcal meningitis, herpes zoster, and secondary P. carinii pneumonia appear to have declined during the past 5 years. Cytomegalovirus, M. avium complex bacteremia, AIDS dementia complex, and toxoplasmosis continue to be relatively common, whereas non-Hodgkin lymphoma, M. tuberculosis disease, and progressive multifocal leukoencephalopathy remain relatively unusual in our patients. Continued progress in the development of preventive therapy for opportunistic disease should lead to further improvements in survival and quality of life in patients with advanced HIV infection.

Dr. Chaisson: Johns Hopkins University, 600 North Wolfe Street, Carnegie 292, Baltimore, MD 21287-6127.