| |
BMI Predicts Survival as well as <200 CD4s
|
| |
| |
"Body Mass Index at Time of HIV Diagnosis: A Strong and Independent Predictor of Survival, as strong as <200 CD4 count"
JAIDS Journal of Acquired Immune Deficiency Syndromes: Volume 37(2) 1 October 2004
van der Sande, Marianne A. B MD, MPH, PhD*; van der Loeff, Maarten F. Schim MD, MSc, PhD*†; Aveika, Akum A BSc*; Sabally, Saihou MD*; Togun, Toyin MD*; Sarge-Njie, Ramu MSc*; Alabi, Abraham S PhD*; Jaye, Assan PhD*; Corrah, Tumani FRCP, PhD*; Whittle, Hilton C FMedSci*
From the *Medical Research Council Laboratories, Fajara, The Gambia; and †London School of Hygiene and Tropical Medicine, London, UK.
SUMMARY
Background: Identification of basic prognostic indicators of HIV infection is essential before widespread antiretroviral therapy can be implemented in low-technology settings. This study assessed how well body mass index (BMI:kg/m2) predicts survival.
There is a pressing need to develop and test simple and affordable clinical criteria to guide interventions such as highly active antiretroviral treatment (HAART) in sub-Saharan Africa. In particular, the optimal starting time for HAART needs to be determined, because the expense of the drugs, potential side effects, and concerns about adherence, safety, and development of resistance must be balanced against the expected benefits of reduced morbidity and mortality. Unfortunately, access to laboratory facilities, which could support such decisions, is often limited in sub-Saharan Africa. The use of plasma viral load (PVL) and CD4+ measurement to guide initiation of treatment in resource-poor environments has been challenged, as there is a lack of data to support such an expensive strategy. Thus, it is important to evaluate simple clinical criteria that might assist decisions about the start of HAART.
Weight loss is a frequent symptom in HIV infection. Severe weight loss of >10% of body weight due to HIV infection itself is a common condition in HIV infection, which is used as a diagnostic criterion in the classification of HIV disease2 (1993 Centers for Disease Control [CDC]). Weight loss can also be due to a range of underlying conditions. Before HAART was available, both the wasting syndrome and malnutrition had been associated with poor survival in studies in industrialized countries, which were independent of immunosuppression as assessed by CD4+ counts.3,4 In most of sub-Saharan Africa, the background level of common infections5 and undernutrition is high, regular recorded weights in adults are rare, and few adults are able to document changes in their body weight. Thus, in sub-Saharan Africa, weight loss is usually not reliably reported or documented.
We considered that a low body mass index (BMI) (weight in kilograms divided by height in meters squared) at the time of initial HIV diagnosis could be a robust, affordable, and easily applicable clinical marker that may correlate well with survival. This analysis aimed to test this hypothesis in a seroprevalent cohort of HIV-1, HIV-2, and HIV dually infected patients.
Methods: BMI within 3 months of HIV diagnosis was obtained from 1657 patients aged >=15 years, recruited in a seroprevalent clinical cohort in The Gambia since 1992 and followed up at least once. Baseline CD4+ counts and clinical assessment at time of diagnosis were done.
Results: The mortality hazard ratio (HR) of those with a baseline BMI <18 compared with those with a baseline BMI >=18 was 3.4 (95% CI, 3.0-3.9). The median survival time of those presenting with a BMI <16 was 0.8 years, in contrast to a median survival of 8.9 years for those with a baseline BMI >=22. Baseline BMI <18 remained a highly significant independent predictor of mortality after adjustment for age, sex, co-trimoxazole prophylaxis, tuberculosis, reported wasting at diagnosis, and baseline CD4+ cell count (adjusted HR = 2.5, 95% CI 2.0-3.0). Sensitivity and specificity of baseline BMI <18 was comparable to that of a CD4+ count <200 in predicting mortality within 6 months of diagnosis.
Discussion: BMI at (within 3 mos) diagnosis is a strong, independent predictor of survival in HIV-infected patients in West Africa. In the absence of sophisticated clinical and laboratory support, BMI may also prove a useful guide for deciding when to initiate antiretroviral therapy.
The strength of the association between BMI and mortality and the predictive value is similar to that of the widely accepted CD4+ count cutoff at 200/μL. Further prospective studies are indicated to evaluate whether BMI could be used to inform intervention decisions, including the decision when to initiate HAART and whether monitoring BMI would be useful in assessing the success of such therapy.
ADDITIONAL RESULTS
Linear regression showed a weak but highly significant correlation between baseline BMI and baseline CD4+ (P < 0.0001; r 2 = 0.10). The median BMI was 17.9 (IQR 15.2-19.4) for patients with <200 CD4+ cells/μL, 19.5 (IQR 17.2-22.0) for those with a CD4+ cell count between 200-500 cells/μL, and 20.9 (IQR 18.7-23.3) for those with a CD4+ cell count >500 cells/μL. The association between baseline BMI and baseline Karnofsky score was also highly significant but weak (P < 0.0001; r 2 = 0.05).
Of the 1657 patients with a baseline BMI within 3 months of diagnosis, 1081 patients (65.3%) were diagnosed with HIV-1, 467 (28.1%) with HIV-2, and 98 (5.9%) were infected with both viruses (HIV-dual). No final HIV type-specific diagnosis could be made for 11 participants (0.7%). None of the patients received antiretroviral treatment during the period under evaluation. Since July 1999, 548 patients (33.1%) received co-trimoxazole prophylaxis initiated with a CD4+ cell count <500/μL. Baseline CD4+ cell counts collected within 3 months of recruitment were available for 1478 (89.2%) of the subjects.
Baseline BMI, CD4+ Cell Count, and Karnofsky Score
Of the 1657 patients eligible for the main analysis, median BMI at diagnosis was 18.8 (interquartile range [IQR], 16.3-21.5), being higher in women (19.6, IQR 16.7-22.4) than in men (18.3, IQR 16.1-20.3) (P < 0.001). The median CD4+ cell count was 250/μL (IQR 90-480) and was higher in women (310/μL, IQR 140-560) than in men (180/μL, IQR 60-370) (P < 0.001). The median Karnofsky score was 80 (IQR 70-90), and this was also higher in women (80, IQR 70-90) than in men (70, IQR 70-80, P < 0.001). At baseline, wasting was reported by 183 (11.0%) of patients; 267 (16.1%) were diagnosed with tuberculosis.
Baseline BMI and Survival
During follow-up, 849 of 1663 patients (51.1%) were known to have died, those with a low BMI being at greatest risk. Three hundred eighty-two patients (22.9%) not known to have died were lost to follow-up before June 30, 2002, because they moved out of the study area (52%), could not be found again (31%) or refused further contact with the study (16%). A total of 3623 person-years of follow-up observation were available for analysis. The median follow-up time was 16 months (range 0-123), and increased significantly according to baseline BMI. Overall mortality rate was 229 (95% CI 203-256) per 1000 person-years of observation, median survival 2.8 years. Survival differed markedly by baseline BMI. The median survival time of those presenting with a BMI <16 was 0.8 years, while for those with a baseline BMI >22 it was 8.9 years. The differences in survival time were highly significant (log-rank test P < 0.0001).
The overall mortality hazard ratio (HR) for patients with a BMI <18 at diagnosis compared with those with a BMI >=18 was 3.4 (95% CI 3.0-3.9). The HR of those in the lowest quintile (BMI <16) compared with those in the highest quintile (BMI >=22) was 6.4 (95% CI 5.0-8.2). This is similar to the HR of those with a CD4+ cell count <200 compared with those with a CD4+ cell count >=500 (HR 6.8, 95% CI 5.3-8.7).
The mortality HR for those with a BMI >18 versus BMI <18 varied with HIV type. In HIV-1 infection the HR was 3.1 (95% CI 2.6-3.7), for HIV-2 infection 4.3 (95% CI 3.2-5.7), and for dual infection 4.4 (95% CI 2.3-8.4).
The HR also varied with a baseline diagnosis of tuberculosis and was less strong for those diagnosed with tuberculosis at baseline (HR 1.8, 95% CI 1.3-2.5), compared with those not diagnosed as such (HR 3.8, 95% CI 3.3-4.4).
Multivariable analysis showed that the high HR of those with a BMI of <18 at enrolment was independent of type of HIV infection, age at diagnosis, sex, and co-trimoxazole prophylaxis; the strong effect on survival was also maintained upon adjustment for reported wasting or tuberculosis at diagnosis.
The effect of baseline BMI on survival was maintained albeit somewhat less pronounced after adjustment for baseline CD4+ count. The effect of low baseline BMI on survival was greatest among those with higher CD4+ cell counts. For patients with a CD4+ cell count <200 cells/μL, the HR was 1.9 (95% CI 1.6-2.4); for those with a CD4 cell count 200-500 cells/μL, it was 2.6 (95% CI 2.0-3.4); and for those with a CD4+ cell count >500 cells/μL, it was 3.6 (95% CI 2.2-5.8). Similar results were obtained when adjustment was made for CD4+ percentage rather than CD4+ cell counts (data not shown).
Comparable results were obtained when the effect on mortality was estimated per unit decrease in BMI: a 1-unit decrease in BMI resulted in a 21% increase in mortality rate (HR 1.21, 95% CI 1.19-1.24, P < 0.001). This association was independent of type of HIV infection, age at diagnosis, sex, co-trimoxazole prophylaxis, or tuberculosis diagnosis at baseline. The increase in mortality rate remained very significant after adjusting for reported wasting at diagnosis (HR 1.18, 95% CI 1.14-1.21, P < 0.001). Inclusion of baseline CD4+ cell count in the multivariable analysis only slightly reduced the effect of baseline BMI on survival, which remained highly significant (HR 1.13, 95% CI 1.10-1.17, P < 0.001).
Baseline RNA PVL measurements, within 3 months of diagnosis, were available in a subset of 36 HIV-1 (median 51,950 copies, range: <500-1,110,000) infected and 44 HIV-2 (median 6762 copies, range: <250-1,320,000) infected patients. Adjustment for baseline PVL did not alter the effect of baseline BMI on survival, but power was limited.
Survival in Patients With a First BMI <18 on Follow-Up
A total of 166 of the patients (17.4%) who had a baseline BMI of >=18 at diagnosis experienced a drop in their BMI to <18 before or on December 31, 2001. A total of 109 of these patients (65.7%) died during the period under observation. Median survival of this group, once their BMI became <18, was 0.8/year (IQR 0.2-2.2). The mortality rate once BMI dropped to <18 was 571/1000 person-years of observation, which is similar to the mortality rate of those with a BMI <18 at diagnosis (560/1000 person-years of observation). The mortality rate among those whose BMI remained >18 was 112/100 person-years of observation.
AUTHOR DISCUSSION
Our data show that baseline BMI recorded within 3 months of the diagnosis of HIV infection is a strong and independent predictor of mortality in this West African cohort. The magnitude of the predictive effect, and the sensitivity and specificity, were similar to those of the CD4+ cell count on mortality. The increased mortality risk for those with a low BMI was strongest for patients with higher CD4+ cell counts. This may have important implications for management and treatment of the increasing numbers of HIV-infected patients in sub-Saharan Africa.
A proportion (12.8%) of enrolled patients had no recorded baseline BMI measurement. Most of these were too weak to stand and would have needed to be measured lying down at diagnosis, which is culturally unacceptable. Many of this group were in very poor health and over half of this group died within 3 months. Conversely, the other group of participants who were excluded from the main survival analysis as their first BMI was obtained >3 months after recruitment appeared to have been in very good health, which might have led to the delay in revisiting the clinic. This group had a mean BMI similar to that in the highest BMI group at diagnosis. Similarly, patients lost to follow-up were concentrated in the group with a BMI >22 and had comparable indicators of good health. Therefore, these missing data may have diluted the observed strength of the association found between baseline BMI and survival.
Studies in industrialized countries before the advent of HAART have shown a correlation between poor nutritional status at diagnosis and clinical progression, but few data are available from sub-Saharan Africa. We are aware of only 3 other smaller studies in adult African populations, which assessed the short-term risk of low BMI at HIV diagnosis on mortality. The first was a study among 460 HIV-1-infected women in Rwanda, in which no data on CD4+ were reported13; the second was a study in Ivory Coast on 510 HIV-1- or HIV-D-infected patients with a total of 421 years of observation.14 Both studies showed a correlation between low BMI at diagnosis and subsequent higher mortality; the second study also found that this association was independent of CD4+ counts. As these studies followed patients for a maximum of 2 years, only the short-term risk of low BMI on mortality could be estimated. A third study among 259 tuberculosis patients in Malawi, of whom 80% were HIV infected, found that a low BMI (<17) at diagnosis was an independent risk factor for early mortality, defined as death within the first 4 weeks of starting tuberculosis treatment.15 Our study is much larger, has a longer follow-up, covers the 3 types of HIV infection, and could adjust for both clinical presentation and CD4+ cell count. It showed a persistent independent strong association between BMI and mortality; indeed the increased risk of a BMI <18 is comparable to the increased risk associated with a CD4+ cell count <200.
In a nation-wide study of BMI among Gambians aged >15 years, mean BMI for men was 20.1 (SD 2.9) and for women 21.3 (SD 3.9). This is nearly 10% higher than that found in newly diagnosed HIV patients and suggests significant weight loss may have preceded the diagnosis. The overall nutritional status of HIV-infected patients as assessed by the BMI at enrollment in this cohort is similar to another African study but lower than that reported from HIV studies in industrialized countries. Weight loss in HIV infection can have many causes. The poor prognosis associated with a low BMI at diagnosis suggests that the weight loss preceded diagnosis. However, adjustment for recorded wasting at diagnosis did not remove the association, which suggests that patients may not perceive significant weight loss to be a recent event. A detailed follow-up study of homosexual HIV-1 seroconverters in the Netherlands found a biphasic pattern of BMI: after a long relatively stable period, a rapid decline occurred in the 6 months preceding the onset of AIDS. The steeper the decline in BMI, the faster was the progression to AIDS. Thus the strong association of a low BMI at diagnosis with poor survival that was found in our study could be explained by presentation with AIDS wasting syndrome, which is associated with a poor outcome; it may also reflect an underlying poor prognosis.
Clinical criteria used as indicators of prognosis may vary according to environment. Thus it is debatable if a diagnosis of tuberculosis, as listed in the widely used CDC criteria, should be an AIDS-defining condition in situations like sub-Saharan Africa, where there is a high background incidence of tuberculosis. Many clinical criteria in the CDC guidelines are difficult to apply in resource-poor environments because they require sophisticated and expensive investigations. They are unsuitable for less qualified health workers who will need to make a clinical diagnosis before deciding to treat. Thus, there is clearly a need for less sophisticated prognostic clinical indicators, and progress has been made to identify these. Measurement of height and length in particular does not require advanced training nor sophisticated technical skills, needs minimal resources, and can be applied in rural health centers.
A study in the United States suggested that high PVL could be causally related to weight loss. In the present analysis, which is limited by small numbers, the effect of baseline BMI on survival appears independent of the baseline PVL. We have previously shown in a subset of this clinic population that CD4+ cell count appeared a stronger predictor of survival than PVL.
A previous survival analysis in this cohort showed that HIV-2-infected patients with CD4+ cell counts >500 cells/μL had a significantly lower mortality rate than HIV-1-infected patients. HIV-2-infected patients with advanced disease had the same poor prognosis as patients with HIV-1. Dually infected patients had mortality rates similar to HIV-1 across all CD4+ cell categories.8 Our data now show that within each type of HIV infection, independent of baseline CD4+ cell count, survival was strongly predicted by BMI at diagnosis. The relatively modest, albeit highly significant, correlation between BMI and CD4+ cell count, as well as the observation that BMI remained an independent predictor of survival after controlling for CD4+ cell count, suggests that BMI captures a different physiological risk for death than the risk reflected by CD4+ cell count.
We conclude that BMI at diagnosis is a low-technology, affordable, prognostic indicator, independent of age, sex, CD4+ cell count, or HIV type. The strength of the association between BMI and mortality and the predictive value is similar to that of the widely accepted CD4+ count cutoff at 200/μL. Further prospective studies are indicated to evaluate whether BMI could be used to inform intervention decisions, including the decision when to initiate HAART and whether monitoring BMI would be useful in assessing the success of such therapy.
|
|
| |
| |
|
|
|
