Abstract
Older HIV-positive patients may experience reduced benefit and increased toxicities from combination antiretroviral therapies (cART) due to late diagnosis, weakened immune systems, and other age-related physiological changes. This study investigates the effect of age on time to virologic suppression and CD4 cell count response to cART. Data were collected from a tertiary care immunodeficiency clinic in Toronto. HIV-positive patients with cART initiation after 1/1/1998 were included. Log logistic accelerated failure time models were used to estimate the effect of age on time to virologic suppression. Mixed linear models were used to evaluate CD4 cell response to treatment. A total of 502 patients were studied; 445 were less than 50 years old and 57 were age 50 or over. Of these 73% were male. Thirty-one percent started treatment with boosted protease inhibitors and 43% with nonnucleoside reverse transcriptase inhibitors. In a log logistic model adjusting for gender, immigration status, AIDS-defining illness, years since HIV diagnosis, baseline CD4 count and viral load, cART type, calendar year of cART initiation, and hepatitis C diagnosis, older age was not associated with time to virologic suppression (n=418, time ratio=0.94, p=0.20). In a multivariable mixed linear regression model adjusting for the same covariates, age was not associated with CD4 cell count response (n=418, β=0.34, p=0.96). Time to virologic suppression and immunologic response were not significantly different among older and younger patients. Further studies should investigate adherence, comorbidities, and regimen changes in addition to the current covariates.
T
Response to combination antiretroviral therapy (cART) may be different among older patients for several reasons. Older patients may present with higher HIV viral load at diagnosis, 3 which may result in longer times to virologic suppression than younger patients. Immunosenescence due to age results in lower CD4 cell counts in older patients, 3 –6 which is exacerbated by HIV disease progression. 3,5,6 Immunologic recovery may therefore be less robust in older patients compared to younger patients. 3,4,6 The effectiveness and toxicities of cART may be different in older patients due to changes in body fat, 4 –7 kidney function, 4 –7 liver function, 4 –7 thymus function, 3 –7 drug pharmacokinetics, 4,6 and comorbidities, 3,4,7 which may lead to decreased virologic and immunologic response. The purpose of the current study was to investigate the effects of age on time to virologic suppression and CD4 cell response among naive HIV-positive individuals initiating cART.
This was an observational retrospective cohort study of HIV-positive patients attending the THIC, a specialist based tertiary care clinic that accepts referrals from primary care physicians and is affiliated with the University of Toronto. CD4 counts and viral loads are measured 1 month after treatment initiation or a regimen change and every 3–4 months afterward. Approval for this project was obtained from the University Health Network Ethics Review Board.
Patients who initiated cART after January 1, 1998 were included in the study, where cART was defined as at least three drugs from at least two classes or three nucleoside reverse transcriptase inhibitors (NRTIs). Demographic and clinical data were collected from manual chart abstraction and monthly downloads from the clinic's electronic laboratory database. Patients were excluded if there was no baseline CD4 count or viral load or if there were fewer than two CD4 counts or viral load measurements in the first year of treatment.
The variable of interest for this study was age at cART initiation. Age was modeled with a continuous variable, which assumed equal effects of age on virologic and immunologic response over the age range. Age was also assessed with a binary variable (<50 vs. ≥50 years) and a categorical variable with four levels (<30, 30 to <40, 40 to <50, and ≥50 years) in order to test for potential nonlinear effects.
Demographic covariates included gender and immigration status (i.e., arrival in Canada within 10 years prior to cART initiation). Risk factor variables for HIV infection were men who have sex with men (MSM), injection drug use (IDU), immigration from an endemic country, and heterosexual transmission. Clinical covariates included hepatitis C diagnosis; time since HIV diagnosis; AIDS-defining illness (ADI) prior to cART; baseline CD4 count and viral load; type of regimen at cART initiation [nonnucleoside reverse transcriptase inhibitor (NNRTI) based, unboosted protease inhibitor (PI) based, boosted PI based, or other]; and calendar year of cART initiation.
Time to virologic suppression was calculated as the time from cART initiation to the first of two consecutive viral load measurements less than 50 copies/ml. Individuals who did not achieve virologic suppression during the study period were censored at the date of their last viral load. Absolute CD4 counts during the first 5 years after cART initiation were used to assess CD4 cell response.
Demographic, baseline, and follow-up variables were compared by age group (<50 vs. ≥50 years) using the Wilcoxon rank sum test for continuous variables, the chi-square test or Fisher's exact test as appropriate for nominal variables, and the Cochran-Armitage Trend test for ordinal variables. A two sample t-test was used to compare the mean change in CD4 count at 1 year after treatment initiation between age groups.
Kaplan–Meier estimates were used to determine median times to virologic suppression. The logrank test was used to test for differences in the distribution of time to virologic suppression between age groups. Since the proportional hazards assumption was violated, accelerated failure time models were considered to model time to virologic suppression. The log-logistic model was chosen as the best fitting parametric model for the data. Univariate models were fit for the following covariates, which were a priori believed to be associated with time to virologic suppression: age, gender, immigration status, ADI prior to treatment, time since HIV diagnosis, baseline viral load, baseline CD4 count, cART type, year of cART initiation, and hepatitis C diagnosis. These covariates were included in a multivariable model with continuous age. Time ratios (TR) were calculated for all models by exponentiating the coefficient estimates. Time ratios represent the multiplicative increase (TR >1) or decrease (TR <1) in time to virologic suppression associated with a unit increase in the covariate of interest. 8
Mixed linear models 9 were used to evaluate the longitudinal CD4 count response to treatment. Based on clinical knowledge of CD4 count increase and exploration of the data, a breakpoint of 1.5 years after treatment initiation was chosen to model the initial and long-term increases in CD4 count separately. Correlation of CD4 counts within patients was accounted for with a spatial Gaussian correlation matrix. This structure takes into account the variable lengths of time between measurements, where measurements further apart in time are less correlated than measurements taken closer together in time. Univariate models were fit for each of the following variables after adjusting for baseline CD4 count: age, gender, immigration status, ADI prior to treatment, time since HIV diagnosis, baseline viral load, cART type, year of cART initiation, hepatitis C diagnosis, time from cART initiation to 1.5 years after cART initiation, and time from 1.5 to 5 years after cART initiation. A multivariable model was fit with age as a continuous variable and all other covariates, which were a priori believed to be associated with CD4 counts. Statistical analyses were performed using SAS Statistical Software Version 9.3 by SAS Institute Inc., Cary, NC.
A total of 628 HIV-positive patients initiated cART at THIC after January 1, 1998. Of these, 22 were missing baseline data and 117 patients had fewer than two CD4 counts or viral loads in the first year of treatment (81 of these patients had less than 1 year of follow-up). Thirteen additional patients initiated cART elsewhere, but had sufficient data at THIC to be included. The final sample included 502 patients. Eighty-three patients were missing data on immigration status and one patient was missing data on HIV diagnosis date. Demographic, risk factor, and baseline clinical characteristics are compared between two age groups in Table 1.
Due to missing data: immigration in previous 10 years n=372 for age <50 and n=47 for age ≥50; immigration from endemic country n=371 for age <50 and n=47 for age ≥50; time since HIV diagnosis n=444 for age <50 and n=57 for age ≥50.
Risk factors are not mutually exclusive.
ADI, AIDS-defining illness; cART, combination antiretroviral therapy; NNRTI, nonnucleoside reverse transcriptase inhibitor; PI, protease inhibitor.
The median duration of follow-up was 4.3 years (IQR=2.0, 7.6). The median number of follow-up visits in the first 5 years of treatment was 14 (IQR=8, 20). No differences were found between age groups (p=0.11 and p=0.25, respectively).
A total of 459 patients (91%) achieved virologic suppression during follow-up. The median time to virologic suppression was 18.9 weeks for patients less than 50 years of age (95% CI=17.4, 20.1) and 20.6 weeks for patients aged 50 and over (95% CI=16.9, 27.7) (p=0.96).
Results of unadjusted and adjusted time-to-event analyses are presented in Table 2 with age as a continuous variable. Age was not associated with time to virologic suppression in an unadjusted model or a model adjusted for covariates. Higher baseline viral load, an unboosted PI cART regimen, and cART initiation in the time period 1998–2003 were significantly associated with longer times to virologic suppression in univariate and multivariable models. Immigration in the 10 years prior to cART initiation was significantly associated with a shorter time to virologic suppression in both models. No significant effects of age were seen with either binary or categorical parameterizations of age. To test for potential bias due to missing data on immigration status, the multivariable model was fit without immigration status, which resulted in similar estimates and conclusions.
Time ratios (TR) are estimated from accelerated failure time models and represent the multiplicative expansion or compression of the average time to virologic suppression compared to a reference group. A TR >1 represents longer times to virologic suppression on average. A TR <1 represents shorter times to virologic suppression.
N=418 due to 83 missing values for immigration status and 1 missing value for time since HIV diagnosis.
The mean CD4 count and total number of patients at baseline and across 5 years of follow-up are plotted by four age groups in Fig. 1. One year after cART initiation, patients less than 50 years old had a mean increase of 175 cells/mm3 (SD=147, n=392) from baseline CD4 count and patients over 50 had a mean increase of 192 cells/mm3 (SD=164, n=50) from baseline (p=0.45).
Mean CD4 counts by age at initiation of combination antiretroviral therapy (cART).
Univariate models adjusting for baseline CD4 count and multivariable mixed linear models are displayed in Table 3 with age as a continuous variable. Age was not associated with CD4 counts over time in a model adjusted for baseline CD4 count or in a model adjusted for all covariates. Male gender and lower baseline CD4 count were significantly associated with lower CD4 counts over time in the multivariable model. Increasing time on cART was associated with increasing CD4 counts in the model, with a greater rate of increase per year in the initial 1.5 years after cART initiation compared to the remaining time period. The effect of age was not significant when included as a binary or categorical variable in the multivariable model.
N=418 due to 83 missing values for immigration status and 1 missing value for time since HIV diagnosis.
In the current study of individuals starting their first cART regimen in a tertiary care, urban center, age was not associated with time to virologic suppression or CD4 response. These findings are consistent with a large retrospective cohort study of over 5000 HIV-positive patients that found no significant association between age and achievement of virologic suppression or CD4 count response (2 to 6 years after treatment initiation) after adjusting for adherence and other covariates. 10 This study adds to the literature a more current dataset with a diverse patient population in a universal health care setting. The literature overall remains inconsistent. 3 –7,10 Most studies show older patients have either increased or similar rates of virologic suppression and have poorer or similar CD4 count responses to treatment compared to younger patients. 3 –7 These discrepancies are most likely due to a wide variety in design, sample size, covariates, and outcomes across studies. 3,4,10
A limitation of the current study was an inability to account for adherence to antiretroviral medication as this information is not captured in the current database. The sample size was not sufficiently large to account for many comorbidities. Treatment regimen changes over time were not accounted for. A strength of this study was the clinic population with a wide range of HIV risk factors and demographic characteristics and the examination of findings outside of the randomized control trial setting.
Although the immunologic and virologic responses of older patients were not found to differ from younger patients in the current study, older patients are a unique and growing population. Future studies should account for adherence, comorbidities, and regimen changes in addition to the current covariates when estimating the effect of age on virologic and immunologic response.
Footnotes
Acknowledgments
J.R. and S.W. were supported by Career Scientist Awards from the Ontario HIV Treatment Network. J.R. was also supported by the Skate the Dream Fund, Toronto and Western Hospital Foundation.
Author Disclosure Statement
No competing financial interests exist.