| |
Trends in the incidence of cancers among HIV-infected persons and the impact of antiretroviral therapy: a 20-year cohort study
|
| |
| |
"For all outcomes, increased age was significantly associated with an increased risk of a cancer event.....Our data suggests that the aging of the HIV population is the primary factor associated with the rising rates of NADCs."...."HIV itself could play a role either by a direct oncogenic effect (e.g., HIV tat gene) [32] or as a consequence of immunosuppression with diminished tumor surveillance"
AIDS:Volume 23(1)2 January 2009p 41-50
Crum-Cianflone, Nancya,b; Hullsiek, Katherine Hupplera,c; Marconi, Vincenta,d; Weintrob, Amya,e; Ganesan, Anuradhaa,f; Barthel, R Vincenta,g; Fraser, Susana,h; Agan, Brian Ka,d; Wegner, Scotta
aTriService AIDS Clinical Consortium, Infectious Disease Clinical Research Program, Uniformed Services University of the Health Sciences, Bethesda, Maryland, USA
bInfectious Disease Clinic, Naval Medical Center San Diego, San Diego, California, USA
cDivision of Biostatistics, University of Minnesota, Minneapolis, Minnesota, USA
dInfectious Disease Clinic, San Antonio Military Medical Center, San Antonio, Texas, USA
eInfectious Disease Clinic, Walter Reed Army Medical Center, Washington, District of Columbia, USA
fInfectious Disease Clinic, National Naval Medical Center, Bethesda, Maryland, USA
gInfectious Disease Clinic, Naval Medical Center Portsmouth, Portsmouth, Virginia, USA
hInfectious Disease Clinic, Tripler Army Medical Center, Honolulu, Hawaii, USA.
"The most frequent cancer in our study cohort was Kaposi's sarcoma followed by NHL, both of which occurred most commonly during the pre-HAART era. Predictors for these cancers included low time-updated CD4 cell counts and the lack of HAART use. These data emphasize the importance of early HIV diagnosis and initiation of therapy before low CD4 cell counts occur......The most common NADCs (non-AIDS cancers) in our cohort were non-Kaposi's sarcoma skin cancers and anal cancer. For both of these cancers, the rates were fairly stable from 1984 to 2000, with increased rates during the period 2001 to 2006. The increasing rate of anal carcinoma was statistically significant (13-fold over the study period).....Reasons for the recent increased importance of NADCs among HIV-infected persons are likely several-fold. Increasing life expectancy [24,25] and the reduction in competitive causes of death [26] are undoubtedly contributory. Viral coinfections such as the human papillomavirus, which may have a relatively long latency before their oncogenic effect, are prevalent in HIV patients and may play a role in cancer development [27-29]. In addition, HIV-infected patients may have higher rates of behaviors, such as tobacco use, which contributes to cancer development [30,31]. Finally, HIV itself could play a role either by a direct oncogenic effect (e.g., HIV tat gene) [32] or as a consequence of immunosuppression with diminished tumor surveillance [33].......in our study, patients with NADCs had robust CD4 cell counts at diagnosis (the median was 430 cells/μl) and 79% of the NADC events in the late post-HAART era occurred at a CD4 cell count more than 350 cells/μl. Other studies have also noted that NADCs, such as lung cancer, may not be related to advanced immunosuppression as measured by CD4 cell counts or HIV viral loads [9,11,30,34-36]. This suggests that strategies beyond achievement of high CD4 cell counts may be necessary for reduction in NADCs. Such strategies may include behavioral modifications, such as smoking cessation, safe-sex practices to reduce viral coinfections, and early recognition and management of viral hepatitis.....Whether initiating HAART earlier than advocated by previous guidelines [38] would be beneficial for cancer prevention via mechanisms such as decreased immune activation or suppression of oncogenic viruses is unknown. Research examining the effect of earlier initiation of antiretroviral therapy on the incidence of AIDS-defining and non-AIDS-defining conditions, such as cancers, is under development.....Our cohort consisted primarily of nondrug users with a low prevalence of hepatitis C; hence, data regarding the effect of drug use or hepatitis coinfection on NADC rates could not be evaluated. Furthermore, we did not collect data on behaviors such as tobacco use, alcohol use, or on family history of cancer. In addition, our population was primarily men; hence, women-specific tumors could not be adequately studied....NADCs may develop independently of CD4 cell count and antiretroviral therapy, concurring with the rising rates of these cancers during the HAART era. Our data suggests that the aging of the HIV population is the primary factor associated with the rising rates of NADCs."
Abstract
Objective: To describe trends in incidence rates of AIDS-defining cancers (ADCs) and non-AIDS-defining cancers (NADCs) during the HIV epidemic and to evaluate predictors, including the impact of antiretroviral therapy, of cancer development.
Design: Retrospective analysis of a multicenter, prospective natural history study including 4498 HIV-infected US military beneficiaries with 33 486 person-years of follow-up.
Methods: Predictors evaluated included demographics, clinical data, time-updated CD4 cell counts, HIV viral loads, and antiretroviral history. Time periods were classified as early pre (1984-1990), late pre (1991-1995), early post (1996-2000), and late post (2001-2006) HAART eras. Cox proportional hazard models were used to evaluate the association of specific factors with cancer.
Results: Ten percent of HIV-infected persons developed cancer. ADC rates increased between the early and late pre-HAART eras (7.6 and 14.2 cases per 1000 person-years) and have since declined from 5.4 to 2.7 in the early and late HAART eras, respectively (P < 0.001). Rates of NADCs have risen over the four periods (2.9, 2.8, 4.2, 6.7, P = 0.0004). During the late HAART era, 71% of cancers were NADCs. Predictors for ADCs included low CD4 cell count, noncancer AIDS diagnosis, and lack of HAART. NADCs were predicted by increasing age and white race (due to skin cancers).
Conclusion: Although the rate of ADCs continues to fall, the rate of NADCs is rising and now accounts for the majority of cancers in HIV-infected persons. The development of NADCs is associated with increasing age among HIV patients. HAART use is protective for ADCs, but did not significantly impact NADCs.
Introduction
Cancers such as Kaposi's sarcoma were among the initial clinical diagnoses that led to the recognition of human immunodeficiency virus (HIV) infections in 1981 [1]. Some experts in the 1980s suggested that malignancies would cause a second epidemic, which was realized with the occurrence of Kaposi's sarcoma and lymphoma [2]. Subsequently, three cancers were classified as AIDS-defining cancers (ADCs), including Kaposi's sarcoma, non-Hodgkin's lymphoma (NHL), and invasive cervical carcinoma (ICC) [3,4].
With the advent of HAART in 1996, the rates of Kaposi's sarcoma and NHL of the central nervous system have dramatically fallen, with less effect on ICC and systemic NHL rates [5-10]. Simultaneously, non-AIDS-defining cancers (NADCs) have accounted for an increasing proportion of cancer cases reported in HIV-infected individuals. Recent studies have reported that NADCs represented 13% of deaths during the HAART era, compared to less than 1% in the pre-HAART era [11], and that fatal NADCs are now more common than fatal ADCs [12]. However, other research has shown conflicting results regarding incidence rates of NADCs [13,14]. Further evaluation of cancer trends in large and diverse HIV-positive cohorts that include early-stage HIV patients is needed.
We evaluated prospectively collected data from the 23-year observational Tri-Service AIDS Clinical Consortium (TACC) HIV Natural History Study (NHS) to further investigate trends in the rates of ADCs and NADCs among HIV-infected persons. Further, given the availability of individual patient data, we assessed whether CD4 cell counts, HIV viral loads, or antiretroviral medications were predictors of cancer occurrence among HIV-infected persons.
Discussion
This study, with extended follow-up in the late-HAART era, provides current data on the trends for AIDS-defining and non-AIDS-defining cancers. We found that rates for ADCs have continued to decline well after the advent of HAART. Conversely, the rates of NADCs have continued to rise over time and now account for the majority of the cancers in our study cohort; the development of NADCs appears most related to increasing age. HAART use was protective for ADCs, but did not significantly impact NADCs.
Studies after the advent of HAART noted dramatic declines in the rates of ADCs [5-10]; however, data on the late post-HAART era are more limited [17-19]. A recent study suggested that ADC rates have stabilized during the HAART era [9], whereas our study demonstrates that the rate of ADCs is progressively decreasing. These contrasting results may be due to the study periods considered (our study has follow-up up to 2006, whereas other studies include follow-up only up to 2002) and the general health of the cohorts considered (our cohort was composed primarily of people with earlier stages of HIV infection, whereas other cohorts included only those with AIDS). A more recent study (1996-2005) performed in an urban HIV clinic had findings similar to ours showing that ADC rates continue to decline [20]. The decreasing rates of ADCs may be related to the improving efficacy and tolerability of antiretroviral therapies.
The most frequent cancer in our study cohort was Kaposi's sarcoma followed by NHL, both of which occurred most commonly during the pre-HAART era. Predictors for these cancers included low time-updated CD4 cell counts and the lack of HAART use. These data emphasize the importance of early HIV diagnosis and initiation of therapy before low CD4 cell counts occur.
The most common NADCs in our cohort were non-Kaposi's sarcoma skin cancers and anal cancer. For both of these cancers, the rates were fairly stable from 1984 to 2000, with increased rates during the period 2001 to 2006. The increasing rate of anal carcinoma was statistically significant (13-fold over the study period). Of note, formal cancer screening practices did not change in our cohort during recent years to explain the increased rate; however, data on trends in the number of MSM in our study cohort is not available. Other cancers, such as renal and prostate carcinoma, reached their highest rates during the late post-HAART period, but the trends were not significant, likely due to the small sample size of the individual cancer types; other studies have also reported an increase in these cancers during the late-HAART era [9,13,21]. We did not find an increase in the incidence of Hodgkin's disease as seen in some studies. [9,22,23].
The proportion of cancer due to NADCs increased over the study period similar to other reports in the literature [13]. Reasons for the recent increased importance of NADCs among HIV-infected persons are likely several-fold. Increasing life expectancy [24,25] and the reduction in competitive causes of death [26] are undoubtedly contributory. Viral coinfections such as the human papillomavirus, which may have a relatively long latency before their oncogenic effect, are prevalent in HIV patients and may play a role in cancer development [27-29]. In addition, HIV-infected patients may have higher rates of behaviors, such as tobacco use, which contributes to cancer development [30,31]. Finally, HIV itself could play a role either by a direct oncogenic effect (e.g., HIV tat gene) [32] or as a consequence of immunosuppression with diminished tumor surveillance [33].
Factors associated with the development of an NADC in our study included increasing age and white race, although the association with race was restricted to skin cancers. We did not find an association with NADCs and prior noncancer AIDS events or time-updated CD4 cell counts. In fact, in our study, patients with NADCs had robust CD4 cell counts at diagnosis (the median was 430 cells/μl) and 79% of the NADC events in the late post-HAART era occurred at a CD4 cell count more than 350 cells/μl. Other studies have also noted that NADCs, such as lung cancer, may not be related to advanced immunosuppression as measured by CD4 cell counts or HIV viral loads [9,11,30,34-36]. This suggests that strategies beyond achievement of high CD4 cell counts may be necessary for reduction in NADCs. Such strategies may include behavioral modifications, such as smoking cessation, safe-sex practices to reduce viral coinfections, and early recognition and management of viral hepatitis.
This study showed an association between HAART use and reduced rates of ADCs among individual HIV patients. Our study had the advantage of examining time-updated HAART data among individual patients; other studies have traditionally lacked individual data regarding HAART use, and simply examined cancer trends during the HAART era [37]. We did not find a significant relationship between HAART use and NADCs. A recent study suggested that HAART use may be associated with a lower risk of NADCs, but did not demonstrate statistical significance [23]. The lack of a demonstrable association suggests that NADCs may develop independently of CD4 cell count and antiretroviral therapy, concurring with the rising rates of these cancers during the HAART era. Our data suggests that the aging of the HIV population is the primary factor associated with the rising rates of NADCs.
It is possible that our study did not find a potential protective effect of HAART on NADCs due to the fact that most of these events occurred among persons with relatively high CD4 cell counts who had not yet initiated HAART. Whether initiating HAART earlier than advocated by previous guidelines [38] would be beneficial for cancer prevention via mechanisms such as decreased immune activation or suppression of oncogenic viruses is unknown. Research examining the effect of earlier initiation of antiretroviral therapy on the incidence of AIDS-defining and non-AIDS-defining conditions, such as cancers, is under development.
The strengths of our study include the long-term follow-up of HIV patients as part our 23-year Natural History Study. Given recommendations for early HIV detection through routine screening [39], this study of patients with early-stage HIV infection provides important data on cancer rates and trends. In addition, our study population was racially diverse, without barriers to healthcare access, and from varied geographical locations in the United States. Moreover, our data included confirmed cancer diagnoses and individual patient data, including precisely defined use of antiretroviral therapy, AIDS-defining events, and serial CD4 cell counts.
Several potential limitations of this study should be noted. Our cohort consisted primarily of nondrug users with a low prevalence of hepatitis C; hence, data regarding the effect of drug use or hepatitis coinfection on NADC rates could not be evaluated. Furthermore, we did not collect data on behaviors such as tobacco use, alcohol use, or on family history of cancer. In addition, our population was primarily men; hence, women-specific tumors could not be adequately studied. Finally, given the limited number of NADCs and that most were skin cancers, examining individual NADCs, such as anal cancer or Hodgkin's disease, was not possible.
In conclusion, although the overall rate of cancer has declined since the HAART era, malignancies remain an important cause of morbidity among HIV-infected persons. The rates of ADCs have continued to fall since the advent of HAART, but the rates of NADCs are rising and now account for the majority of cancer cases. The increasing rates of NADCs appear most related to the aging of the HIV population. Antiretroviral therapy appears protective for the development of ADCs, but had no significant impact on NADCs.
Results
Between 1984 and 2006, 4498 participants were followed for a total of 33 486 person-years. The study cohort had a median age at HIV diagnosis of 28 (IQR 24-33) years; 91% were men. Race was reported as 45% African-American, 44% white/non-Hispanic, and 11% others (Table 1). HIV-seropositive date was prior to 1996 for 2443 (77%) of the participants. The median length of follow-up was 6.6 (IQR 3.7-10.1) years. Median baseline CD4 cell count at HIV diagnosis was 510 cells/μl (IQR 353-680). During the study period, 24% experienced an AIDS-defining event other than cancer (Table 1).
At least one cancer event was recorded for 446 individuals (10%). The first cancer was AIDS-defining for 311 (70%) individuals and non-AIDS-defining for 135 (30%). The median time from HIV diagnosis to an ADC was 5.6 (IQR 3.6-7.9) years and for a NADC was 6.0 (IQR 2.8-11.7) years. Figure 1 reports the specific types of cancers diagnosed. Of those whose first cancer was AIDS-defining, Kaposi's sarcoma was the most frequent (73%; n = 227); the most common NADC was skin cancer (47%; n = 63). Skin cancers were mostly basal carcinomas (n = 48), followed by melanoma (n = 10) and squamous (n = 5). Thirty-three persons (7.4% of those with cancer) developed two different cancers during the study period, as shown in Fig. 1. Eleven people had two different ADCs: 10 persons had Kaposi's sarcoma followed by NHL and one had NHL followed by Kaposi's sarcoma. Four people with an initial ADC subsequently developed a NADC (lung, anal, Hodgkin's, and skin cancer). Eight people with an initial NADC later developed an ADC (five had Kaposi's sarcoma and three had NHL), which was most commonly an initial skin cancer followed by Kaposi's sarcoma. Finally, 10 people developed two NADCs; most commonly, this was the development of two different types of skin cancer.
Of those who developed cancer, the diagnosis occurred in the pre-HAART era for 302 (68%) and in the post-HAART era for 144 (32%), resulting in pre-HAART and post-HAART cancer rates (per 1000 person-years) of 16.1 and 9.8, respectively. The rate of ADCs increased significantly between the early and late pre-HAART eras (7.6 and 14.2, respectively) and then declined significantly during the early and late post-HAART intervals (5.4 and 2.7, respectively) (Table 2). The rates for NADCs were stable in the pre-HAART era at approximately three cases per 1000 person-years. However, since the availability of HAART, NADC rates increased to 4.2 and 6.7 in the early and late post-HAART eras, respectively (P = 0.004). The rates of nonskin NADCs were less than two cases per 1000 person-years in the pre-HAART and early post-HAART eras, but increased to 4.1 per 1000 person-years during the late post-HAART era (P = 0.0003). Furthermore, the proportion of cancers that were NADCs significantly increased from 20% in the pre-HAART era to 36% in the early post-HAART era and 71% in the late post-HAART era (P < 0.0001).
The rates over time for the most common cancer types are shown in Table 2 and Fig. 2. Rates for both Kaposi's sarcoma and NHL increased significantly before HAART, but have steadily declined since 1996. The rate of anal cancer was stable in the pre-HAART era (0.1-0.2 cases per 1000 person-years), but significantly increased to 1.3 in the late post-HAART era (P = 0.001). Skin, renal, and prostate cancers also had the highest rates during the late post-HAART era.
The age-adjusted incidence among men in our HIV cohort was 13.0 per 1000 person-years for any cancer event (excluding basal cell and squamous), compared to 5.5 for men in the general population. Limiting the events to NADCs (excluding basal cell and squamous), the age-adjusted incidence among men in our cohort was 6.5 per 1000 person-years. The age-adjusted incidence among HIV-positive men (compared to the general population) was 4.0 (vs. 0.004) for Kaposi's sarcoma, 2.7 (vs. 0.2) for NHL, 1.6 (vs. 0.03) for Hodgkin's disease, and 0.2 (vs. 0.01) for anal carcinoma. For each type of event, the age-adjusted incidence for our overall cohort was similar to the age-adjusted incidence for men (data not shown).
The univariate proportional hazards regression models for any cancer, ADC, NADC, and nonskin NADC are shown in Table 3. For all outcomes, increased age was significantly associated with an increased risk of a cancer event. African-American race (compared with whites) was significantly associated with a decreased risk of an ADC or NADC event, but were not associated with a nonskin NADC. ADCs were also associated with male sex, occurrence of a noncancer AIDS diagnosis, lower CD4 cell counts, higher HIV viral loads and lack of HAART use. The association of ADCs and male sex was due to Kaposi's sarcoma (data not shown).
In the multivariate model, the predictors of any cancer and an ADC included male sex and a noncancer AIDS event, whereas factors associated with a reduced risk of a cancer event included African-American race, increased CD4 cell counts, and HAART (Table 4). HIV viral load was not included in the multivariate models because measurements were unavailable for 28% of the cohort.
NADCs were associated with increasing age and white race in the multivariate model; there was no association with sex, CD4 cell counts, HAART use, or prior noncancer AIDS events (Table 4). The relationship between white race and an elevated risk of NADCs was due to the high number of skin cancers; of the 63 cases of skin cancers, 93% occurred in the white race group. Models were repeated for nonskin NADCs (Tables 3 and 4). From the multivariate model for nonskin NADCs, age was still associated with cancer development.
Compared to those with ADCs at the time of cancer diagnosis, those with NADC at the time of diagnosis were more likely to be older (median 42 vs. 35 years, P < 0.0001) and white (68 vs. 52%, P = 0.006), have an HIV diagnosis date after 1996 (13 vs. 3%, P < 0.0001), have higher median CD4 cell counts (430 vs. 72 cells/μl, P < 0.0001) and lower median HIV viral loads (3.4 vs. 4.8 log, P < 0.0001), have received HAART for a greater percentage of their follow-up time (15 vs. 3%, P < 0.0001), and were less likely to have a prior AIDS-defining event (16 vs. 45%, P < 0.0001). The median CD4 cell counts at cancer occurrence for those with an ADC during the early pre-HAART, late pre-HAART, early post-HAART, and late post-HAART eras were 40, 80, 44, and 242 cells/μl, respectively (P = 0.14). For the same intervals, those with NADCs at cancer diagnosis had median CD4 cell counts of 410, 370, 361, and 474 cells/μl, respectively (P = 0.05).
During the study period, 1523 (34%) participants of the overall cohort died. Death occurred among 85% of patients with an ADC, 40% of those with an NADC, 46% with a nonskin NADC, and 30% without cancer (P < 0.0001). For patients with an ADC compared with those with a NADC, the hazard ratio for time from cancer diagnosis to death was 3.7 (95% CI 2.7-4.9, P < 0.0001). The estimated mortality at 1, 3, and 5 years after cancer diagnosis was 51, 76, and 84%, respectively, for those with an ADC, 15, 29, and 41% for those with a NADC, and 26, 38, and 49% for those with a nonskin NADC.
Methods
We examined data collected from the TACC NHS, a multicenter, prospective, observational study, which enrolled 4566 HIV-positive persons from 1984 to 2006 at seven geographic locations in the United States. Participants were military beneficiaries (active duty members, retirees, and dependents) evaluated on a biannual basis utilizing standardized data collection procedures, and all have free access to care, including medications. Our study was approved by central and local Institutional Review Boards, and patients provided informed consent.
All active duty US military personnel undergo routine HIV screening every 1-5 years; prior to enlistment, all members are confirmed HIV negative. For this study, among those with a last known HIV-negative date (58%), the median time from last HIV-negative date to first HIV-positive date [enzyme-linked immunosorbent assay (ELISA) confirmed by a western blot test] was 16 months. Baseline was defined as the time of HIV seroconversion, conservatively estimated as 6 months prior to the first documented HIV-positive test. Cancer cases were only included if they occurred after this point. The diagnosis of cancer was based on physician diagnosis supported by laboratory, radiologic, and/or histopathologic results. Cancer events were identified in our database by searching for specific cancer codes and for the terms 'cancer', 'malignancy', 'tumor', or 'neoplasm'. Cancers were defined as ADCs (Kaposi's sarcoma, NHL, or ICC) or NADCs (all others). Participants excluded from this analysis were those without a documented HIV-positive test (n = 10), with cancer prior to HIV seroconversion (n = 51), and with a benign tumor or a clinical diagnosis that could not be confirmed as malignant (n = 7), yielding 4498 participants for our study (Fig. 1).
Data collected included demographics, including self-reported race/ethnicity, CD4 cell counts and HIV viral load tests at baseline and sequentially over time, history of AIDS-defining conditions other than cancer, and antiretroviral therapy. HAART was defined as two or more nucleoside reverse transcriptase inhibitors (NRTIs) in combination with at least one protease inhibitor or one nonnucleoside reverse transcriptase inhibitor (NNRTI), one NRTI in combination with at least one protease inhibitor and at least one NNRTI, or an abacavir-containing or tenofovir-containing regimen of three or more NRTIs. For those with cancer, the event date was determined by the first cancer diagnosis date for the specific type of cancer considered. For those without cancer, the censoring date was the last study visit or the date of death. Follow-up for this report ended 31 December 2006.
Statistical analyses included descriptive statistics to compare those with and without cancer events. Medians are presented with interquartile ranges (IQRs). Kruskal-Wallis tests were used to compare medians, and chi-squared tests were used to compare proportions. The number of events, person-years at risk, and rates of events (per 1000 person-years of follow-up) were calculated for the overall study period and for specific time intervals: the early pre-HAART era (1984-1990), the late pre-HAART era (1991-1995), the early post-HAART era (1996-2000), or the late post-HAART era (2001-2006). Each participant contributed to the person-years at risk for every time interval from HIV diagnosis until the event or censoring time. Poisson regression analyses were used to test the hypothesis that the cancer rates remained constant over those intervals. In order to compare the risk of cancer in our HIV-positive cohort with the risk seen in the general population, age-adjusted incidence (per 1000 person-years) over the study period (1984-2006) was calculated for any cancer event (excluding basal cell and squamous) and then separately for a NADC event (excluding basal cell and squamous), Kaposi's sarcoma, NHL, Hodgkin's disease, and anal carcinoma. For each event, the incidence was age-adjusted to the US 2000 standard population [15] and calculated for the overall cohort and for males only. The age-adjusted incidence was compared to data provided in the National Cancer Institute Surveillance Epidemiology and End Results (SEER) Cancer Statistics Review [16] for the period 1974-2004; for Kaposi's sarcoma, SEER results from 1975 to 1979 were used, as most cases of Kaposi's sarcoma occurred among HIV patients after 1980. Basal cell and squamous were excluded because SEER does not collect data on those events.
Participants were classified by HIV diagnosis era: pre-HAART (prior to 1996) or post-HAART (at or after 1996). Univariate and multivariate Cox proportional hazard models, stratified by HIV diagnosis eras, were used to evaluate the association of specific factors with cancer. The multivariate models were adjusted for demographics at the time of HIV diagnosis (age, sex, ethnicity, and year of HIV diagnosis). Variables that may change during follow-up (CD4 cell count, HIV RNA levels, HAART use, and noncancer AIDS event) were considered as time-updated covariates, using all available data between HIV diagnosis time and the event or censoring time. For those with cancer, time from cancer diagnosis to death was evaluated with unadjusted proportional hazards models and Kaplan-Meier survival estimates. Hazard ratios are reported with 95% confidence interval (CI). All analyses were conducted using SAS (version 8.2, SAS Institute, Cary, North Carolina, USA).
|
|
| |
| |
|
|
|
