Secondary Logo

Journal Logo

Epidemiology and Prevention

Comorbidity Acquired Before HIV Diagnosis and Mortality in Persons Infected and Uninfected With HIV: A Danish Population-Based Cohort Study

Lohse, Nicolai MD, PhD*; Gerstoft, Jan MD, DMSc; Kronborg, Gitte MD, DMSc; Larsen, Carsten Schade MD, DMSc§; Pedersen, Court MD, DMSc; Pedersen, Gitte MD, PhD; Nielsen, Lars MD, PhD#; Sørensen, Henrik Toft MD, PhD, DMSc*; Obel, Niels MD, DMSc

Author Information
JAIDS Journal of Acquired Immune Deficiency Syndromes: August 1st, 2011 - Volume 57 - Issue 4 - p 334-339
doi: 10.1097/QAI.0b013e31821d34ed



Mortality and morbidity among persons infected with HIV has decreased dramatically since the introduction of highly active antiretroviral therapy in 1996-1997.1 Despite this progress, mortality remains markedly higher than in uninfected persons.2 The proportion of all deaths among persons infected with HIV due to non-AIDS and noninfectious causes such as drug-induced toxicity,3 cardiovascular disease,4 renal disease,5 cancer,6 and HIV-induced morbidity other than AIDS7 has increased to as much as 46%.2,8-12 Several recent studies have focused on coexisting non-AIDS and noninfectious morbidity as prognostic factors in these patients,6,8-15 but the impact of diseases acquired by patients before their HIV diagnosis remains poorly understood, with the exception of hepatitis C virus (HCV) infection.16,17 We aimed to estimate the impact of comorbidity acquired before HIV diagnosis on mortality in individuals infected with HIV.


Study Type

This matched cohort study investigates the impact of non-HIV comorbidity on all-cause mortality in a cohort of persons infected with HIV compared with a cohort of persons from the general population.

Study Population

Our study population consisted of all persons who were diagnosed with HIV in Denmark between January 1, 1997, and May 1, 2005; who were at least 16 years old; and who were Danish residents at the time of HIV diagnosis. These inclusion dates allowed for at least 2 years of observation time for all subjects. For each patient with HIV, we sampled general population controls, matched on sex and year and month of birth, who were alive and living in the same municipality as the patient on the patient's diagnosis date. Denmark has a population of approximately 5.3 million persons living in 270 municipalities. We aimed to sample 99 controls for each individual infected with HIV, but because of a shortage of eligible controls in some municipalities, the mean number of controls per patient in the final general population cohort was 95.5.

Data Sources

Persons with HIV were identified from the Danish HIV Cohort Study (DHCS), which has established an open population-based cohort that includes all persons infected with HIV seen in 8 HIV clinics in Denmark since January 1, 1995.18,19 Denmark provides free, tax-supported health care including treatment for HIV. Antiretroviral drugs and specialist treatment are available exclusively through the HIV clinics, and persons diagnosed with HIV are referred directly to one of these clinics for care and treatment. It is estimated that more than 99% of patients accept this offer, so the cohort is virtually complete regarding incident cases of HIV in Denmark. The annually updated data of DHCS include antiretroviral treatment, development of AIDS-defining illnesses, CD4+ cell counts, HIV RNA [viral load (VL)], and deaths, in addition to sex, age, ethnicity, most likely mode of infection, HCV coinfection at time of diagnosis, and preexisting AIDS-defining illness.18 General population control subjects were sampled from the Danish Civil Registration System (DCRS), which has maintained information on all Danish residents since 1968, using unique personal identifiers (CPR numbers).20 The DCRS records date of birth, sex, place of residence, date of migration, and date of death.20 Its database is updated within a week of a person's birth, address change, death, or emigration. Comorbidity before an HIV diagnosis was assessed both for persons with HIV and for general population controls through the Danish National Patient Registry (DNPR). This registry, established in 1977, covers all Danish hospitals and records all hospital admissions, diagnoses, and, since 1995, outpatient and emergency visits.21 The DNPR covers private and public hospitals. We used CPR numbers to link data between these registries.

Study Variables

Outcome Variable

Death from any cause was obtained through the DCRS.

Main Explanatory Variables

We computed a Charlson comorbidity index (CCI) score for all study subjects based on their complete hospital history in the DNPR. The CCI22 was developed to classify comorbid conditions that alter the risk for 1-year mortality after hospitalization in longitudinal studies. The CCI has been adapted and validated for use with hospital discharge data in International Classification of Diseases databases for the prediction of short- and long-term mortality.23 The following 19 disease categories are included in the CCI: myocardial infarction (1 point), congestive heart failure (1 point), peripheral vascular disease (1 point), cerebrovascular disease (1 point), dementia (1 point), chronic pulmonary disease (1 point), connective tissue disease (1 point), ulcer disease (1 point), mild liver disease (1 point), diabetes (1 point), hemiplegia (2 points), moderate to severe renal disease (2 points), diabetes with end-organ failure (2 points), any tumor (2 points), leukemia (2 points), lymphoma (2 points), moderate to severe liver disease (3 points), metastatic solid tumor (6 points), and AIDS (6 points). A CCI score was calculated for each patient infected with HIV as of the date of HIV diagnosis. For each patient's general population control, we computed a CCI score as of the same date. The disease category “AIDS” was excluded in our computations because it was closely related to the main exposure in our study.

HCV infection is not included in the CCI score but is common in individuals infected with HIV due to shared routes of transmission. We thus included it as a study variable, assuming that HCV infection occurred at the same time or before HIV acquisition. Patients with at least 1 positive HCV antibody test or a positive HCV RNA test were considered HCV-positive; others were considered HCV-negative. HCV status was available for 98.0% of study patients infected with HIV. We did not have individual data on HCV infection among the general population controls mainly because HCV infection is often asymptomatic and therefore not diagnosed; also, it was not a separate diagnosis in the International Classification of Diseases, 8th Revision, which was used in the DNPR until 1993. As the estimated prevalence of HCV in Denmark is just 3 per 1000,24 all individuals in the general population cohort were considered HCV-negative in our analyses.

Other Explanatory Variables

Information on the sex (male/female), age (in years), ethnicity (white yes/no), mode of infection (injecting drug use yes/no, men who have sex with men yes/no, heterosexual yes/no, blood transfusion yes/no, hemophilia yes/no, other yes/no, unknown yes/no), CD4+ T-cell count at time of diagnosis (median, interquartile range), VL at time of diagnosis (median, interquartile range), and HCV serostatus of the subjects was obtained from the DHCS.

Statistical Analyses

Basal Characteristics

Chi-square test was used to examine differences in distribution of characteristics between persons with HIV and the general population.

Survival Analysis

We linked all study subjects to the DCRS and computed observation times from their date of HIV diagnosis until death from any cause, migration, or June 1, 2007, whichever came first. We used Cox regression analysis to compute unadjusted hazard ratios as a measure of mortality rate ratios (MRRs) for patients with HIV diagnosed after January 1, 1997, comparing different covariate strata (model A). In addition to CCI (0 points vs ≥1 points), the following covariates were included in the analysis: CD4+ T-cell count at time of diagnosis, defined as the first measurement within 90 days after diagnosis (>200 cells/μL, 50-200 cells/μL, <50 cells/μL, or missing); age at time of diagnosis; history of injecting drug use; white race; sex; and log(VL) at time of diagnosis, defined as the first measurement within 90 days after diagnosis (<4 copies/mL, 4-5 copies/mL, >5 copies/mL, or missing). We estimated the MRRs in a multivariate model including all covariates (model B), and in adjusted models including CCI and one of each of the other covariates successively (models C and D). Model C examined the influence of each covariate on CCI score. Model D examined the influence of CCI score on each of the other covariates. To assess the potential bias introduced by immigrants with previous hospital admissions not registered in the DNPR, we carried out a sensitivity analysis restricting the population to individuals infected with HIV born in Denmark. Another sensitivity analysis excluded lymphoma from the CCI score because some lymphomas might have been AIDS-defining events.

Interaction Between HIV and Comorbidity Effects

Mortality rates (MRs) were computed within 4 CCI strata (0, 1, 2, and >3 points), separately for patients and for general population controls. Kaplan-Meier curves were constructed, and log-rank test was used to compare survival within each stratum. For controls, observation time was computed starting from the date of HIV diagnosis of the corresponding HIV-infected person. To estimate the interaction between baseline comorbidity and HIV effects, defined as deviation of MRs from an additive model,25 we assessed the effect measure modification on MR differences. The interaction between the prognostic effects of HIV and comorbidity was computed as the interaction risk (IAR)25 with a bootstrap 95% confidence interval (CI)26: IARhiv,x = (MRhiv,x − MRhiv,0) − (MRpop,x − MRpop,0), where IARhiv,x is the IAR difference for persons with HIV and CCI = X; MRhiv,x and MRhiv,0 are the MRs for patients with HIV with CCI = X (X can take the values 1, 2, or 3) and CCI = 0, respectively; and MRpop,x and MRpop,0 are the MRs for population controls with CCI = X and CCI = 0, respectively. The proportion of excess deaths in persons infected with HIV due to interaction within each comorbidity stratum CCI = X was IARhiv,x/(MRhiv,x − MRhiv,0) = 1 − (MRpop,x − MRpop,0)/(MRhiv,x − MRhiv,0). A sensitivity analysis restricted the population to persons not coinfected with HCV.

Population Attributable Risk

The population attributable risk (PAR) of death due to CCI points in the population infected with HIV was calculated as follows27: If MRhiv is the MR in the population with HIV and MRhiv,0 is the MR in the unexposed population (no CCI points), then PAR is (MRhiv - MRhiv,0)/MRhiv.


Among the 1638 patients with HIV diagnosed from January 1, 1997, to May 1, 2005, we observed 195 deaths in 9350 person-years of follow-up, yielding an MR of 2.09% (95% CI: 1.81 to 2.40) per year. In comparison, the 3676 deaths observed among 156,506 general population control cohort members followed for 939,560 person-years yielded an MR of 0.39% (95% CI: 0.38 to 0.40) per year. Morbidity was present before HIV diagnosis in 354 (21.6%) of 1638 patients with HIV, of whom 185 (11.3%) had acquired at least 1 CCI point and 211 (12.9%) were coinfected with HCV; morbidity was present in 12,531 (8.0%) of 156,506 general population controls (Table 1). The most common CCI comorbidities were liver disease in 48 (2.9%) patients vs 793 (0.51%) general population controls, lymphoma in 11 (0.67%) patients vs 232 (0.15%) controls, and solid malignant tumors in 22 (1.34%) patients vs 1773 (1.13%) controls. Among 1638 patients with HIV, 155 (9.5%) did not have available CD4+ T-cell count data and 187 (11.4%) did not have available VL data at the time of diagnosis.

Characteristics of the Study Populations

Comorbidity as a Predictor of All-Cause Mortality in Patients With HIV

The MRR for patients with HIV with at least 1 CCI point compared with patients with HIV with no CCI points was 3.27 (95% CI: 2.38 to 4.50; Table 2, model A). This estimate decreased after adjustment for other prognostic factors for death (Table 2, model C), but the CCI score remained a predictor of death in the fully adjusted model (MRR = 1.84, 95% CI: 1.32 to 2.57; Table 2, Model B). CCI score did not substantially affect the MRR of other risk factors known to be associated with death in patients infected with HIV (Table 2, model D). In a sensitivity analysis in which lymphoma diagnoses—some of which are AIDS-defining—were excluded from the CCI score, we obtained similar MRR estimates (1.90, 95% CI: 1.36 to 2.66). In an analysis restricted to patients with HIV of Danish origin (69% of all patients), the effect of CCI score in the fully adjusted model decreased slightly to 1.74 (95% CI: 1.22 to 2.48).

Cox Regression of Time to Death for Patients Diagnosed After January 1, 1997

Interaction Between HIV and Comorbidity Effects on All-Cause Mortality

MRs for persons with HIV compared with the general population within CCI strata were 1.70 per 100 person-years at risk (95% CI: 1.44 to 2.00) vs 0.27 per 100 person-years at risk (95% CI: 0.26 to 0.28) for CCI = 0; 4.37 (3.01 to 6.32) vs 1.36 (1.26 to 1.47) for CCI = 1; 8.06 (4.94 to 13.16) vs 2.44 (2.22 to 2.68) for CCI = 2; and 10.15 (5.08 to 20.30) vs 5.84 (5.19 to 6.58) for CCI = 3+. Survival curves are depicted in Figure 1. There was a high IAR on rate differences in all CCI strata, indicating sizeable interaction between HIV and comorbidity effects (ie, the joint effects of HIV and comorbidity exceeded the sum of their individual effects on mortality). Compared with patients with no CCI points, 58.8% of the excess deaths in patients with a CCI score of 1, 66.0% of the excess deaths in patients with a CCI score of 2, and 34.1% of the excess deaths in patients with a CCI score of 3 or higher resulted from this interaction (Table 3). Restricting the analyses to the subgroup not coinfected with HCV produced similar IAR estimates in the 3 CCI point strata.

Kaplan-Meier survival curves and log-rank tests comparing persons with HIV with the general population within 4 CCI point strata.
Interaction Between HIV and Comorbidity Effects

Mortality Due to Conditions Existing Before HIV Diagnosis

The PAR of death due to comorbidity acquired before HIV diagnosis was 0.19, that due to HCV coinfection was 0.16, and that due to baseline comorbidity and HCV coinfection combined was 0.32. Thus, 32% of mortality in the Danish population with HIV could be attributed to comorbidity acquired before HIV diagnosis or to HCV coinfection. This is illustrated in Figure 2 as the relative difference between the mortality for all patients (MR = 2.09% per year, 95% CI: 1.81 to 2.40) and that for HCV-negative patients with no CCI points (MR = 1.43% per year, 95% CI: 1.18 to 1.73). Considering that the mortality in the general population with no CCI points (MR = 0.28% per year, 95% CI: 0.27 to 0.30) comprised 20% of the latter group, an estimated 45% [(2.09 − 1.43 + 0.28)/2.09] of total mortality in the population infected with HIV could not be ascribed to HIV itself, nor to toxicity from antiretroviral drugs.

Visual presentation of the total mortality in persons with HIV and the relative contribution of background mortality (0.28/2.09 = 13.6%), HCV coinfection and CCI (2.09 − 1.43/2.09 = 31.5%), and HIV (1.43 − 0.28/2.09 = 55%).


In this population-based study, we found that morbidity acquired before HIV diagnosis was an independent risk factor for death. Almost half of mortality in persons diagnosed with HIV in a health care setting with free access to highly active antiretroviral therapy stemmed from factors unrelated to the HIV disease or associated factors such as toxicity from antiretroviral drugs. Furthermore, comorbidity acquired before HIV diagnosis acted synergistically with HIV as a risk factor for death.

Our study had a number of strengths. First, the population-based setting of DHCS allowed us to minimize selection bias by including all patients infected with HIV residing in Denmark at the time of diagnosis. Second, the DCRS with its complete follow-up allowed us to generate a matched general population comparison cohort and to collect accurate information on time of death and emigration, which further minimized selection bias. Third, the DNPR provided access to information on all diseases treated in a hospital going back almost 20 years before the HIV diagnosis. Thus, information on morbidity before HIV diagnosis was prospectively collected and not influenced by recall bias. Finally, a unique personal identifier allowed us to link all registries.

Our study also had limitations. Persons may be infected with HIV for years before they are diagnosed; thus, a disease registered before HIV diagnosis may actually have been acquired after HIV infection and could be a consequence of HIV rather than a risk factor acquired before HIV. However, we believe this was a minor source of bias. Only diagnoses obtained before the hospital visit at which HIV was diagnosed were used in computing the CCI score, and no points were counted for HIV and/or AIDS diagnosis. The only AIDS-defining illness that may have contributed to the CCI score was lymphoma, and a sensitivity analysis excluding CCI points for lymphoma did not change any estimates or conclusions. Underestimation of comorbidity might have happened because of lack of information on diagnoses made in the primary care sector or because of the presence of mild diseases that would go undiagnosed. Relative overestimation of comorbidity might have happened in some high-risk groups later diagnosed with HIV, for example, injecting drug users or men who have sex with men. These groups might be in more frequent contact with the health care system, which would increase the chances of any disease being diagnosed and lead to information bias. In both cases, restricting diagnoses to those causing hospitalization have most likely reduced bias and strengthened our findings rather than weakening them. Another concern is that discharge diagnoses from hospital visits outside Denmark were not captured, so CCI scores could be underestimated in immigrants. However, a sensitivity analysis excluding patients of non-Danish origin only slightly changed the risk estimates for CCI score.

Few studies have looked at deaths due to conditions present before HIV diagnosis.14,28 Delpierre et al14 found a 3.75 times higher risk for death in persons unemployed at the time of HIV diagnosis. Assuming that comorbidities are associated with unemployment, these results support our findings. HCV coinfection affects mortality both directly17 and indirectly28 through family-related risk factors. It is normally diagnosed during the initial screening of patients newly diagnosed with HIV and most often assumed to be acquired at the same time as or before HIV through an identical route of transmission. No previous studies have quantified the proportion of mortality attributable to comorbidity acquired before HIV diagnosis.

Part of the excess mortality in persons with comorbidity acquired before HIV diagnosis may be due directly to the comorbidity itself. As we have shown previously for HIV/HCV-coinfected patients, the excess mortality may also be due to lifestyle and other family-related risk factors28 such as tobacco use, alcohol intake, drug abuse, and risk behaviors. Finally, the interaction between baseline comorbidity and HIV contributed considerably to the excess mortality. This could be purely biological but could also stem from suboptimal treatment occurring when physicians and patients face the complexity of HIV in combination with another serious disease. HCV coinfection did not seem to underlie the observed synergistic effects.

Depending on the distribution of age and mode of infection among persons infected with HIV, we would expect similar findings in other high-income settings. Illness and death in an HIV-infected population with access to up-to-date antiretroviral therapy can be lowered through earlier diagnosis, increased adherence, and optimization of antiretroviral regimens. Recent studies suggest, however, that relying solely on these measures may not suffice as patients get older and acquire age-related and possibly also drug toxicity-related diseases.2,3,6,29-32 Prevention of causes of non-AIDS noninfectious death is important but can decrease mortality only to a certain extent. The considerable burden conferred by diseases acquired before HIV diagnosis, found in more than 1 in 5 patients in this study, calls for a comprehensive approach to treatment and care. Involvement of a team of medical specialists is clearly needed. Still, additional mortality reduction may not always be possible.

In conclusion, persons infected with HIV in Denmark have a considerable burden of disease acquired before HIV diagnosis. HIV and its treatment seem to account directly for little more than half of the deaths in this population. In addition, HIV infection seems to increase the impact of comorbidities diagnosed before HIV diagnosis on risk for death. Further studies aiming to identify biological and sociocultural risk factors for comorbidity are required to increase our understanding of the complex interaction between HIV and diseases acquired before HIV.


1. Sterne JA, Hernan MA, Ledergerber B, et al. Long-term effectiveness of potent antiretroviral therapy in preventing AIDS and death: a prospective cohort study. Lancet. 2005;366:378-384.
2. Lohse N, Hansen AB, Pedersen G, et al. Survival of persons with and without HIV infection in Denmark, 1995-2005. Ann Intern Med. 2007;146:87-95.
3. Friis-Moller N, Reiss P, Sabin CA, et al. Class of antiretroviral drugs and the risk of myocardial infarction. N Engl J Med. 2007;356:1723-1735.
4. Obel N, Thomsen HF, Kronborg G, et al. Ischemic heart disease in HIV-infected and HIV-uninfected individuals: a population-based cohort study. Clin Infect Dis. 2007;44:1625-1631.
5. Wyatt CM, Rosenstiel PE, Klotman PE. HIV-associated nephropathy. Contrib Nephrol. 2008;159:151-161.
6. Pantanowitz L, Schlecht HP, Dezube BJ. The growing problem of non-AIDS-defining malignancies in HIV. Curr Opin Oncol. 2006;18:469-478.
7. Phillips A. Morbidity and mortality in the HAART era. In: 15th Conference on Retroviruses and Opportunistic Infections; 2008. Abstract 8.
8. Lucas SB, Curtis H, Johnson MA. National review of deaths among HIV-infected adults. Clin Med. 2008;8:250-252.
9. Panos G, Samonis G, Alexiou VG, et al. Mortality and morbidity of HIV infected patients receiving HAART: a cohort study. Curr HIV Res. 2008;6:257-260.
10. Novoa AM, de Olalla PG, Clos R, et al. Increase in the non-HIV-related deaths among AIDS cases in the HAART era. Curr HIV Res. 2008;6:77-81.
11. Martinez E, Milinkovic A, Buira E, et al. Incidence and causes of death in HIV-infected persons receiving highly active antiretroviral therapy compared with estimates for the general population of similar age and from the same geographical area. HIV Med. 2007;8:251-258.
12. Braithwaite RS, Justice AC, Chang CC, et al. Estimating the proportion of patients infected with HIV who will die of comorbid diseases. Am J Med. 2005;118:890-898.
13. Hansen AB, Gerstoft J, Kronborg G, et al. Mortality in siblings of patients coinfected with HIV and hepatitis C virus. J Infect Dis. 2007;195:230-235.
14. Delpierre C, Cuzin L, Lauwers-Cances V, et al. Unemployment as a risk factor for AIDS and death for HIV-infected patients in the era of highly active antiretroviral therapy. Sex Transm Infect. 2008;84:183-186.
15. Tai JH, Udoji MA, Barkanic G, et al. Pregnancy and HIV disease progression during the era of highly active antiretroviral therapy. J Infect Dis. 2007;196:1044-1052.
16. Matthews GV, Dore GJ. HIV and hepatitis C coinfection. J Gastroenterol Hepatol. 2008;23:1000-1008.
17. Weis N, Lindhardt BO, Kronborg G, et al. Impact of hepatitis C virus coinfection on response to highly active antiretroviral therapy and outcome in HIV-infected individuals: a nationwide cohort study. Clin Infect Dis. 2006;42:1481-1487.
18. Obel N, Engsig FN, Rasmussen LD, et al. Cohort profile: the Danish HIV cohort study. Int J Epidemiol. 2009;38:1202-1206.
19. Lohse N, Hansen AB, Jensen-Fangel S, et al. Demographics of HIV-1 infection in Denmark: results from the Danish HIV Cohort Study. Scand J Infect Dis. 2005;37:338-343.
20. Pedersen CB, Gøtzsche H, Møller JØ, et al. The Danish Civil Registration System. A cohort of eight million persons. Dan Med Bull. 2006;53:441-449.
21. Andersen TF, Madsen M, Jorgensen J, et al. The Danish National Hospital Register. A valuable source of data for modern health sciences. Dan Med Bull. 1999;46:263-268.
22. Charlson ME, Pompei P, Ales KL, et al. A new method of classifying prognostic comorbidity in longitudinal studies: development and validation. J Chronic Dis. 1987;40:373-383.
23. de Groot V, Beckerman H, Lankhorst GJ, et al. How to measure comorbidity. A critical review of available methods. J Clin Epidemiol. 2003;56:221-229.
24. Christensen PB. Epidemiology of hepatitis C [in Danish]. Ugeskr Laeger. 1998;160:3529-3532.
25. Rothman KJ. Epidemiology: An Introduction. New York, NY: Oxford University Press; 2002.
26. Assmann SF, Hosmer DW, Lemeshow S, et al. Confidence intervals for measures of interaction. Epidemiology. 1996;7:286-290.
27. Last JM. A Dictionary of Epidemiology. New York, NY: Oxford University Press; 2000.
28. Hansen AB, Lohse N, Gerstoft J, et al. Cause-specific excess mortality in siblings of patients co-infected with HIV and hepatitis C virus. PLoS One. 2007;2:e738.
29. Lohse N, Hansen AB, Gerstoft J, et al. Improved survival in HIV-infected persons: consequences and perspectives. J Antimicrob Chemother. 2007;60:461-463.
30. Bhavan KP, Kampalath VN, Overton ET. The aging of the HIV epidemic. Curr HIV /AIDS Rep. 2008;5:150-158.
31. Goulet JL, Fultz SL, Rimland D, et al. Aging and infectious diseases: do patterns of comorbidity vary by HIV status, age, and HIV severity? Clin Infect Dis. 2007;45:1593-1601.
32. Wierzbicki AS, Purdon SD, Hardman TC, et al. HIV lipodystrophy and its metabolic consequences: implications for clinical practice. Curr Med Res Opin. 2008;24:609-624.

HIV; comorbidity; mortality; cohort study; population based

© 2011 Lippincott Williams & Wilkins, Inc.