Introduction Whether HIV influences pulmonary function remains controversial. We assessed dynamic pulmonary function in people living with HIV (PLWHIV) and uninfected controls.
Methods A total of 1098 PLWHIV from the Copenhagen Co-morbidity in HIV infection study and 12 161 age-matched and sex-matched controls from the Copenhagen General Population Study were included. Lung function was assessed using FEV1 and FVC, while airflow limitation was defined by the lower limit of normal (LLN) of FEV1/FVC and by FEV1/FVC<0.7 with FEV1predicted <80% (fixed). Logistic and linear regression models were used to determine the association between HIV and pulmonary function adjusting for potential confounders (including smoking and socioeconomic status).
Results In predominantly white men with mean (SD) age of 50.6 (11.1) the prevalence of airflow limitation (LLN) was 10.6% (95% CI 8.9% to 12.6%) in PLWHIV and 10.6% (95% CI 10.0 to 11.1) in uninfected controls. The multivariable adjusted OR for airflow limitation defined by LLN for HIV was 0.97 (0.77–1.21, P<0.78) and 1.71 (1.34–2.16, P<0.0001) when defined by the fixed criteria. We found no evidence of interaction between HIV and cumulative smoking in these models (P interaction: 0.25 and 0.17 for LLN and fixed criteria, respectively). HIV was independently associated with 197 mL (152–242, P<0.0001) lower FEV1 and 395 mL (344–447, P<0.0001) lower FVC, and 100 cells/mm3 lower CD4 nadir was associated with 30 mL (7–52, P<0.01) lower FEV1 and 51 mL (24–78, P<0.001) lower FVC.
Conclusion HIV is a risk factor for concurrently decreased FEV1 and FVC. This excess risk is not explained by smoking or socioeconomic status and may be mediated by prior immunodeficiency.
Trial registration number NCT02382822.
- COPD epidemiology
Statistics from Altmetric.com
What is the key question?
Is pulmonary function lower in people with HIV compared with epidemiologically well-matched uninfected controls?
What is the bottom line?
HIV is independently associated with lower FEV1 and FVC after adjustment for early-life and late-life risk factors.
Why read on?
This study represents the largest HIV spirometry study and provides detailed information on respiratory risk factors and self-reported morbidity.
Morbidity and mortality due to non-AIDS comorbidity have become a major focus in clinical care of people living with HIV (PLWHIV). Cardiovascular disease, bone, metabolic and renal diseases have received much attention, and specific recommendations for prevention are available for these conditions.1 COPD is a world-leading cause of life years lost in the general population,2 and is associated with modifiable risk factors.3 Despite this, COPD has received less attention in clinical management of PLWHIV.
The prevalence of COPD in PLWHIV is likely high with estimates from the combination antiretroviral therapy (cART) era exceeding 15% (summarised in online supplementary figure S1).4–9 This may be reflected by the increased utilisation of tobacco smoking among PLWHIV, but HIV-associated drivers, such as chronic inflammation and immune activation, may also contribute to the high burden of COPD.10 Few studies have used spirometry and included non-HIV controls making it difficult to assess the independent effect of HIV on pulmonary function.4 7 11 12 In addition, most studies have been small and have not appropriately adjusted for confounders. Some smaller studies using non-conventional pulmonary function tests and CT have found evidence that HIV is a risk factor for airway and alveolar abnormalities.13 14
Supplementary file 1
In this cross-sectional study we evaluate the prevalence of airflow limitation and associated risk factors in a large cohort of PLWHIV from the Copenhagen Co-morbidity in HIV infection (COCOMO) study and matched uninfected controls from the Copenhagen General Population Study (CGPS).15 Uniform data collection in these cohorts and detailed information on risk factors enabled us to more accurately explore the association between HIV and pulmonary function.
Study design, study subjects and ethics
The COCOMO study has been described elsewhere.15 In brief, the COCOMO study (NCT02382822) is a prospective study evaluating prevalence, incidence and pathogenesis of non-AIDS comorbidity in PLWHIV in Copenhagen, Denmark. Data collection for PLWHIV was performed at the Department of Infectious Diseases, Rigshospitalet, Copenhagen, and the Department of Infectious Diseases, Hvidovre Hospital, Copenhagen, between March 2015 and November 2016, and includes >40% of PLWHIV in the capital region of Copenhagen. Matched uninfected controls were recruited between January 2013 and December 2016 at Herlev Hospital, Copenhagen. Controls were not HIV tested but the expected prevalence of HIV among the general population in Denmark is small (~0.1%). The CGPS was initiated in 2003 and follows >100 000 volunteers from the greater Copenhagen area. Of all inhabitants in this area, 25% of the 20–40 years old and 100% of the >40 years old were invited.
The procedure and calibration of spirometry have previously been described.15 Briefly, an EasyOne ultrasonic spirometer (ndd Medical, Zürich, Switzerland) was used in accordance with American Thoracic Society/European Respiratory Society guidelines. Predicted values, z-scores and lower limit of normal (LLN) were calculated for prebronchodilator volumes using multiethnic prediction equations and R macros provided by the Global Lung Function Initiative.16 Instructions for spirometry were graded by the examiner as ‘easy’, ‘difficult’ or ‘very difficult’ which was used for sensitivity analysis.
Detailed information related to respiratory health was obtained through self-report using identical questionnaires in both cohorts and included tobacco smoking history, environmental tobacco smoke exposure, educational status and information about early-life and late-life events. Self-reported respiratory morbidity included questions about breathlessness (ie, modified British Medical Research Council (mMRC), cough (for the past 2 months), sputum production (over 3 months within a year) and wheezing (intermittently)). Self-reported COPD, self-reported asthma and use of inhaled medications were recorded as well. HIV-related variables were obtained from patient records. Biochemical analyses were measured uniformly in all participants using standardised hospital assays at Herlev Hospital, Copenhagen.17
Study endpoints and selection of covariates
The primary endpoint of the study was airflow limitation defined according to the LLN, that is, the fifth percentile of the reference population (equivalent to −1.64 z-scores).16 As secondary endpoint we assessed FEV1/FVC<0.7 defined according to the Global Initiative for Chronic Obstructive Lung Disease (GOLD) with FEV1predicted below 80% (grade 2 or higher COPD3) which has previously been used in the general population and in PLWHIV.18 19 Severity of airflow limitation was also graded according to GOLD: ≥80 (mild), 50–79 (moderate), 30–49 (severe), <30% FEV1predicted (very severe).3 We also assessed absolute FEV1 and FVC and their z-scores. A priori selection of confounders was based on clinical assumptions about the relationship between HIV and airflow limitation using directed acyclic graphs.
Uninfected controls were frequency matched with PLWHIV by gender and five age year strata. We randomly identified 14 unique controls for every person with HIV in each age and sex stratum. For men aged 30–35 it was only possible to identify three controls in each 5-year age interval, due to differences in the age and sex distribution between the HIV and the source population for the controls (CGPS) (online supplementary figure S2). Differences in clinical characteristics and spirometry outcomes between those with HIV and matched controls were assessed using t-tests and Mann-Whitney comparisons for continuous data, and Χ2 tests were used for categorical data. For pulmonary function analyses we omitted tests with FEV1 and FVC below 500 and 400 mL, respectively. 95% binomial CIs were computed for the prevalence of airflow limitation among PLWHIV and uninfected controls using the Wilson method. Logistic regression analyses were performed to determine risk factors associated with airflow limitation (LLN and FEV1/FVC<0.7 with FEV1predicted <80%). Univariable and multivariable ORs with 95% CIs were computed for these analyses. The association between HIV status and FEV1 and FVC as continuous outcomes was assessed using linear regression. For all analyses, we considered a simple demographic model 1 (adjusted for age, sex, ancestry, pack-years of smoking) and model 2 (which included additional confounders, ie, preterm delivery,20 passive smoking exposure during childhood,21 educational status22 and previous pneumonia23). An interaction term between HIV and pack-years was included to determine whether HIV modifies the effect of smoking on airflow limitation. Self-reported symptoms were compared between those with HIV and matched controls and stratified by smoking status.
Missing values were rare (smoking status was missing for 1.7%) and analyses were based on a complete case scenario. A few variables (ie, educational status, self-reported COPD, e-cigarette use and marijuana use, and passive smoking during childhood) were not introduced during the first half of recruitment (n=5658) for the CGPS, which reduced the df for model 2 compared with model 1, in which all covariates were available.
As a subanalysis, we analysed PLWHIV separately to determine whether HIV-associated variables are associated with spirometric indices (current and nadir CD4+ T cell count, viral load, time living with HIV, previous AIDS defining condition and prior Pneumocystis jirovecii pneumonia (PCP)). As a sensitivity analysis, we repeated logistic and linear regression analyses confining our primary analyses to spirometry measurements that were graded as easy by the examiner. A P value <0.05 was used to infer statistical significance. Statistical analyses were performed using R software V.3.3.2 (R Foundation for Statistical Computing, Vienna, Austria).
A total of 1098 PLWHIV and 12 161 uninfected controls performed spirometry. We excluded 15 (1.3%) PLWHIV and 87 (0.7%) uninfected controls from our final analyses due to FEV1 and FVC values below 400 and 500 mL, respectively. Clinical characteristics are depicted in tables 1 and 2. Most PLWHIV were Scandinavian men with mean (SD) age of 50.6 (11.3). There was a larger proportion of current smokers and more cumulative tobacco smoking in PLWHIV. The majority of PLWHIV were receiving cART treatment with suppression of viral replication (table 2).
Prevalence estimates of airflow limitation
Dynamic pulmonary function indices are depicted in table 3. FEV1/FVC<LLN was equally common in PLWHIV of 10.6% (95% CI 8.9 to 12.6) and uninfected controls of 10.6% (95% CI 10.1 to 11.1), P=1.0. FEV1/FVC<LLN was found in 4.2% (2.5–6.8) and 6.6% (6.0–7.3) in PLWHIV and controls who never smoked, respectively. In current and former smokers, the prevalence of FEV1/FVC<LLN was 14.0% (11.5–16.8) and 12.5% (11.6–13.3), P=0.31, respectively. A different prevalence estimate for airflow limitation was obtained when FEV1/FVC<0.7 with FEV1predicted <80% were used as 10.0% (8.3–11.9) of PLWHIV and 5.8% (5.4–6.2) of controls, P<0.0001, were found to have airflow limitation (table 3). In current and former smokers, the prevalence for this outcome was 13.1% (10.7–15.8) and 8.4% (7.7–9.1), P<0.0001, respectively. Venn diagrams of the two outcomes are depicted in online supplementary figure S3. The prevalence of FEV1/FVC<0.7 was 15.6% vs 17.8%, P=0.08. Severity grades of airflow limitation tended to be higher among PLWHIV (table 3).
Z-scores of spirometric indices
Histograms for FEV1, FVC and FEV1/FVC z-scores are depicted in figure 1. Mean (SD) FEV1 and FVC z-scores were lower for PLWHIV compared with controls (−0.7 (1.1) vs −0.1 (1.1), P<0.0001) and (−0.5 (1.0) vs 0.1 (1.0), P<0.0001). Mean FEV1/FVC z-scores were slightly lower for controls (−0.4 (1.0) vs −0.3 (1.1), P<0.0001).
Risk factors associated with pulmonary function
We performed univariate and multivariate logistic regression analyses to determine risk factors associated with FEV1/FVC<LLN (table 4). HIV was not associated with airflow limitation based on LLN in any of the two models (table 4). We also found no evidence for interaction of HIV with cumulative smoking on risk of airflow limitation (P=0.25), which shows that treated HIV does not modify the effect of smoking (online supplementary figure S4). HIV was independently associated with FEV1/FVC<0.7 with FEV1predicted <80% in both models considered, that is, OR 1.71 (1.34–2.16, P<0.0001) (model 1: adjusted for age, sex, ancestry and cumulative smoking) and OR 1.40 (1.05–1.85, P=0.02) (model 2: adjusted for age, sex, ancestry, cumulative smoking, educational level, previous pneumonia, preterm delivery and passive smoking during childhood). Similar results were obtained in analyses that were confined to spirometric procedures that were graded as easy (OR 1.03 (0.77–1.34, P=0.85) for LLN and OR 1.79 (1.32–2.39, P<0.001) for FEV1/FVC<0.7 with FEV1predicted <80%) and when smoking was modelled as smoking status (current, former and never smoking) or categorised according to pack-years (0–10, 10–30 and >30 pack-years).
Using continuous outcome variables of absolute FEV1 and FVC in linear regressions models, HIV was associated with a 253 mL (212–294, P<0.0001) or −6.8% (−7.8 to (−5.8), P<0.0001) lower FEV1 in model 1 and 197 mL (152–242, P<0.0001) or −6.3% (−7.4 to (−5.1), P<0.0001) lower FEV1 in model 2 (online supplementary table S1). HIV was also associated with a 395 mL (344–447, P<0.0001) or −8.2% (−9.2 to (−7.3), P<0.0001) lower FVC in model 1 and 328 mL (272–385, P<0.0001) lower FVC or −7.9% (−9.0 to (−6.9), P<0.0001) in model 2. Similar results were obtained in analyses that were confined to high-graded spirometries, after including marijuana smoke as a covariate, and after excluding PLWH with a prior diagnosis of PCP or TB (data not shown).
Self-reported morbidity (mMRC, cough and sputum) and self-reported COPD, but not asthma, tended to be more common in PLWHIV, independent of smoking history (table 5). A total of 17.5% (11.6–25.8) of PLWHIV with FEV1/FVC<LLN reported a previous diagnosis of COPD, whereas only 12.8% (10.4–15.6) of controls with airflow limitation based on LLN reported COPD. Moreover, 21.0% (14.3–29.7) of PLWHIV and 15.2% (13.2–17.4) of controls with FEV1/FVC<LLN reported use of any inhaled medication.
HIV-associated factors and pulmonary function
Considering PLWHIV separately, most recent CD4+ T cell count, CD4+ T cell nadir, CD4/CD8 ratio, detectable viral replication (>50 copies/mL), previous AIDS defining condition, previous PCP and time living with HIV were not associated with airflow limitation (LLN). Of these variables, only CD4+ T cell nadir was significantly associated with lower FEV1 and FVC after adjusting for age, sex, ancestry and cumulative smoking. Thus, a 100 cells/mm3 lower CD4+ T cell nadir was associated with a 30 mL (7–52, P<0.01) lower FEV1 and 51 mL (24–78, P<0.001) lower FVC.
The epidemiological landscape of respiratory disease in HIV has changed, and studies are needed to ascertain the long-term respiratory health consequences of HIV. To our knowledge, this study represents the largest analysis of pulmonary function in PLWHIV and uninfected controls. We found a similar prevalence of FEV1/FVC below LLN in PLWHIV compared with age-matched and sex-matched uninfected controls. However, HIV was independently associated with approximately 200 and 350 mL lower FEV1 and FVC, respectively, which seems to be associated with prior immune deficiency. These findings suggest that HIV-specific disease mechanisms contribute to lower dynamic pulmonary function.
Several factors influence COPD development and progression. Although cigarette smoking is the most well-studied risk factor, individuals who never smoke may develop chronic airflow limitation. In fact, approximately half of patients with moderate to severe COPD have developed the disease due to abnormal lung growth and development rather than due to an accelerated lung function decline during adult life.18 More concrete issues such as equipment use, seasonal and circadian variation may also affect results of spirometry.24 For this reason, strictly uniformly collected epidemiological and physiological data were prioritised in this study, and enabled us to explore the independent effect of HIV infection on pulmonary function.
The Veterans Aging Cohort study from the current cART era found HIV to be an independent risk factor for COPD using International Classification of Diseases, Ninth Revision, diagnosis codes and self-reports.25 26 However, these studies relied on diagnostic codes without spirometry evidence and lacked information on confounders associated with lung function decline, or were not able to adjust for cumulative smoking,25 which may be a better predictor of COPD than smoking categorisation.27 Spirometric studies may draw a more precise estimate of COPD and be more suitable for evaluating the pulmonary effects of HIV. Several smaller studies have assessed pulmonary function in PLWHIV (online supplementary figure S1). The four largest spirometric studies include the Strategic Timing of Antiretroviral Treatment (START) pulmonary substudy (n=1026),28 the Lung-HIV consortium study (n=908),9 the France REcherche Nord & Sud Sida-HIV et He’patites (ANRS) EP48 HIV-CHEST cohort of smokers with prior immune deficiency (CD4 nadir count <350 cells/mm3) above 40 years of age (n=351)19 and the AIDS Linked to the IntraVenous Experience (ALIVE) study of injection drug users (n=303 PLWHIV).29 The latter two studies also included uninfected controls and were able to evaluate the effect of HIV. Although HIV status itself was not found to be a risk factor for FEV1/FVC<LLN in the ALIVE study, poorly controlled HIV (viral load≥200 000 copies/mL) was associated with higher odds of airflow limitation.29 Due to the very limited number of individuals with ongoing viral replication in our study, we were not able to evaluate the effect of very high levels of viraemia, and viral loads above 50 copies/mL were not associated with any of our outcomes. The ANRS EP48 study did not assess the LLN but found HIV to be independently associated with FEV1/FVC<0.7 with FEV1predicted <80% (OR 1.72), which was similar to our results.
Our prevalence estimate of airflow limitation in PLWHIV (10.6%) is comparable to earlier reports,30–32 but several studies have found estimates exceeding 15%.4–9 These variations may be explained by uneven distribution of conventional risk factors, primarily age and tobacco smoking, and the use of different criteria for airflow limitation. The variations may also capture uneven distribution of other non-HIV and HIV-related risk factors. Such risk factors may deserve attention as they can be affected by preventive measures. In our primary analyses, educational status and pneumonia seemed to be important (dose–response dependent) predictors of both airflow limitation (LLN) and lower absolute FEV1 and FVC. None of the HIV variables studied, however, seemed to predict FEV1/FVC<LLN, but a low CD4 nadir was associated with lower FEV1 and FVC. These findings coincide with the START pulmonary substudy. This study included PLWHIV with initial CD4 cell counts >500 cells/µL and randomised participants to either immediate or deferred cART. However, the average CD4 count in the deferred arm was >600 cells per µL during follow-up. Thus, the timing of cART had no effect on the rate of lung function decline when CD4 cell counts were high, which may be in line with the notion that immune deficiency associated with HIV, rather than HIV per se, is a driver for loss of pulmonary function.
Individuals who refrain from smoking may provide additional insight into HIV disease mechanisms. We previously found evidence in well-treated PLWHIV who never smoked of small airway abnormalities, as evaluated by multiple breath washouts of nitrogen, despite having normal spirometric indices.14 In the present study, 4% of individuals who never smoked had FEV1/FVC<LLN. Notably, in the multicontinental START study, where participants were screened for good health and had a low median age at 36, half of those with FEV1/FVC<LLN were found to be never smokers.28 This observation may be explained by exposure to biomass fuel exposure, and respiratory tract infections during childhood, which are important risk factors in developing countries.33 Although these risk factors also seem to be important in our setting (13% of PLWHIV with airflow limitation never smoked), they may have greater significance in low-resource settings.
We were unable to determine why PLWHIV have a similar prevalence of FEV1/FVC<LLN (our primary outcome) as uninfected controls but a higher prevalence of FEV1/FVC<0.7 with FEV1predicted <80%. Estimates of airflow limitation for PLWHIV based on the two criteria were very similar (10.6% vs 10.0%) whereas they seemed to differ more substantially in uninfected controls (10.0% vs 5.8%). A larger number of uninfected controls meet LLN criteria while having FEV1>80% predicted (58.5%) which was less frequently occurring in PLWHIV (25.4%). Thus, differences in severity of airflow limitation between the two cohorts may explain the discrepancy using the two outcome measures. Moreover, PLWHIV did not have a higher prevalence of FEV1/FVC<0.7. The age discrepancy in the two cohorts may also have accounted for some of the difference although the discrepancy was still evident when analyses were confined to individuals above the age of 50 (13.6% vs 10.3% with FEV1/FVC<LLN and 14.8% vs 7.9% with FEV1/FVC<0.7 with FEV1 predicted <80%, respectively). Finally, the discrepancy in ethnicity did not affect the results and restricting analyses to Scandinavians yielded similar estimates (11.2% vs 10.7% with FEV1/FVC<LLN and 10.5% vs 5.9% with FEV1/FVC<0.7 with FEV1predicted <80% for PLWHIV and uninfected controls, respectively).
A large proportion of PLWHIV (82%) and controls (87%) with FEV1/FVC<LLN had no self-reported COPD, suggesting that COPD is grossly underdiagnosed in both cohorts. Screening of COPD in PLWHIV, however, cannot be recommended based on these findings, and there are currently no data indicating that such an approach would lead to improved outcomes. We also found no evidence of disparity in inhaled therapies or pharmacological treatment of tobacco dependence for PLWHIV. A total of 21% and 15% of PLWHIV and controls with FEV1/FVC<LLN were receiving some kind of inhaled therapy, respectively, and 9% current smokers with HIV received nicotine replacement therapy compared with 8% for uninfected controls.
Few mechanistic studies have assessed the respiratory effects of well-treated HIV and most studies are limited by cross-sectional designs. Initiation of cART reduces viral load in the lung compartment,34 but some replication in the lungs may still occur despite systemic suppression.35 This may, theoretically, cause direct or indirect lung damage, impair immune functions and render the host susceptible to noxious stimuli. Although we did not find interaction between smoking and treated HIV, one small study found depletion of lung mucosal CD4+ T cells, deficient functional responses of these cells, as well as increased expression of proapoptotic markers, in PLWHIV with COPD compared with uninfected individuals with COPD.36 Other evidence suggests that altered lung microbiome may be a contributor to COPD pathogenesis. P. jirovecii colonisation has received attention as it was found to be associated with airflow limitation.37 However, we were not able to confirm this association,38 and another recent study has shown that microbial populations in the lungs of well-treated PLWHIV did not differ from uninfected individuals.39
Our study has limitations. First, spirometry is effort dependent and PLWHIV may perform worse, although our findings were similar in analysis confined to spirometries where instructions were graded as easy. Second, although HIV was associated with decreased FVC (a feature of both restrictive and obstructive ventilator defects), we did not confirm a restrictive ventilatory defect by body plethysmography, which would be essential for determination of total lung capacity.40 Chest CT would also have provided further insight into the underlying pathology. Third, several covariates, including previous episodes of pneumonia, were based on self-reports and recall bias cannot be precluded. Fourth, the COCOMO cohort comprised predominantly white men that may have a lower burden of respiratory risk factors compared with other cohorts, and our findings may not be generalisable to PLWHIV in other parts of the world. Fifth, our results may be subjected to unmeasured residual confounding, although we tried to adjust for several early-life and late-life risk factors in multivariate analyses. Finally, the two cohorts may be subjected to selection bias such as healthy volunteer bias, as CGPS participants were recruited by postal mail (response rate ≈45%), whereas COCOMO participants were recruited at ambulatory visits (response rate ≈42%).
In summary, in a well-treated cohort of PLWHIV, HIV was not associated with FEV1/FVC<LLN but rather seemed to be a driver for concurrently decreased FEV1 and FVC. Although tobacco smoking cessation should be a priority in any individual, it may have even greater priority in the care of PLWHIV. Cohort studies exploring longitudinal change in spirometry, other measures of lung function and CT imaging in PLWHIV and uninfected controls are needed to better characterise lung abnormalities and elucidate their causes. Moreover, these findings have to be related to symptoms in PLWHIV in different geographical settings who have acquired HIV infection via different routes of transmission.
We thank all the study subjects for their participation. We thank the staff at the Department of Infectious Diseases at Rigshospitalet and at Hvidovre Hospital for their dedicated participation.
Contributors AnR was responsible for concept, data collection, and statistical analysis, and drafted the manuscript. JL, TB, JG, BGN and JV were responsible for concept and have had content review and editing input. SA was partly responsible for statistical analysis and has had content review and editing input. AR and AM had content review and editing input, and provided statistical support. SDN was the project leader, was responsible for concept and data collection, and has had content review and editing input.
Funding This work was supported by Rigshospitalet Research Council, Region Hovedstaden, the Lundbeck Foundation, the Novo Nordisk Foundation, and the Danish National Research Foundation grant 126. The study was designed, conducted, analysed and written by the authors without involvement of any commercial party.
Competing interests AnR: Travelling grants from Gilead. TB: Personal fees from Bristol Myers Squibb (BMS) and from Gilead and non-financial support from BMS, and from Gilead. AsR: No conflicts of interests. AM: Honoraria, lecture fees and travel support from BMS, BI, Pfizer, Merck, ViiV and Wragge. JG: Honoraria for consulting and presenting paid to his institution from Gilead, Abbvie, ViiV, BMS, MSD, Janssen and Medivir. JV: Honoraria for consulting and presenting from AstraZeneca, Boehringer-Ingelheim, Chiesi, GlaskoSmithKline and Novartis. SDN: Unrestricted research grants from the Novo Nordisk Foundation, the Lundbeck Foundation and the Rigshospitalet Research Council; travelling grants from Gilead, MSD, BMS and GSK/ViiV; advisory board activity for Gilead and GSK/ViiV.
Ethics approval The study was approved by the Regional Ethics Committee of Copenhagen (COCOMO: H-15017350; CGPS: H-KF-01-144/01). Written informed consent was obtained from all participants.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.