Spectrum of chronic kidney disease in HIV-infected patients


Dr Frank Post, Academic Department of HIV/GUM, King's College London School of Medicine at Guy's, King's College and St Thomas' Hospitals, Weston Education Centre, Cutcombe Road, London SE5 9RJ, UK. Tel: +44 207 848 5779; fax: +44 207 848 5769; e-mail: frank.post@kcl.ac.uk



The aim of the study was to investigate the prevalence and aetiology of chronic kidney disease (CKD) and trends in estimated glomerular filtration rate (eGFR) in HIV-infected patients.


Ascertainment and review of CKD cases among patients attending King's College and Brighton Hospitals, UK were carried out. CKD was defined as eGFR <60 mL/min for ≥3 months. Longitudinal eGFR slopes were produced to examine trends in renal function before, during and after exposure to indinavir (IDV) or tenofovir (TFV).


CKD prevalence was 2.4%. While HIV-associated nephropathy accounted for 62% of CKD in black patients, 95% of CKD in white/other patients was associated with diabetes mellitus, hypertension, atherosclerosis and/or drug toxicity. Exposure to IDV or TFV was associated with an accelerated decline in renal function (4.6-fold and 3.7-fold, respectively) in patients with CKD. In patients initiating IDV, age ≥50 years increased the odds of CKD [odds ratio (OR) 4.9], while in patients initiating TFV, age ≥50 years (OR 5.4) and eGFR 60–75 mL/min (OR 17.2) were associated with developing CKD.


This study highlights the importance of metabolic and vascular disease to the burden of CKD in an ageing HIV-infected cohort. In patients who developed CKD, treatment with IDV or TFV was associated with an accelerated decline in renal function.


Acute renal failure (ARF) and chronic kidney disease (CKD) are important complications of HIV infection [1]. Approximately 6% of HIV-infected patients develop one or multiple episodes of ARF [2], and 15% of patients have evidence of CKD [3,4]. ARF usually occurs in the setting of severe (opportunistic) infections, malignancy or liver disease [2, 5], and both ARF and CKD are associated with advanced immunodeficiency [2–6]. Black HIV-infected patients are at risk of developing HIV-associated nephropathy (HIVAN), which is characterized by heavy proteinuria and rapid progression to end-stage renal disease (ESRD) [7–9]. HIVAN, idiopathic noncollapsing focal and segmental glomerulonephritis (FSGS), immune complex kidney disease and other glomerulopathies predominate in biopsy series and among black patients with ESRD [7,10–13].

In the general, HIV-uninfected population, the prevalence of CKD increases dramatically in those aged 50 years and over [14]. CKD in these patients is associated with diabetes mellitus, hypertension and atherosclerosis [15], and is an independent risk factor for coronary heart disease [16]. CKD in this setting is often insidious in onset and may take decades to become clinically manifest. The dramatic improvement in life expectancy afforded by antiretroviral therapy (ART) has resulted in many HIV-infected persons living long enough to develop age-related morbidities including cardiovascular disease and CKD [17]. The combined effects of HIV-induced immune activation, adverse metabolic consequences of ART [18–20], and the propensity of specific antiretrovirals to cause renal injury [21,22] may increase the risk of vascular and renal morbidity. Indeed, a large cross-sectional cohort study found age [odds ratio (OR) 5.47 per 10 years older], any indinavir (IDV) use (OR 2.49) and any tenofovir (TFV) use (OR 2.18) to be most strongly associated with CKD [23].

The objectives of this study were to describe the prevalence and aetiology of CKD in HIV-infected patients receiving care in the UK, and to examine trends in renal function before, during and after exposure to IDV or TFV in patients with CKD.


Case ascertainment

All HIV-infected adults with CKD who attended King's College Hospital and Brighton and Sussex University Hospitals between January 1998 and December 2005 were identified. King's College Hospital serves an ethnically diverse population in South London and the HIV clinic in Brighton serves predominantly white men. At both sites, demographic, clinical and laboratory data are routinely collected prospectively on all clinic attendees in computerized databases, and blood samples for renal function, CD4 T-cell count and HIV-1 RNA level are obtained at clinic visits every 3–4 months. Serum creatinine values were obtained from the hospital electronic patient record system, and converted into estimated glomerular filtration rate (eGFR; standardized for body surface area), using age, gender, ethnicity and the four-variable modification of diet in renal disease (MDRD) equation [24]. CKD was defined as eGFR <60 mL/min for ≥3 months, and the severity of CKD categorized as stages 3–5 (eGFR 30–59, 15–29 and ≤15 mL/min, respectively) [24]. Patients with a single eGFR value, and those whose eGFR values spanned <3 months, were excluded from all analyses. The study was approved by the NHS Multi-Centre Research Ethics Committee (MREC).

Data collection

The following demographic, clinical and laboratory parameters were abstracted from the HIV clinical database: age, gender, ethnicity, risk factor for HIV acquisition, date of HIV diagnosis, dates of first and last attendance, CD4 T-cell counts, HIV-1 RNA levels, hepatitis B virus surface antigen and hepatitis C virus antibody status, AIDS-defining conditions and details of antiretroviral therapy. Medical notes were reviewed for all patients with CKD, and the presence of severe hypertension [use of antihypertensive medication or systolic blood pressure (BP)≥160 and/or diastolic BP≥100 at CKD diagnosis], diabetes mellitus (glucose intolerance requiring dietary or pharmacological intervention), clinically significant atherosclerosis (renal artery stenosis, ischaemic cerebrovascular disease, ischaemic heart disease and/or peripheral vascular disease) [23, 25] and other co-morbidities, exposure to potentially nephrotoxic medications, requirement for renal replacement therapy and status at last visit were recorded.

As susceptibility to CKD in general, and HIVAN in particular, is strongly influenced by ethnicity, patients were stratified by ethnicity as black (including ‘mixed-black’) or white/other (of whom >90% were white). Clinical, biochemical, radiological and, where available, histological information was used to determine the aetiology of CKD. Kidney disease was stratified into HIVAN, idiopathic glomerulopathies and congenital kidney disease, and CKD associated with diabetes, hypertension, vascular disease, urinary tract obstruction and drug toxicity (‘non-primary CKD’). In black patients with CKD who did not undergo kidney biopsy, HIVAN was defined clinically by all of the following criteria: (1) proteinuria >1.5 g/24 h, (2) echogenic, unobstructed kidneys on ultrasound, and (3) absence of diabetes mellitus, hypertension, collagen vascular disease, liver cirrhosis, pregnancy and organ transplant [9]. As CKD aetiology in patients with metabolic or vascular disease and drug-induced renal injury was often multifactorial and the relative contribution of each of these factors to the development and progression of CKD difficult to define, we reported all factors that may have contributed to the development or progression of nonprimary CKD without speculating on their relative importance.

Data analysis

Demographic, clinical and laboratory parameters were described for patients (1) with and without CKD, (2) with HIVAN and nonprimary CKD, and (3) with a history of IDV or TFV exposure for ≥3 months during the study period, and compared using Fisher's exact test (categorical variables), and Student's t-test or Wilcoxon's rank sum test (continuous variables). Kaplan–Meier survival analysis was performed to estimate time from CKD diagnosis to reaching stage 5 CKD (first date of eGFR <15 mL/min, sustained for >3 months, or initiation of permanent renal replacement therapy) for patients with HIVAN or nonprimary CKD. Patients were censored at death or the last clinic visit (up to November 2007), and survival curves were compared by log rank test.

Logistic regression was used to define factors associated with CKD among patients who received IDV and TNF. To reduce the risk of type 1 error for multiple comparisons, we chose α=0.02. As a consequence, only differences between groups in the univariate analysis in which the P-value was <0.01 were considered to be statistically significant, and taken forward into multivariate analysis.

To examine trends in renal function before, during and after exposure to IDV or TFV in patients with or without CKD, longitudinal eGFR slopes with 95% confidence intervals (CIs) were produced for each time period using a mixed effects linear model. This model was chosen to allow for variations in the number and temporal spacing of eGFR measurements and variable duration of drug exposure. For each patient, the intercept was variable and the only random effect in the model. Analysis was restricted to patients who had received IDV or TFV for ≥3 months and who had at least two eGFR measurements during each time period. For the eGFR slopes reflecting the post drug exposure period, we discarded eGFR values obtained within 3 months of drug cessation to reduce the residual effect of drug exposure. Slopes were adjusted for age and eGFR at the start of each time period. stata 10 (STATA Corp, College Station, TX, USA) was used for all analyses.


CKD prevalence, aetiology, and patient characteristics

During the study period, 3439 patients attended the HIV services for ≥3 months, 81 (2.4%; 95% CI 2.1–2.7%) of whom were diagnosed with CKD. The prevalence of CKD was similar among black and white/other patients (2.0%vs. 2.6%, respectively; P=0.22). Irrespective of ethnicity, patients with CKD were older, had lower nadir CD4 T-cell counts, and had more frequently experienced an AIDS-defining illness. Gender, history of injecting drug use, and hepatitis B or C virus coinfection were not associated with CKD. IDV and TFV use was more common among white/other patients with CKD (Table 1).

Table 1.   Patient characteristics
 Black/mixed-black ethnicityWhite/other ethnicity
CKD (n=26)No CKD (n=1305)P-valueCKD (n=55)No CKD (n=2053)P-value
  • CDC, Centers for Disease Control and Prevention; CKD, chronic kidney disease, defined as estimated glomerular filtration rate (eGFR) <60 mL/min for >3 months; IDU, injecting drug use; HBsAg, hepatitis B virus surface antigen; HCV Ab, hepatitis C virus antibody; IQR, interquartile range; SD, standard deviation.

  • *

    Age at CKD diagnosis or last visit (up to 31 December 2005) for patients without CKD.

  • Risk factor for HIV infection was unknown/other for 388 patients (11%), HBsAg status was available for 2303 patients (67%), and HCV Ab status was available for 2452 patients (71%).

  • For at least 3 months, up to 31 December 2005.

Age (years) [mean (SD)]*39.5 (10.6)35.4 (8.3)<0.000153.8 (12.7)39.1 (9.3)<0.0001
Female gender [n (%)]13 (50)771 (59)0.356 (11)170 (8)0.41
Risk for HIV infection [n (%)]  0.81  0.32
 Heterosexual23 (88)1158 (91) 10 (19)239 (14) 
 Homo/bisexual3 (12)106 (8) 43 (80)1334 (78) 
 IDU05 (0.4) 1 (2)131 (8) 
Nadir CD4 T-cell count (cells/μL) [median (IQR)]68 (22-148)188 (77–311)<0.000177 (38-189)207 (104-320)<0.0001
AIDS (CDC-C) [n (%)]13 (50)283 (25)0.0124 (44)427 (22)0.0001
HBsAg-positive [n (%)]2 (8)61 (9)0.935 (10)68 (5)0.45
HCV Ab-positive [n (%)]1 (4)32 (4)0.991 (2)141 (10)0.07
Indinavir-experienced [n (%)]2 (8)41 (3)0.1912 (22)150 (7)0.0001
Tenofovir-experienced [n (%)]3 (12)259 (20)0.2923 (42)558 (29)0.03

As expected, ethnicity had a major impact on CKD aetiology. HIVAN was present in 16 (62%) of 26 black patients with CKD. In 35% of black patients and 95% of white/other patients, renal failure was associated with diabetes mellitus, hypertension, atherosclerosis and/or drug toxicity (Table 2). While the diagnosis of HIVAN was pathologically confirmed in 81% of patients with HIVAN, renal biopsy was infrequently performed in patients with nonprimary CKD, and the aetiology was confirmed in only 10% of cases with nonprimary CKD. The 61 patients with nonprimary CKD had less severe renal dysfunction at the time of HIV diagnosis (median eGFR 74 vs. 25 mL/min) and at CKD diagnosis (median eGFR 55 vs. 25 mL/min), and reduced progression to stage 5 CKD (eGFR <15 mL/min or initiation of dialysis) compared with patients with HIVAN (Fig. 1). When compared with patients without CKD, these 61 patients were older (mean age 53.6 vs. 37.7 years), had lower median nadir CD4 T-cell counts (76 vs. 197 cells/μL), more often received an AIDS diagnosis (48%vs. 23%) (all P<0.001), and more often received treatment with IDV (24%vs. 6%; P<0.001) or TFV (43%vs. 26%; P=0.002). In contrast, none of the 16 patients with HIVAN or the four patients with idiopathic glomerulopathies or congenital kidney disease had received IDV- or TFV-containing highly active antiretroviral therapy (HAART).

Table 2.   Confirmed or suspected aetiology of chronic kidney disease in 81 HIV-infected patients, stratified by ethnicity
 All patients
  • CKD, chronic kidney disease [estimated glomerular filtration rate (eGFR) <60 mL/min for >3 months]; HIVAN, HIV-associated nephropathy.

  • *

    Noncollapsing focal and segmental glomerulosclerosis (FSGS; n=1), mesangiocapillary glomerulonephritis (n=1), immunoglobulin A (IgA) nephropathy (n=1), or congenital kidney disease (n=1).

  • CKD associated with metabolic/vascular disease, drug toxicity, etc. (multiple aetiologies may be present).

  • Renal artery stenosis, ischaemic heart/cerebrovascular disease or peripheral vascular disease.

  • §

    Acute tubular necrosis with incomplete resolution (n=4), recurrent urinary tract infection (n=2), kidney stones or ureteric obstruction (n=6), congestive cardiac failure/diuretics (n=1), or aetiology unclear (n=1).

  • At or prior to CKD diagnosis, up to 31 December 2005.

HIVAN [n (%)]16 (20)16 (62)0
Idiopathic CKD* [n (%)]4 (5)1 (4)3 (5)
Nonprimary CKD [n (%)]61 (75)9 (35)52 (95)
 Diabetes mellitus18216
 Clinically significant atherosclerosis17116
 Nephrotoxic medication (any drugs)49643
  Nonantiretroviral medication28226
  Antiretroviral medication39534
Figure 1.

 Natural history of chronic kidney disease (CKD) in HIV-infected patients: progression from CKD diagnosis [estimated glomerular filtration rate (eGFR) <60 mL/min] to stage 5 CKD (eGFR <15 mL/min or initiation of dialysis) among 16 patients with HIV-associated nephropathy (HIVAN) (broken line) and 61 patients with nonprimary CKD (CKD associated with metabolic/vascular disease, drug toxicity, etc.) (solid line). *CKD associated with diabetes mellitus, hypertension, atherosclerosis, urological disease, drug toxicity or other non-HIV-related illnesses.

Characteristics of patients with CKD and a history of IDV or TFV exposure

Of the 205 patients who received IDV during the study period, 14 (6.8%) developed CKD while receiving IDV. IDV was stopped because of renal concern in 13 patients, and one patient was lost to follow-up while receiving IDV. Patients with IDV-associated CKD were older (mean age 49.6 vs. 39.9 years; P<0.001), had poorer renal function (median eGFR 99 vs. 108 mL/min; P=0.03) at the time of IDV initiation, and had received IDV for longer (median 3.0 vs. 1.2 years; P=0.007) by the end of the study period compared with patients who did not develop CKD, while a similarly low proportion of patients had received ritonavir-boosted IDV (15%vs. 12%; P=0.93). Ten patients with IDV-associated CKD (71%) had at least one other CKD risk factor.

Of the 843 patients who received TFV during the study period, 26 (3.1%) developed CKD. CKD developed during treatment with TFV in 21 patients, while CKD was present at the time of TFV initiation in four patients and CKD developed post-TFV discontinuation in one patient. TFV was discontinued in 22 patients (in 21 because of renal concern), and four patients died while receiving TFV. TFV was appropriately dose adjusted in nine of 13 patients with creatinine clearance below 50 mL/min, and 22 patients (85%) with TFV-associated CKD had at least one other CKD risk factor.

Patients with TFV-associated CKD were older (mean age 53.6 vs. 39.9 years; P<0.001), had more prolonged known HIV infection (median duration 10.1 vs. 5.4 years; P=0.001), had more prolonged exposure to ART (median duration 6.2 vs. 2.9 years; P<0.001) and had impaired renal function (median eGFR 69 vs. 102 mL/min; P<0.001) at the time of TFV initiation compared with those who did not develop CKD while receiving TFV during the study period. The median duration of TFV exposure (up to 31 December 2005) was similar in patients with and without CKD (1.78 vs. 1.83 years; P=0.7), and similar proportions had received prior treatment with IDV (16%vs. 11%, respectively; P=0.43), while none had received IDV and TFV concomitantly.

eGFR slopes prior to, during and after IDV or TFV exposure in patients with and without CKD

To examine a potential effect of IDV or TFV on renal function, longitudinal eGFR slopes were produced for patients with and without CKD who had received these drugs for a minimum of 3 months. A modest decline in eGFR (−2.6 and −2.3 mL/min/year, respectively) was observed in patients with CKD prior to IDV or TFV exposure. An accelerated decline in renal function was observed in patients with CKD during exposure to IDV or TFV (4.6-fold and 3.7-fold, respectively), while IDV or TFV discontinuation was associated with a reduced eGFR decline. Patients without CKD generally had stable renal function. In these patients, a modest decline in eGFR was observed during treatment with IDV, and a modest improvement in eGFR was observed following discontinuation of TFV (Fig. 2).

Figure 2.

 Estimated glomerular filtration rate (eGFR) slopes in patients with chronic kidney disease (CKD; eGFR <60 mL/min for >3 months) and patients without CKD prior to, during and after exposure to indinavir- or tenofovir-containing antiretroviral therapy. Numbers in parentheses reflect patients in each stratum.

Factors associated with IDV- or TFV-associated CKD

Factors associated with development of CKD among patients who had received IDV or TFV were analysed in multivariate logistic regression analysis. For patients who received IDV, age ≥50 years was associated with increased odds of CKD, while among those who initiated TFV, age ≥50 years and reduced eGFR increased the odds of CKD (Table 3).

Table 3.   Logistic regression analysis of factors associated with chronic kidney disease (CKD)
 Univariate analysisMultivariate analysis
OR95% CIP-valueOR95% CIP-value
  • In multivariate analysis, the model for indinavir (IDV) was adjusted for age, estimated glomerular filtration rate (eGFR) at IDV initiation, and duration of IDV exposure and the model for tenofovir (TFV) adjusted for age, duration of HIV infection, duration of antiretroviral therapy, and eGFR at TFV initiation.

  • *

    At IDV/TFV initiation.

  • Up to 31 December 2005.

  • Per additional year of therapy.

  • OR, odds ratio.

Indinavir (IDV)
 Age <50 years*1  1  
 Age ≥50 years*6.00(1.70–21.4)0.0064.92(1.31–18.4)0.02
 Duration of IDV therapy†‡1.34(1.06–1.69)0.021.29(1.00–1.65)0.047
Tenofovir (TNF)
 Age <50 years*1  1  
 Age ≥50 years*12.1(4.82–30.2)<0.0015.42(1.71–16.8)0.004
 Duration of HIV infection*‡1.10(1.02–1.19)0.010.90(0.80–1.11)0.594
 Duration of prior ART*‡1.24(1.11–1.37)<0.0011.22(1.00–1.50)0.048
 GFR ≥90 mL/min*1  1  
 GFR 75–89 mL/min*7.10(1.67–29.9)0.0084.10(0.91–18.3)0.066
 GFR 60–74 mL/min*39.0(10.2–148.7)<0.000117.2(3.91–74.9)<0.0001


This study describes the clinical epidemiology of HIV-associated CKD in the HAART era, and explores the previously reported association between CKD and use of IDV and TFV. In black patients, HIVAN accounted for two-thirds of the cases of CKD and was associated with low eGFR at initial presentation, advanced immunodeficiency and poor renal outcome. In white/other patients, approximately 95% of CKD cases were associated with metabolic, vascular and urological disease and exposure to nephrotoxic medications. These nonprimary forms of CKD were commonly absent or subclinical at the time of HIV diagnosis, and associated with older age, more advanced HIV disease, and less severe renal failure. While the aetiology of CKD associated with IDV or TFV use was often multifactorial, accelerated eGFR decline was observed while these patients were exposed to IDV or TFV.

Most of the literature on HIV-associated CKD has focused on patients with biopsy-defined CKD or those who progressed to ESRD. As HIVAN almost exclusively affects black patients [7,9] and frequently results in ESRD [7,9,26], severe CKD is strongly associated with black ethnicity [3,27–31]. However, when CKD was systematically ascertained, similar prevalence rates were noted in black and white/other patients [23]. Our study suggests that the susceptibility to HIVAN of black patients may be matched by an increased susceptibility of older, nonblack patients to nonprimary CKD. Indeed, increasing age and co-morbidities associated with CKD in the general population, including diabetes mellitus, clinically significant atherosclerosis and hypertension, have been associated with CKD in HIV-infected patients [23,29,32,33].

CKD in HIV-infected patients is associated with a low nadir CD4 T-cell count [3,4,6,23]. In our cohort, HIVAN and primary forms of CKD were almost invariably present at the time of HIV diagnosis, and advanced immunodeficiency reflected late diagnosis of HIV infection [9]. While all initiated ART, IDV and TFV were generally avoided in these patients. In contrast, many patients with nonprimary CKD were in HIV care and receiving ART when CKD manifested. In some, CKD or the vascular and metabolic risk factors for CKD were present at the time of HIV diagnosis, while in others ageing, HIV infection, immunodeficiency, ART and nephrotoxic medications contributed to the development of CKD. Our results underscore the importance of current recommendations to assess renal function (eGFR and urinalysis) in HIV-infected patients at baseline, and annually thereafter if risk factors such as diabetes mellitus or hypertension are present, and to monitor renal function in patients with CKD to ensure appropriate dosing of renally excreted (antiretroviral) medications [24].

The previously reported association between use of IDV or TFV and CKD [23] was confirmed in our cohort. We studied the eGFR slopes prior to, during and after discontinuation of IDV or TFV in patients with CKD and noted an accelerated decline in renal function while patients received these drugs. Renal complications, including gradual loss of renal function and progressive or irreversible renal failure, have been reported in 7–33% of patients who receive IDV, and largely relate to crystallization of IDV in the urinary tract [34]. Reduced eGFR at IDV initiation did not increase the odds of developing CKD, and renal function was affected in all recipients, confirming a propensity of IDV to cause renal injury in the absence of pre-existing renal disease. In contrast, the odds of developing CKD and a decline in renal function while receiving TFV were strongly associated with reduced renal function at baseline, raising the possibility that renal toxicity was mediated by reduced TFV excretion. In several studies, TFV use has been associated with a mild and generally nonprogressive decline in creatinine-based measures of renal function [34–37]. Clinically significant TFV-associated renal toxicity in cohort studies is observed in approximately 1% (range 0.3–2.3%) of patients [35,38], and may present as progressive decline in renal function, Fanconi syndrome, nephrogenic diabetes insipidus, hypophosphataemia, hypokalaemia, reduced bone mineral density and/or acute tubular necrosis [22,39,40]. While the risk factors for TFV-associated renal toxicity have not been fully defined, prolonged administration of TFV, co-administration of TFV with didanosine, (boosted) protease inhibitors or nephrotoxic agents, age >50 years, advanced HIV infection, pre-existing renal failure, and low body weight may predispose to TFV-associated renal toxicity [22,40,41]. Our results suggest that older patients and those with impaired renal function at the time of TFV initiation may be more susceptible to renal injury while receiving TFV, while TFV appeared to have no deleterious effect on eGFR in patients without CKD, as observed in a recent randomized clinical trial [36]. Nonetheless, renal function improved in patients without CKD who discontinued TFV, suggesting that TFV may affect renal function in a subset of patients without CKD. However, as the effects of TFV on the proximal tubular organic cation transporter responsible for tubular secretion of creatinine have not been well defined, changes in creatinine-based measures of renal function in patients receiving TFV should be interpreted with caution, and noncreatinine-based GFR measurements may provide a more reliable estimate of the contribution of TFV to CKD in HIV-infected patients.

This study has several limitations. Despite the sizeable cohorts, the number of CKD cases was relatively small. The lack of histological confirmation of the aetiology of most cases of nonprimary CKD raises the possibility that some patients categorized as having nonprimary CKD may in fact have had HIVAN or idiopathic glomerulopathies. During the study period, screening for proteinuria was not routine clinical practice. The lack of information on proteinuria will have resulted in substantial underestimation of CKD prevalence [3,4]. Further limitations include the limited availability of information on diabetes, hypertension and clinically significant atherosclerosis in patients without CKD, limited sensitivity of clinical parameters to exclude the presence of clinically significant atherosclerosis, and limited information on the use of nephrotoxic nonantiretroviral medications.

In conclusion, our study highlights the important contribution of IDV- and TFV-associated nephrotoxicity and metabolic and vascular disease to the burden of CKD in an ageing HIV-infected cohort. Older patients, and those with reduced renal function on ART, are at risk of an accelerated decline in renal function following initiation of IDV or TFV. IDV and TFV should be used with caution in these patients, and renal function carefully monitored.


The authors wish to thank Stephen Duffell, Nicky Perry, Anthony Pullin, Richard Shanahan and Elizabeth Tissingh for their assistance with data collection, IT, R&D submission, and/or access to hospital notes.

Funding: Funding for this study was obtained from GlaxoSmithKline (grant COL109793 to FAP). The funder was not involved in study design, data analysis or interpretation of the results.

Conflict of interest: FAP, SGH, MF and BMH have received research grants or honoraria from Gilead Sciences and GlaxoSmithKline. None of the other authors has a conflict of interest to report.