Neurological opportunistic infections and neurological immune reconstitution syndrome: impact of one decade of highly active antiretroviral treatment in a tertiary hospital

Authors


Correspondence: Dr Vicenç Falcó, Infectious Diseases Department, Hospital Vall d'Hebron, P° Vall d'Hebron 119-129, 08035 Barcelona, Spain. Tel: 34 93 274 60 90; fax: 34 93 489 40 91; e-mail: vfalco@vhebron.net

Abstract

Background

Despite the reported decrease in the incidence and mortality rates of central nervous system (CNS) infections after the introduction of highly active antiretroviral therapy (HAART), few studies have focused on the global incidence and the relationship of these diseases with immune reconstitution inflammatory syndrome (IRIS) in the developed world.

Methods

A descriptive cohort study of all consecutive adult HIV-infected patients with CNS opportunistic infections diagnosed between 2000 and 2010 in a tertiary hospital in Spain was carried out. Demographic, clinical, laboratory, and microbiological data were recorded. Patients were followed up until death or loss to follow-up or until 30 July 2011, when the study finished. The significance of differences in the incidence rate between early and late HAART periods was determined using the Mantel–Haenszel test. Survival distribution was estimated using the Kaplan–Meier method.

Results

A total of 110 cases of CNS infections were diagnosed. The incidence of CNS opportunistic infections decreased from 9 cases per 1000 HIV-infected patients per year in the early HAART period to 3.8 in the late HAART period (P = 0.04). Overall, the estimated mean survival time was 58.8 months (95% confidence interval 47.1–70.6 months). Of the 110 patients, 18 (16.4%) met the criteria of IRIS, 10 (55.6%) were paradoxical and eight (44.4%) were unmasking. IRIS was not associated with a higher mortality rate.

Conclusions

The annual incidence of CNS infections decreased progressively during the period of study. The mortality rate associated with these diseases remains high despite HAART. The development of IRIS associated with neurological infections had no influence on prognosis.

Introduction

The widespread use of highly active antiretroviral therapy (HAART) has led to a dramatic decline in the incidence of new AIDS cases and most opportunistic illnesses [1-3]. In the developed world, cases of opportunistic neurological infections such as cryptococcal meningitis, tuberculous meningitis, cerebral toxoplasmosis and progressive multifocal leukoencephalopathy (PML) are nowadays becoming infrequent [4-6]. For this reason, in the last decade, most studies on opportunistic infections have been performed in limited-resource settings where their incidence is still high as a consequence of the lack of availability of HAART.

Some patients beginning HAART experience clinical deterioration despite decreasing HIV RNA levels and rising CD4 cell counts. These symptoms are the results of a paradoxical inflammatory response to both infectious and noninfectious antigens attributable to the recovery of the immune system. This inflammation has been termed immune reconstitution inflammatory syndrome (IRIS) [7-12]. Reports of cases of IRIS involving the central nervous system (CNS) are increasing and the outcomes for these patients seem to be worse than those for patients with non-neurological IRIS [8, 10]. At present there is some uncertainty about the clinical significance of neurological IRIS, and in particular about the optimal time to initiate HAART in patients with CNS opportunistic infections. In this context, a randomized clinical trial performed in sub-Saharian Africa concluded that early initiation of HAART resulted in increased mortality in patients with cryptococcal meningitis [13].

Because information about clinical outcomes in HIV-infected patients with a CNS opportunistic infection, and the effect of IRIS on their prognosis, has been scarce in developed countries in the last decade, we conducted an observational study of patients diagnosed with a CNS opportunistic infection. The aim of this study was to investigate the incidence and survival of patients with CNS opportunistic infections and the characteristics of IRIS related to these infections during the first decade of the 21st Century in a setting in which the use of HAART has become the standard of care for HIV-infected patients.

Patients and methods

Study design

A descriptive cohort study of all consecutive adult HIV-infected patients with CNS opportunistic infections diagnosed between 1 January 2000 and 31 December 2010, in a single-centre tertiary hospital in Barcelona, Spain, was carried out.

Case definitions

Diagnosis of PML was based on clinical and radiological findings. Neuroradiological diagnosis of PML was established by magnetic resonance imaging (MRI) when the following abnormalities were present: asymmetric and well-demarcated lesions hyperintense in T2 and hypointense in T1, with no mass effect and with location in white matter [14-17].

Cerebral toxoplasmosis was diagnosed when the following criteria were present: (1) progressive neurological deficits, (2) a contrast-enhancing mass lesion in imaging findings [computed tomography (CT)/MRI] and (3) a successful response (defined as a significant improvement in clinical and neuroradiological findings with a CT or MRI performed at 2 weeks) to specific treatment within 2 weeks [16].

Diagnosis of cryptococcal meningitis was suspected in patients with clinical manifestations of meningitis and was confirmed by any of the following methods: (1) visualizing the fungus in the cerebrospinal fluid (CSF) using India ink, (2) detecting cryptococcal antigen using a latex agglutination assay in the CSF or (3) a positive CSF culture for Cryptococcus neoformans [16].

Tuberculous meningitis was diagnosed in patients with clinical meningitis and demonstration of Mycobacterium tuberculosis in the CSF (culture, CSF smear or polymerase chain reaction) [15].

An IRIS diagnosis was made when all the following criteria were present: (1) recurrence of symptoms and signs of a previously identified and treated CNS infection (paradoxical IRIS) or new onset of clinical and neuroradiological findings (unmasking IRIS) within 6 months after HAART initiation, (2) a decrease in the plasma viral load of ≥ 1 log10 HIV-1 RNA copies/ml, and (3) the presence of symptoms not explained by a newly acquired disease, by the usual course of a previously acquired illness or by pharmacological toxicity [7-12, 18, 19].

Study variables and data collection

The following data were recorded: demographic, clinical, laboratory, radiological and microbiological data, antiretroviral therapy, clinical course and mortality. Patients were followed up until death or loss to follow-up or until 30 July 2011, when the database was closed.

Estimation of the incidence of CNS opportunistic infections

Incidences of CNS opportunistic diseases were estimated using as the denominator the number of HIV-infected persons alive registered in the database of our hospital, and were expressed as cases per 1000 HIV-infected people per year.

In order to explore a change in the incidence trend during the study, two treatment periods were defined: the ‘early HAART period’, from 1 January 2000 to 31 June 2005, and the ‘late HAART period’, from 1 July 2005 to 31 December 2010. The incidence was calculated as the number of events per 1000 HIV-infected persons per year. To increase reliability, because the numbers of patients with some of the infections studied were small, the incidence taking into account the total number of new cases of CNS infections on every period was also calculated.

Statistical analysis

Statistical analyses were performed using the statistical software package spss for Windows, version 19.0 (SPSS, Chicago, IL). The significance of differences in mean incidences between the early and late HAART periods was determined using the Mantel–Haenszel test. Changes in incidences were reported with their associated 95% confidence intervals (CIs). Continuous variables are expressed as the median and interquartile range (IQR) or mean and standard deviation, as appropriate, and were compared using the Student t-test or the Mann–Whitney U-test. Categorical variables were compared using the χ2 test or the Fisher exact test. The survival distribution was estimated using the Kaplan–Meier method. The comparison of survival between the different subject groups was performed using the log-rank test. A P-value < 0.05 was considered statistically significant.

Results

Baseline characteristics and incidence of CNS opportunistic infections

One hundred and ten patients with a CNS opportunistic infection were diagnosed between 2000 and 2010: 37 cases of cerebral toxoplasmosis, 23 of cryptococcal meningitis, 10 of tuberculous meningitis and 40 of PML. The baseline characteristics of the patients are shown in Table 1. The median CD4 lymphocyte count at diagnosis of CNS infection was 38 cells/μL (IQR 12–108 cells/μL). Eleven out of 110 (10%) patients had a CD4 lymphocyte count > 200 cells/μL when CNS infection was diagnosed. Those patients had tuberculous meningitis or PML, mainly associated with unmasking IRIS.

Table 1. Baseline characteristics of patients with central nervous system (CNS) infections
  1. Results are expressed as median (IQR) or n (%).
  2. MSM, men who have sex with men.
  3. aData were missing for five patients..
Age (years)39 (34–42)
Male sex81 (73.6%)
Risk factors for HIV infection 
Injecting drug use57 (51.8%)
Heterosexual transmission25 (22.8%)
MSM13 (11.8%)
Blood transfusion9 (8.2%)
Vertical transmission6 (5.5%)
Previous diagnoses of HIV infection94 (85.5%)
Prior AIDS-defining conditions50 (45.5%)
Current HAART31 (28.2%)
Coinfection with hepatitis C virus73 (66.4%)
CD4 lymphocyte count at diagnosis (cells/μL)38 (12–108)a
Number of patients with CD4 count < 100 cells/μL77 (70%)a
Number of patients with CD4 count > 200 cells/μL11 (10%)a
HIV-1 viral load (log10 copies/mL)5.04 (4.1–5.7)a
Number of patients with HIV-1 viral load < 50 copies/mL9 (8.2%)a

In 16 (14.5%) patients, HIV and CNS opportunistic infections were diagnosed simultaneously. Thirty-one out of 94 (33%) patients with a previously known HIV infection were receiving HAART at the onset of CNS infection; 19 of them had detectable levels of serum HIV-1 viral load, mainly as a result of poor adherence.

The annual incidence and the linear tendency are represented in Figure 1. The incidence of CNS opportunistic infections decreased from 9 cases per 1000 HIV-infected-patient-years in the ‘early HAART period’ to 3.8 in the ‘late HAART period’ (P = 0.04). When we calculated the incidence as the total number of cases per 1000 HIV-infected-patient on each period, there was a decrease from 49.5 cases in the early HAART period to 20.9 cases in the late HAART period (P < 0.001). During the study period, the proportion of patients on HAART in the overall cohort did not change significantly (75.7% vs. 77.2% in the early and late periods, respectively). However, the proportion of patients with CD4 lymphocyte counts of < 200 cells/μL decreased from 17.7% to 10.1% and the proportion of patients with undetectable viral load increased from 64.1% to 78.6%. In Table 2 we show the incidences of the different CNS infections. Regarding the comparison between the early and late periods, the incidence of all CNS infections decreased significantly, except for PML, the incidence of which remained stable.

Figure 1.

Annual incidence (per 1000 HIV-infected patients) of central nervous system (CNS) opportunistic infections and linear tendency.

Table 2. Number of cases and mean annual and global incidences categorized by highly active antiretroviral therapy (HAART) period
 Early HAART periodLate HAART periodEarly vs. late HAART periods
 Number of cases per periodCases/1000 HIV-infected patients per yearCases/1000 HIV-infected patients per periodNumber of cases per periodCases/1000 HIV-infected patients per yearCases/1000 HIV-infected patients per period1000 HIV-infected-year1000 HIV-infected-period
Change [% (95% CI)] P Change [% (95% CI)] P
  1. CI, confidence interval; CNS, central nervous system; PML, progressive multifocal leukoencephalopathy.
All CNS infections708.9949.50403.8020.895.5 (0.1 to 12.3)0.04138.6 (15.6 to 41.6)<0.001
Cerebral toxoplasmosis253.2117.67121.146.272.5 (−0.9 to 5.9)0.12111.4 (3.7 to 19.1)0.002
Cryptococcal meningitis151.9310.6080.764.181.6 (−1 to 4.2)0.1886.4 (0.4 to 12.5)0.027
Tuberculous meningitis81.035.6520.191.040.7 (−0.7 to 2)0.2454.6 (0.4 to 8.8)0.016
PML222.8315.55181.719.401.3 (−2.0 to 4.5)0.4326.1 (−1.6 to 13.9)0.107

Outcomes for patients with CNS opportunistic infections

The median duration of follow-up was 22 months (IQR 4–54 months). Thirty-four patients (31%) died and 32 (29%) were lost to follow-up during the study period. The two groups of patients were considered together in all survival analyses. At 3 months, 56 patients had clinical improvement and 16 remained stable. However, in 14 patients neurological damage worsened and 24 died or were lost to follow-up. In the early HAART period, 25 of 70 patients (35.7%) died compared with nine of 40 (22.5%) in the late HAART period (P = 0.15).

Overall, the estimated mean survival time was 58.8 months [95% confidence interval (CI) 47.1–70.6 months]. The Kaplan–Meier estimates of probability of survival were 79% (95% CI 71.5–86.7%) at 3 months, 71.8% (95% CI 63.4–80.2%) at 6 months, 61.7% (95% CI 52.7–70.7%) at 12 months, 48.3% (95% CI 38.9–57.7%) at 36 months and 36.7% (95% CI 26.9–46.5%) at 60 months. The estimated median survival time expressed in months and the cumulative survival time for the different CNS opportunistic infections are shown in Figure 2. Differences in the survival time among the CNS infections did not reach statistical significance.

Figure 2.

Cumulative probability of survival categorized by the central nervous system (CNS) infection and mean survival time [months; 95% confidence interval (CI)]. PML, progressive multifocal leukoencephalopathy.

IRIS

Eighteen of 110 cases (16.4%) met the criteria of IRIS. Of these, 10 patients (55.6%) had PML. IRIS was diagnosed in four of 36 (11.1%) patients with cerebral toxoplasmosis, three of 21 (14.3%) with cryptococcal meningitis, one of 10 (10%) with tuberculous meningitis and 10 of 35 with PML (28.6%). The proportions of IRIS cases in the early and late HAART periods were similar (P = 0.8). The infection was newly identified after the initiation of HAART (unmasking IRIS) in eight out of 18 cases (44.4%). In the remaining 10 cases (55.6%), IRIS was diagnosed after worsening of a previously treated CNS infection (paradoxical IRIS). The median interval from HAART initiation to diagnosis of IRIS was 39 days (IQR 20–90 days).

Table 3 shows demographic, clinical and immunological characteristics of patients who developed paradoxical and unmasking IRIS.

Table 3. Demographic and clinical baseline characteristics of patients according to the diagnosis of immune reconstitution inflammatory syndrome (IRIS)
 IRISNo IRIS (n = 92)Paradoxical IRIS vs. no IRIS P
Paradoxical (n = 10)Unmasking (n = 8)
  1. CNS, central nervous system; HAART, highly active antiretroviral therapy; IQR, interquartile range.
  2. *Data were missing for nine patients. †Only for patients with unmasking IRIS.
Age (years) [median (IQR)]39 (37–45)40.5 (38–41.5)38.5 (33–42)0.210
Male sex [n (%)]8 (80)4 (50)69 (75)0.727
Previous diagnoses of HIV infection [n (%)]8 (80)8 (100)78 (84.8)0.693
Prior AIDS-defining condition [n (%)]0 (0)3 (37.5)47 (51.1)0.002
Current HAART [n (%)]0 (0)8 (100)23 (25)0.072
CD4 lymphocyte count at diagnosis of CNS infection (cells/μL) [median (IQR)]80.5 (48–108)30 (10–86)* 0.881
CD4 lymphocyte count at HAART initiation (cells/μL) [median (IQR)] 64 (13.5–151.5)
CD4 lymphocyte count at diagnosis of IRIS (cells/μL) [median (IQR)] 169 (157–240)139.5 (113.5–231)  
CD4 lymphocyte count at 3 months of CNS infection (cells/μL) [median (IQR)]190 (160–213)139.5 (110.3–353)106 (38.5–192.5)0.100
CD4 lymphocyte count at 6 months of CNS infection (cells/μL) [median (IQR)]230 (184.3–416.3)173 (83.8–469)145 (80–227.5)0.078
HIV-1 viral load at HAART initiation (log10 copies/mL) [median (IQR)] 5.6 (4.6-6.3)  
HIV-1 viral load at diagnosis of CNS infection (log10 copies/mL) [median (IQR)]5.1 (4.7-5.9)3.1 (1.7-3.9)5.15 (4.5-5.7)* 0.639
Patients with HIV-1 viral load < 50 copies/mL at diagnosis of CNS infection [n (%)]0 (0)3 (42.9)6 (7.2)* 0.379
Patients with HIV-1 viral load < 50 copies/mL at 3 months of CNS infection [n (%)]2 (28.6)2 (33.3)19 (29.7)0.951
Patients with HIV-1 viral load < 50 copies/mL at 6 months of CNS infection [n (%)]5 (83.33)5 (71.4)33 (56.9)0.209
Early onset of HAART (within 2 weeks after the diagnosis of CNS infections) [n (%)]5 (50)52 (65.8)0.326
HAART based on protease inhibitors [n (%)]6 (60)3 (37.5)47 (58.8)0.940
Late HAART period [n (%)]2 (20)5 (62.5)33 (35.9)0.315

In order to identify pretreatment variables associated with the risk of developing paradoxical IRIS, these patients were compared with those who did not experience a paradoxical reaction. We found, as the only difference between the two groups, that patients who did not develop IRIS were more likely to have had a previous AIDS-defining condition (51.1% vs. 0% for those developing paradoxical IRIS; P = 0.002). Patients developing IRIS had a more rapid immunological recovery than patients who did not develop IRIS, as evidenced by a greater increase in CD4 count after 3 months of antiretroviral therapy (ART) (170 vs. 62 cells/μL, respectively; P < 0.025). At this time-point, the decrease in viral load was also greater among patients with paradoxical IRIS, but differences did not reach statistical significance (–2.6 vs. −1.8 log10 for those with paradoxical IRIS; P = 0.10). Patients who began HAART within 2 weeks after the diagnosis of a CNS infection were not at higher risk of developing paradoxical IRIS (50% vs. 65.8% for those who began HAART more than 2 weeks after diagnosis; P = 0.32). Figure 3 shows the cumulative probabilities of survival and the median survival time categorized by the development and type of IRIS. We did not find significant differences in survival between patients who developed paradoxical IRIS and those without IRIS.

Figure 3.

Cumulative probability of survival categorized by the development and type of immune reconstitution inflammatory syndrome (IRIS). CI, confidence interval.

Eight (44.4%) of the 18 patients with IRIS received therapy with steroids for a variable period depending on the response to therapy and other individual patient characteristics. None of the 10 patients who were not treated with steroids died, while three of the eight who received steroids died. In those three cases, mortality was directly attributed to IRIS. These three patients had PML.

Discussion

In our study, we observed a progressive decline in the incidence of CNS opportunistic infections during the first decade of the 21st Century. The overall rate of CNS infections decreased significantly from 8.3 cases per 1000 HIV-infected patients in the year 2000 to 1.4 in 2010. Since HAART became available, many studies have reported a decrease in the incidence of most opportunistic conditions related to HIV infection, including neurological infections [1-6, 20-22]. For example, a study performed in France by Abgrall et al. in 2001 showed a reduction of 34% in the risk of cerebral toxoplasmosis after the introduction of protease inhibitors [5]. However, such tendency has not been the same for all opportunistic infections. We observed a decline in the incidence of all CNS opportunistic infections except for PML. Different studies performed in France, Spain and Denmark have also shown a stabilization in the incidence of PML despite the widespread use of HAART [17, 23, 24]. This may be partly explained by the appearance of new cases of PML after the introduction of HAART associated with unmasking IRIS, as previously noted [25].

Different studies have shown a higher survival rate for CNS infections after the introduction of HAART [26, 27]. Indeed, patients with PML, which is considered the most devastating CNS disorder associated with HIV, have shown improved prognoses [27-29]. Before the introduction of HAART, the median survival time for PML was 8–15 weeks [30], in contrast to the 44.5 months of estimated survival in our cohort. These data are similar to those obtained in other cohort studies performed in the HAART era [17, 24, 26, 27, 31, 32]. However, despite the improvement in survival and the reduction in the incidence, it is important to point out that overall prognosis of patients with CNS opportunistic infections is still poor and most patients experience mild to severe neurological impairment and require long-term care [24, 25, 31, 32]. In our cohort, 31% of patients died and 29% were lost to follow-up. During the first 3 months after diagnosis of the CNS infection, the condition of 14 patients worsened and 24 died or were lost to follow-up. Finally, the estimated probability of survival was only 48% at 3 years. Taken together, these data indicate the necessity of early diagnosis of HIV infection and HAART in order to avoid the possibility of developing a CNS opportunistic infection.

The incidence of IRIS in our cohort was 16.4%. This observation agrees with those in other cohorts, where between 17 and 25% of patients developed one or more manifestations as a consequence of the inflammatory syndrome after starting HAART [8, 33, 34]. A prospective study performed in South Africa showed an incidence of 10% for patients initiating ART, including both unmasking and paradoxical forms of IRIS [35]. In our series, IRIS presented as paradoxical IRIS in 55.5% of cases and the remaining 44.5% had unmasking IRIS. This finding is consistent with data from a multicentre cohort in which each type of IRIS represented 50% of cases [34].

Regarding the different neurological infections, two prospective studies reported that 13–17% of HIV-infected patients with cryptoccocal meningitis developed paradoxical IRIS after initiation of HAART [9, 36]. Of the 44 cases of IRIS described by Murdoch et al., 6.8% corresponded to cryptoccocal meninigitis, all of them unmasking IRIS [35]. Concerning PML, which has been the disease most commonly related to the development of IRIS, 25% of our cases met the criteria of IRIS, similar to the 18–23% described in previous observational studies [17, 27]. In our cohort, five of 40 (12.5%) cases of PML met the criteria of unmasking IRIS, a percentage similar to those given in previous reports [27]. With regard to tuberculosis, although paradoxical IRIS associated with Mycobacterium tuberculosis occurs in over 45% of individuals, CNS complications are presumed to be much lower [37]. In our cohort, just one case of IRIS was related to tuberculosis. In two recent studies of 80 and 144 patients coinfected with tuberculosis and HIV, no cases of CNS-associated complications were reported [38].

In our study, none of the patients who presented a paradoxical IRIS had a previous history of any AIDS-defining opportunistic infection. Previous studies had suggested that patients previously undiagnosed with HIV infection presented a higher risk for the development of paradoxical IRIS if a CNS opportunistic infection was the AIDS-defining illness [4]. We did not find any other clinical or immunological parameter at baseline that predicted the development of IRIS after initiation of HAART. As previously described, in our cohort baseline CD4 cell count was not predictive of developing paradoxical IRIS [4, 39, 40]. However, patients who developed paradoxical IRIS had a more rapid immune restoration in response to HAART than patients who did not. This is consistent with findings from both retrospective and prospective analyses which revealed a greater CD4 response in patients developing IRIS [7, 9, 40-42]. As previously indicated, no differences in the risk of developing IRIS were observed when protease inhibitor-containing regimens were compared with other regimens [11, 39].

In our cohort, initiation of HAART during the first 2 weeks after the diagnosis of a neurological infection was not associated with a higher risk for the development of paradoxical IRIS. However, the retrospective design of our study and the low number of patients with IRIS limit the significance of this observation. Optimal timing for initiation of HAART is still a controversial issue in patients with CNS opportunistic infections. Nowadays we have consistent data that indicate the benefits of early HAART in patients with opportunistic infections, even in those with tuberculosis [43]. However, in none of these studies were patients with CNS infections sufficiently represented.

In some retrospective studies of patients with cryptococcal meningitis, beginning HAART within 30–60 days of the treatment of meningitis has been associated with a higher risk of paradoxical IRIS and a higher mortality rate [4, 40]. In contrast, a prospective multicentre study did not identify an association between the timing of HAART and the development of IRIS [36]. Finally, a recent prospective study performed in sub-Saharian Africa showed a risk of mortality 3 times greater in HIV-infected patients who had begun ART within 72 h after cryptococcal meningitis diagnosis than in those in whom HAART was delayed for 10 weeks [13].

Development of IRIS seems not to worsen prognosis in patients with CNS opportunistic infections. The mortality rates of patients with and without IRIS were not different in our study. This observation is consistent with findings from other studies [9, 11, 17, 36, 37, 40, 44]. A recent study performed in Uganda found that only four out of 54 HIV-related deaths could be attributed to IRIS in a cohort of patients followed within the first 3 years after ART initiation [44]. It should be noted that three of these four deaths were attributable to CNS infections.

Regarding treatment of IRIS, we cannot draw any conclusions about the benefits of steroids from our study. Three of eight patients who received steroids died while none of the 10 patients who were not treated with steroids died. However, it should be noted that we probably treated patients with more severe clinical manifestations, because we did not have a previous protocol for steroid use in cases of IRIS.

The limitations of our study are those inherent to its retrospective design. However, in our opinion, the results of this study, which covers 11 years after the introduction of HAART, provide interesting information on the prognosis of patients with CNS opportunistic infections in the developed world. Nowadays it is difficult to obtain this kind of data because the number of incident cases is fortunately low. Another limitation is that the results of the study cannot be generalized world-wide, although the data are applicable to the developed world.

In conclusion, despite the reduction in the incidence of CNS opportunistic infections, the prognosis is still poor in many cases. Efforts should be made to improve adherence to HAART and to diagnose HIV infection early in order to avoid the development of new cases of CNS infections.

Acknowledgements

This study was supported by Red Temática de Investigación en SIDA (RIS G03/173-RETIC RD06/0006/0039).

Conflicts of interest: The authors declare that they have no conflicts of interest in relation to this study.

Ancillary