A randomized controlled trial of two dialysate sodium concentrations in hospitalized hemodialysis patients

ABSTRACT

Background

Several large dialysis organizations have lowered the dialysate sodium concentration (DNa) in an effort to ameliorate hypervolemia. The implications of lower DNa on intra-dialytic hypotension (IDH) during hospitalizations of hemodialysis (HD) patients is unclear.

Methods

In this double-blind, single center, randomized controlled trial (RCT), hospitalized maintenance HD patients were randomized to receive higher (142 mmol/L) or lower (138 mmol/L) DNa for up to six sessions. Blood pressure (BP) was measured in a standardized fashion pre-HD, post-HD and every 15 min during HD. The endpoints were: (i) the average decline in systolic BP (pre-HD minus lowest intra-HD, primary endpoint) and (ii) the proportion of total sessions complicated by IDH (drop of ≥20 mmHg from the pre-HD systolic BP, secondary endpoint).

Results

A total of 139 patients completed the trial, contributing 311 study visits. There were no significant differences in the average systolic blood pressure (SBP) decline between the higher and lower DNa groups (23 ± 16 versus 26 ± 16 mmHg; P = 0.57). The proportion of total sessions complicated by IDH was similar in the higher DNa group, compared with the lower DNa group [54% versus 59%; odds ratio 0.72; 95% confidence interval (95% CI) 0.36–1.44; P = 0.35]. In post hoc analyses adjusting for imbalances in baseline characteristics, higher DNa was associated with 8 mmHg (95% CI 2–13 mmHg) less decline in SBP, compared with lower DNa. Patient symptoms and adverse events were similar between the groups.

Conclusions

In this RCT for hospitalized maintenance of HD patients, we found no difference in the absolute SBP decline between those who received higher versus lower DNa in intention-to-treat analyses. Post hoc adjusted analyses suggested a lower risk of IDH with higher DNa; thus, larger, multi-center studies to confirm these findings are warranted.

Graphical Abstract

Graphical Abstract.

KEY LEARNING POINTS.

What is already known about this subject?

• The optimal dialysate sodium concentration for patients receiving maintenance hemodialysis is not clear. In hospitalized patients the evidence-base is particularly lacking.

What this study adds?

• In this randomized controlled trial, we tested the effect of higher versus lower dialysate sodium on the magnitude of intra-dialytic blood pressure decline.

What impact this may have on practice or policy?

• We did not find significant differences in the primary analyses according to randomized treatment arm. There were no differences in adverse effects. In post hoc adjusted analyses there was a suggestion of less blood pressure decline in patients receiving higher dialysate sodium concentration (DNa) (versus lower), supporting the need for larger, multi-center studies powered for hard clinical outcomes.

INTRODUCTION

Hypervolemia (excess sodium and water) is common among patients receiving maintenance hemodialysis (HD) and is associated with adverse cardiovascular complications and death [1]. However, overly rapid volume removal can be complicated by intra-dialytic hypotension (IDH), which itself is associated with higher morbidity and mortality [2]. The dialysate sodium concentration may influence diffusive sodium flux during HD and thus play an important role in the control of extracellular volume and blood pressure (BP).

During HD, the pattern of decline in BP during HD is not linear [3, 4], with the greatest relative decline occurring in the earlier part of the session. This early BP decline is independent of the ultrafiltration rate, suggesting that non-volume-related factors may be important [4]. We previously reported that higher pre-HD blood urea nitrogen (BUN) and serum osmolality are associated with IDH [5, 6], suggesting that changes in osmolality may play a role in BP changes. Indeed, clinicians have used higher dialysate sodium concentrations for decades in an effort to promote BP stability and improve patient symptoms, but its use has been tempered by concerns of increased thirst, inter-dialytic weight gain and inter-dialytic hypertension [7].

Notwithstanding the lack of large-scale randomized clinical trials powered for hard clinical outcomes, many large dialysis organizations in the USA lowered the ‘standard’ dialysate sodium concentration for outpatient use in recent years [8–10]. The implications of such prescriptions for HD-related BP control in hospitalized patients is unclear. Therefore, we tested if higher versus lower dialysate sodium would result in less IDH in hospitalized patients on maintenance HD.

MATERIALS AND METHODS

Study design and population

The study was designed as a double-blind, single-center, randomized controlled clinical trial. It was registered at ClinicalTrials.gov (NCT02145260). All participants provided written informed consent, and the study was approved by the hospital institutional review board (approval number 2014P000629).

Adult patients on maintenance HD (>90 days) who were hospitalized at Brigham and Women's Hospital and who had received a maximum of one inpatient HD session were eligible for inclusion. Only index admissions were considered. Reasons for exclusion included: use of pressors or midodrine; pre-HD serum sodium ≤128 mmol/L or >145 mmol/L; pre-HD systolic BP >180 mmHg; intensive care unit stay earlier in admission; expected length of stay <24 h; acute coronary syndrome within 7 days; acute stroke; institutionalized individuals; pregnancy; or enrolment in other trials or judgment that participation would be detrimental to the patient (Figure 1). Exclusion criteria were based around safety and optimizing the ability to detect treatment effects with the following rationale. (i) Acute coronary syndrome—given the potential hemodynamic effects of the interventions on cardiac perfusion. (ii) Recent stroke—due to the potential hemodynamic effects on cerebral perfusion, as well as the theoretical potential for exacerbating cerebral edema with lower dialysate sodium concentration (DNa). (iii) Baseline systolic blood pressure (SBP) >180 mmHg—Higher DNa could exacerbate hypertension. (iv) More than one inpatient HD sessions/prior intensive care unit stay—effect sizes may have been minimized if excessive volume had been removed in pre-study sessions.

Figure 1:

Enrollment flow chart. CNS, central nervous system.

Study procedures

The trial consisted of a screening period and a double-blind randomized treatment period. The randomization schedule was created prior to study initiation by the Investigational Drug Service of Brigham and Women's Hospital and concealed from all study staff. Eligible patients were randomized in a 1:1 fashion to receive either dialysate sodium of 142 or 138 mmol/L, up to a maximum of six consecutive inpatient sessions. These concentrations were designed to achieve higher and lower separation from the historical ‘standard’ DNa of 140 mmol/L used at our institution. Patient and study staff blinding was maintained throughout the study—only the treating registered nurse was aware of the dialysate sodium prescription. Detailed data related to demographics, comorbidities and medications were collected at baseline. Hemodynamic, dialysate prescription, weight and blood pressure data (every 15 min) were collected at each study session. The Principal Investigator assumed responsibility for data and safety monitoring, as approved by the hospital institutional review board.

Evaluation of outcomes

The primary endpoint was defined as the intra-dialytic decline in systolic BP (in mmHg), calculated as pre-HD systolic BP minus the lowest intra-dialytic systolic BP. The secondary endpoint was defined as the proportion of sessions complicated by IDH (defined as any drop of ≥20 mmHg from the pre-HD systolic BP). Using data from a prior observational study [11], assuming a difference in mean systolic BP decline of 7 mmHg between the groups and standard deviation of 14 mmHg, a sample size of 128 patients would be required to achieve 80% power with an alpha = 0.05. To allow for potential drop-outs, we aimed to recruit 140 patients.

Patient symptoms were ascertained via two questionnaires, adapted from the hemodialysis study [12], that were administered to study patients at the end of each HD study session. The first was an 18-part yes/no questionnaire that enquired about intra-dialytic symptoms, with a possible range of 0–18 (Supplementary data, Table S1). The second was an 18-part questionnaire that enquired about thirst, scored on a semi-quantitative scale (never/almost never; occasionally; fairly often/often), with a possible range of 3–54 (Supplementary data, Table S2).

Statistical analyses

Data were reported as mean (± standard deviation), median (25th–75th percentile) or frequencies and percentages, as appropriate. The Student's t-test, Wilcoxon Rank Sum, or χ² tests were used to determine differences in baseline variables between randomized treatment arms, according to data distribution.

To account for within-person correlations, the difference in systolic BP decline according to treatment arm was analyzed by fitting a repeated-measures mixed-effects model that used patient ID as a random effect. Initially, unadjusted models were fit according to the intention-to-treat principle. Subsequently, as pre-HD systolic BP is a major determinant of intradialytic decline in systolic BP, and as there were differences in baseline characteristics related to heart failure, sex and pre-dialysis weight, an additional exploratory post hoc model was fit, adjusting for these variables. For the secondary endpoint, the difference in the total proportion of sessions complicated by at least one episode of IDH was assessed using the χ² test. Subsequently, mixed-effects logistic regression models were fit to account for within-person correlations, with sensitivity analyses using mixed-effects Poisson regression. Differences in patient symptoms and thirst were also assessed by fitting repeated-measures mixed effects models, as described above.

All analyses were performed at the nominal alpha level of 0.05, without correction for multiple hypothesis testing or imputation for missing data. Statistical analyses were performed using STATA (version 16.0, Stata Corp., College Station, TX, USA).

RESULTS

Enrollment, randomization and follow-up

From July 2014 to July 2020, we screened 335 patients at Brigham and Women's Hospital, Boston. A total of 144 were randomly assigned to receive either DNa of 138 or 142 mmol/L. Five patients were discharged prior to completing the first study session (Figure 1). No patients were lost to follow-up. The median number of study sessions was 2/patient (25th–75th percentile 1–3 sessions), with no appreciable difference between randomized treatment arms (P-difference = 0.34) or inter-dialytic interval (P-difference = 0.76).

The average age of participants was 60 ± 14 years; 57% were male and 33% were Black. The characteristics of the patients at baseline were broadly similar, although there were notable between-group differences in sex, history of heart failure and pre-HD weight (Table 1). The baseline characteristics of the hemodialysis sessions were similar among patients, according to the randomized treatment arms (Table 2).

Table 1.

Baseline characteristics according to randomized treatment arm

Characteristic	Lower DNa (138 mmol/L) (n = 69)	Higher DNa (142 mmol/L) (n = 70)
Age, years	61 ± 14	58 ± 15
Male, n (%)	43 (62)	36 (51)
Black, n (%)	21 (30)	25 (36)
Diabetes, n (%)	36 (52)	39 (56)
Heart failure, n (%)	29 (42)	22 (31)
Catheter access, n (%)	17 (25)	14 (20)
Session length, min	213 ± 31	208 ± 43
Pre-HD SBP, mmHg	135 ± 24	139 ± 24
Anti-hypertensive use, n (%)	42 (61)	46 (66)
Pre-HD weight, kg	80.4 ± 22.6	73.5 ± 20.9
Serum sodium, mmol/L	137 ± 4	137 ± 3
Serum potassium, mmol/L	4.7 ± 0.7	4.8 ± 0.7
Serum albumin, g/dL	3.2 ± 0.5	3.3 ± 0.5
BUN, mg/dL	49 ± 19	48 ± 16
Hemoglobin, g/dL	9.6 ± 1.5	9.6 ± 1.6

Values for continuous variables are given as mean ± standard deviation; values for categorical variables are given as count (percentage).

Values were missing for 62 patients for pre-HD weight, 89 patients for serum albumin, one patient for serum sodium, one patient for BUN and three patients for hemoglobin.

Table 2.

Baseline hemodialysis treatment characteristics according to randomized treatment arm

Characteristic	Lower DNa (138 mmol/L) (n = 69)	Higher DNa (142 mmol/L) (n = 70)
Dialysate potassium, n (%)
2 mEq/L	31 (45)	30 (44)
3 mEq/L	31 (45)	35 (51)
4 mEq/L	7 (10)	4 (6)
Dialysate calcium, n (%)
2.5 mEq/L	65 (94)	66 (96)
3.0 mEq/L	3 (4)	0 (0)
3.5 mEq/L	1 (1)	3 (4)
Dialysate bicarbonate, n (%)
30 mEq/L	0 (0)	2 (3)
32–33 mEq/L	0 (0)	2 (3)
35 mEq/L	69 (100)	65 (94)
Ultrafiltration volume, L	1.8 ± 1.0	1.8 ± 1.0
Session length, min	213 ± 31	208 ± 43
Blood flow, mL/min	400 (350, 400)	400 (350, 400)
Dialysate flow, mL/min	800 (800, 800)	800 (800, 800)
Dialysate temperature, °C	36.5 ± 0.5	36.5 ± 0.6

Values for continuous variables are given as mean ± standard deviation or median (25th, 75th percentile); values for categorical variables are given as count (percentage).

Percentages may not add to 100 due to rounding.

Values were missing for one patient for dialysate potassium, calcium and bicarbonate; two patients for ultrafiltration volume; one patient for blood flow and dialysate flow; 10 patients for dialysate temperature.

Outcomes

When averaged by patient, the mean pre-HD systolic BP was 138 ± 22 mmHg and mean nadir intra-dialytic systolic BP was 113 ± 21 mmHg. The mean systolic BP decline in those randomized to the higher dialysate sodium arm was 26 ± 16, compared with 23 ± 16 mmHg in those randomized to the lower dialysate sodium arm (P = 0.57; Figure 2).

Figure 2:

Mean SBP according to randomized treatment arm.

The total number of HD sessions complicated by IDH was 82 out of 152 (54%) in those randomized to the higher dialysate sodium arm, compared with 92 out of 157 (59%) in those randomized to the lower dialysate sodium arm. The odds of developing IDH in the higher (versus lower) dialysate sodium arm was 0.72 [95% confidence interval (95% CI) 0.36–1.44].

Post hoc models

In a post hoc multivariable mixed-effects model, adjusting for pre-HD systolic BP, sex and heart failure, there was less SBP decline in those randomized to the higher dialysate sodium arm (−4.4 mmHg; 95% CI −9.1 to 0.3 mmHg). This was further accentuated upon additional adjustment for pre-HD weight (−7.7 mmHg; 95% CI −13.4 to −1.9 mmHg; Table 3).

Table 3.

Dialysate sodium—association with SBP BP decline and IDH

	Difference in SBP decline (95% CI)		Risk of IDH (SBP decline ≥20 mmHg)
	for higher versus lower DNa		for higher versus lower DNab
	Absolute change	Relative change	Odds ratio	Prevalence ratio
	(mmHg)	(%)	(95% CI)	(95% CI)
Unadjusted	−1.5 (−6.6 to 3.6)	−2.0 (−5.3 to 1.4)	0.72 (0.36 to 1.44)	0.92 (0.68 to 1.24)
Model 1	−1.8 (−6.9 to 3.2)	−2.1 (−5.5 to 1.3)	0.71 (0.36 to 1.43)	0.92 (0.68 to 1.23)
Model 2	−3.2 (−7.9 to 1.5)	N/Aa	0.58 (0.25 to 1.36)	0.92 (0.68 to 1.23)
Model 3	−4.4 (−9.1 to 0.3)	−2.9 (−6.3 to 0.6)	0.48 (0.20 to 1.14)	0.86 (0.64 to 1.17)
Model 4	−7.7 (−13.4 to −1.9)	−4.7 (−8.6 to −0.8)	0.37 (0.15 to 0.94)	0.73 (0.48 to 1.12)

Model 1 adjusted for baseline heart failure; Model 2 adjusted for heart failure and pre-HD systolic BP; Model 3 adjusted for heart failure, pre-HD systolic BP and sex; Model 4 adjusted for heart failure, pre-HD systolic BP, sex and pre-HD weight.

^aPre-HD systolic BP was not included in the models for relative change due to collinearity.

^bOdds ratios were estimated using mixed-effects logistic regression and prevalence ratios with mixed-effects Poisson regression.

For the binary outcome of IDH, in a multivariable mixed-effects model, adjusting for pre-HD systolic BP, sex and heart failure, there was a lower odds of developing IDH in those randomized to the higher dialysate sodium arm [odds ratio (OR) 0.48; 95% CI 0.20–1.14]. This was further accentuated upon additional adjustment for the pre-HD weight (OR 0.37; 95% CI 0.15–0.94). Similar, but attenuated, patterns of association were noted in sensitivity analyses reporting prevalence ratios (Table 3).

Patient symptoms

When averaged by patient, the mean HD symptom score was 3.3 ± 2.7 in those randomized to the higher dialysate sodium arm and 3.1 ± 2.4 in those randomized to the lower dialysate sodium arm. The difference was not significant in the unadjusted mixed effects model (P = 0.59).

The mean thirst score was 31 ± 8 in those randomized to the higher dialysate sodium arm and 29 ± 8 in those randomized to the lower dialysate sodium arm. The difference was not significant in the unadjusted mixed effects model (P = 0.30).

Adverse events

After randomization, a total of 37 patients experienced 56 adverse events among those randomized to the lower DNa arm, while 36 patients experienced 53 events in those randomized to the higher DNa arm. The most common events were extremes of blood pressure, cramping and pain. There were no major differences in AEs between the groups (Table 4).

Table 4.

Adverse events according to randomized treatment arm

	Lower DNa	Higher DNa
Adverse event	(n = 69)	(n = 70)
	(No. of participants)	(No. of events)
Hypotension (SBP <90 mmHg)	9 (9)	6 (7)
Hypertension (SBP >180 mmHg)	3 (3)	7 (7)
Chest pain/MI	4 (4)	1 (1)
Tachycardia	1 (1)	1 (1)
Diaphoresis	1 (1)	1 (1)
Dyspnea	0 (0)	1 (1)
Cough	2 (3)	0 (0)
Nausea	3 (3)	3 (4)
Abdo/back pain	3 (3)	3 (3)
Bloating	1 (1)	0 (0)
Diarrhea	0 (0)	1 (1)
Perirectal abscess	1 (1)	0 (0)
Cramping	7 (10)	6 (8)
Altered mental status	2 (2)	2 (2)
Headache	0 (0)	2 (3)
Dizziness	1 (1)	0 (0)
Anxiety	2 (2)	0 (0)
Restless legs	1 (1)	0 (0)
Trembling	0 (0)	1 (1)
Itch	1 (1)	1 (1)
Access dysfunction	2 (2)	2 (3)
Other pain	5 (8)	4 (6)
Eye pain	0 (0)	1 (1)

MI, myocardial infarction.

DISCUSSION

In this double-blind, parallel-group, randomized controlled trial in hospitalized patients receiving maintenance HD, we found no significant difference in the absolute SBP decline between those who received higher, compared with lower, dialysate sodium concentrations. The frequency of patient symptoms and adverse events was similar between the higher and lower dialysate sodium arms.

Patients receiving maintenance HD are at markedly higher risk of all-cause and cardiovascular mortality, compared with age-matched patients not requiring HD [13]. Prior studies have shown that rates of hospitalization for cardiovascular events tend to cluster around the days on which HD occurs, suggesting an association with the HD procedure itself [14–16]. While there are many factors that could contribute to this observation (and participants of our study were already hospitalized), the development of intra-dialytic hypotension has received increasing attention, given its association with development of myocardial stunning [17], progressive decline in contractile function [18] and development of cardiac arrhythmia [19]. Furthermore, IDH is potentially modifiable and is known to be associated with a higher risk of mortality [2], making it an attractive surrogate outcome for interventional studies.

The dialysate sodium prescription has evolved over time, initially being lower than the serum sodium concentration (when treatments were mostly reliant on diffusive clearance), to eventually being higher (∼140 mEq/L; to minimize patients symptoms and promote hemodynamic instability), in the setting of shorter treatment times and hydrostatic-driven ultrafiltration [7, 20]. Indeed, controversy has dogged the choice of dialysate sodium concentration over time, with proponents of lower dialysate sodium highlighting the benefits in terms of reducing thirst, inter-dialytic weight gain and BP, while others argue against widespread use of lower dialysate sodium, pointing to a higher risk of hemodynamic instability [8]. In terms of harder clinical outcomes, some prior studies have reported a lower risk of death and hospitalization with higher dialysate sodium use, particularly in those with lower pre-HD serum sodium concentration [8, 21–23]. However, based on largely observational evidence, many large dialysis organizations have moved to lower the ‘standard’ dialysate sodium to around 137–138 mEq/L, though this suggestion has not been uniformly supported [9, 24]. Indeed, a recent meta-regression of clinical trials specifically highlights the risk of IDH with DNa <136 mmol/L and calls for robust evidence before any further lowering of ‘default’ DNa prescriptions [8]. A large pragmatic trial in outpatients, powered to detect differences in mortality, is currently underway (NCT02823821).

Our study is important, as it examines the effect of higher versus lower dialysate sodium in a hospitalized patient group, who are potentially at higher risk for hemodynamic instability. The average ultrafiltration volume and pre-HD systolic BP were lower in our trial, compared with prior studies of outpatients [4, 25], which may have contributed to less decline in systolic BP and a lower frequency of IDH events (e.g., 56% overall IDH events, compared with 68% in an outpatient cohort) [2]. This lower than expected intra-dialytic decline in systolic BP may have impacted our ability to detect significant differences between randomized arms, which was further limited by some baseline imbalances in patient characteristics. For example, lower body weight has been more frequently observed among patients with IDH in contemporary US hemodialysis cohorts, perhaps reflecting nutritional deficiency or even higher ultrafiltration rates [2]. Conversely, pre-HD systolic BP was marginally higher in our higher DNa group, which may bias toward less frequent IDH, as reported in a prior pilot trial of different pre-HD BP targets [26]. Notably, despite not achieving statistical significance, all effect estimates in our trial were in favor of less hemodynamic instability with the higher dialysate sodium arm. Indeed, in post hoc analyses that adjusted for imbalances in baseline characteristics, higher dialysate sodium was associated with 8 mmHg less decline in SBP and 63% lower odds of IDH, compared with lower dialysate sodium. While these results should only be considered exploratory, as the adjustments were not pre-specified, the effect estimates are all consistent with a lower risk of IDH with higher DNa. Furthermore, they provide additional support and rationale for performance of a larger trial in hospitalized HD patients that is powered for hard clinical outcomes.

To our knowledge, this is the largest inpatient trial testing the effects of higher versus lower dialysate sodium. We were able to recruit and randomize a diverse population in a large tertiary referral center, without evidence for an increased risk of patient symptoms or adverse events in either arm. However, this trial has some limitations. We randomized hospitalized patients receiving maintenance HD according to strict inclusion/exclusion criteria over a short duration, which limits generalizability. By virtue of their hospitalization status and potential predisposition to hemodynamic instability, estimated dry weights of patients were often dynamic, which may have added variability to the risk of IDH. Furthermore, not all patients were able to be weighed accurately pre- and post-HD, which may have biased results from adjusted models. Questionnaires were not developed specifically for use on a per-session basis and in the inpatient setting, where waxing and waning mental status in some patients could also contribute to inaccuracies. Limitations in access to outpatient records precluded the calculation of dialysis vintage, predisposition to IDH or comparisons of inter-dialytic weight gain prior to admission. Smaller than expected effect sizes, along with some baseline imbalances in randomized arms, contributed to the trial being underpowered to detect significant differences in the intention-to-treat approach. In this respect, whether the primary and secondary definitions of IDH considered in this trial are appropriate to the inpatient setting is not clear. Future approaches should consider examination of more continuous measurements of BP and protocolization of other aspects of the dialysate prescription, such as timing of anti-hypertensive medications, use of fixed dialysate calcium and lowering of the dialysate temperature, as well as incorporation of objective measures of volume status.

CONCLUSION

In conclusion, the use of higher dialysate sodium concentration in hospitalized patients receiving maintenance HD was safe, compared with lower dialysate sodium, but did not result in a meaningful difference in the magnitude of IDH. The suggestion of benefit in post hoc adjusted analyses provides rationale for performance of a large-scale, event-driven trial, powered to detect differences in hard clinical outcomes such as 30-day readmissions and mortality.

FUNDING

This study was funded through National Institutes of Health grant K23DK102511 to F.R.M.C.

AUTHORS’ CONTRIBUTIONS

F.R.M.C. and S.S.W. contributed in research idea and study design. F.R.M.C., K.S.R., Z.A.K., S.A.P.S., A.T.S., K.A.C. and S.C. contributed in data acquisition. Data analysis/interpretation was done by F.R.M.C. and S.S.W. Supervision or mentorship was done by S.S.W. Each author contributed in important intellectual content during manuscript drafting or revision, and accepts accountability for the overall work by ensuring that questions pertaining to the accuracy or integrity of any portion of the work are appropriately investigated and resolved.

CONFLICT OF INTEREST STATEMENT

F.R.M.C. Causland reports funding from National Institute of Diabetes and Digestive and Kidney Diseases grants U01DK096189, R03DK122240 and K23DK102511 and a grant to his institution from Advanced Instruments. S.S.W. reports personal fees from Public Health Advocacy Institute, personal fees from CVS, personal fees from Roth Capital Partners, personal fees from Kantum Pharma, personal fees from Mallinckrodt, personal fees from Wolters Kluwer, personal fees from GE Health Care, personal fees from GSK, grants and personal fees from Allena Pharmaceuticals, personal fees from Mass Medical International, personal fees from Barron and Budd (versus Fresenius), personal fees from JNJ, personal fees from Venbio, personal fees from Strataca, personal fees from Takeda, personal fees from Cerus, personal fees from Pfizer, personal fees from Bunch and James, personal fees from Harvard Clinical Research Institute (aka Baim), personal fees from Oxidien, personal fees from Sironax, personal fees from Metro Biotechnology, personal fees from Biomarin, personal fees from Bain and personal fees from Regeneron, outside the submitted work. All other authors report no relevant conflicts of interest.