A significant UK contribution to renal anaemia management

 E J Will

 Introduction

Clinical decision support using CCL mini-computer software made an important UK contribution to the modern management of renal anaemia, but that has scarcely been acknowledged. Most of the development of clinical computing was in regional renal units after 1981 (Liverpool, Nottingham, Leeds, Glasgow and Bristol, in particular). That experience may have been less appreciated in metropolitan and academic centres, some of whom were preoccupied with attempting to develop their own clinical software.  As detailed in the report of the 2016 BRCG Clinical Computing seminar, most of the regional computing was concerned with intra-unit patient and data management and there was little attempt at formal, wider publication in the specialist literature. Journals anyway expected formal  scientific studies and usually rejected reports of technological advance in the medical specialties. Perhaps, as a result, the evidence of active research and development in IT was inapparent in the conventional sites of specialty communication.

The UK predicament

The management of renal anaemia in the typically large UK renal units was always a significant logistical problem, often solved by delegation to specialist nurses or pharmacists, with or without formal guidance. From 1998 onwards the UK Renal Registry (UKRR) reflected those efforts by compiling and displaying point prevalent data distributions of relevant clinical variables. The presentation of whole unit data offered a new, aggregated expression of local clinical management and outcomes.   That activity was resulting in a wide range of outcomes, for example in haemoglobin values (Hb). Data turned out to be normally distributed, regardless of absolute values, with characteristic ranges of dispersion (Standard Deviation, SD). The UKRR then began to present  Rose-Day derived graphics, which flagged precisely the necessary shift in each unit outcome distribution for compliance with clinical practice guidelines (see figures 1 and 2).

FIGURE 1. Rose-Day plots of Model Normal Distributions. Notional Mean Hb vs. Per cent over 10g/dl. Standard Deviation set at 1.0, 1.5, 2.0 (yellow).

The Rose-Day sigmoid curves were usefully linear over the intermediate range of inadequate compliance, especially shown at the time for Hb and Urea Reduction Ratios (URR). (As national performance improved to high levels of compliance, the range of unit SDs created a scatter of data points at the top of the sigmoid curve that was no longer instructive).

Figure 2. Rose-Day plot of URR >65% showing the pattern of linear and scattered data from each renal unit

When units recognised  underperformance against minimum Hb values recommended by guidelines, they typically adopted piecemeal incremental dosing, using regimens with expensive Erythropoietin (EPO). The suitable scale and frequency of iron supplements was uncertain at that time.

A novel UK solution – Computerised Decision Support

To rationalise this clinical area a wide-ranging computer project (dubbed ICEBIRG – not an acronym, but a reminder of the theory below the surface of the algorithms) was established at St James’s University Hospital, Leeds (SJUH) to develop medical decision support for the cohort management of renal anaemia. The large, multi-satellite, Haemodialysis population at SJUH was under very uniform, digitally monitored, clinical supervision. The project was based on a novel description of clinical management, designed to reproduce typical, guideline compliant, outcome data distributions.^2,8,9 A series of calibrating studies were necessary, exploring the clinical intervention at predefined threshold and ceiling  values of Hb and serum Ferritin , involving both Haemodialysis and Peritoneal dialysis patients. ^1,4-6 The intervention was explored as a fixed part of management (PRO-active) or in response to particular values (RE-active). Threshold and ceiling values for clinical intervention could be identified to shape predictable and reproducible outcome distributions.^4,5 This approach was in contrast to  the clinical methodology of ‘targeting’, that depended on predefined, desirable outcome values of laboratory variables .  The word ‘target’ is often placed in inverted commas, as a compliment to the uncertainty of its meaning in clinical methodology. That meaning depends on the preposition attached to the idea of aiming  (aim AT, aim TO ACHIEVE, or aim FOR).⁸

For such reasons, there were several counter-intuitive findings to overcome to allow the fashioning of desirable  outcome distributions. An essentially reactive delay of correction, waiting until low values had been allowed to occur, was shown to be a recipe for underachievement. If a reactive methodology was preferred then over-provision was preferable, using raised pre-emptive thresholds of intervention. Established trends (downward for Hb) and everyday technical inefficiencies had to be balanced by a margin of over-aspiration. Studies clearly demonstrated the sparing of EPO by iron supplements when outcome Hb was stabilised, and a demand for iron could be reliably detected by current laboratory assays.⁶ Regular Proactive dosing was another means of obviating the underperformance of reactive clinical management. CRP was measured routinely to detect inflammation that might cause ESA resistance.

It became possible to manipulate the values of Hb and iron indices, stabilising the values in the patient cohorts at designated levels. That potential for sustained control  allowed the system to be deployed in several RCTs in haemodialysis treated populations, which were otherwise  managed uniformly.^11,18 It was interesting that the dispersion (SD) of point prevalent outcome Hbs did not change significantly using computer decision support.

The construction and application of the theory were  published in peer-reviewed,   journals, and a summary article reported the accumulated research experience.²¹   The approach was more  productive than several attempts at decision support in the USA and Canada, and was widely presented.^10,14,16,17

The computerised routines

Automatically downloaded laboratory data were treated algorithmically to offer advice on doses of erythropoiesis-stimulating agents (ESA) and Iron for the entire population of haemodialysis and peritoneal dialysis patients, considering previous and current laboratory results, and prescribed therapy. This could be achieved individually or in a batch mode, typically monthly, with immediate printing of the suggested prescriptions. The system could be over-ruled or modified (personalised) for each patient, in terms of preferred Hb and medication. Satellite dialysis populations (typically 40 patients each) were managed according to the staggered timing of their local routines of data sampling and pharmacy supply.  Clinical information was fed in through weekly MDTs, focussed serially on the main and satellite units. Quick and flexible graphics routinely displayed trends and intercurrent changes in the MDT, prompted by standardised exception reports of clinical variables in the cohort being reviewed. Nursing staff acted in an extended role in daily supervision of the satellites.

The consequences

There were several practical consequences, both local and remote. It was apparent that patients who had previously been overlooked, after failed OPD attendance or for other reasons, could be readily identified and brought back into anaemia management. It was not a trivial benefit in such a large UK renal unit to achieve such complete ascertainment of the patient population against a master index. Secondly, because of the predictability of Hb values under decision support, any patients thus managed showing anaemia or unanticipated trends in ESA/Iron dosing came very quickly to attention and the causes could be explored and remedied.³ The anomalies fell out, as it were ‘netted’, from among the typical responders. It was also useful in unit administration that a predictable, precise costing for the ESA and iron components of unit budgeting became possible.

The system was effective in the other Leeds renal unit (at Leeds General Infirmary) and York, with a variety of routines and ESAs.^10,14,16 NICE subsequently recommended an algorithmic threshold system.

Software copies were solicited by Nottingham, Newcastle and Oxford but that did not produce a useful enlargement of the patient population under study.

The anaemia management system was modified subsequently at Leeds, using different predictive algorithms within the same operational infrastructure. That study showed the benefit of variably extended EPO dosing intervals, in promoting a reduced frequency of prescription, and permitted a further RCT with stabilised anaemia management.^24,25

Attempted generalisation and subsequent withdrawal

In order to establish a wide research network, allowing further study of the value of decision support of renal anaemia management in the UK and Europe, a project was developed with Leeds University, funded by Amgen. The writing of transferable computer software (AMIE) was well in hand when the project was torpedoed by international reaction to the CHOIR study, which compared the consequences of higher than usual (normalised) and more routine Hb outcomes. The study, in US CKD patients, was published in the New England Journal of Medicine in insufficient detail,  but achieved celebrity  status at the annual ASN meeting (2006-7).²⁶  Its conclusion cast doubt on the safety of Hb values over 12g/dl, which led to the withdrawal  of  funding to projects that might possibly, intended or not, have resulted in such values. This response seems to have been accepted with less than usual criticism in the USA, perhaps because it was timely in reducing the trend to progressive over-prescription of ESA in their many haemodialysis centres. European agencies and NICE followed suit, without mounting their own critical examination of the work. The subsequent ERA renal anaemia guideline at least acknowledged that it was only intended Hb values >12g/dl that had become suspect.

It was unsurprising that CHOIR was shown later to be flawed when examined in greater detail. Several other two-level comparative Hb ‘normalisation’ studies were also unconvincing, for various reasons discussed elsewhere.²⁷  All these studies fell short of their designated higher Hb objectives, using target-based methodologies. A plausible basis for pursuing incomplete correction of Hb in renal patients has not been developed. In none of the studies were the achieved Hb values related to any harms. The apparent risks associated with very high ESA doses seem likely to have depended on the clinical need for ESA itself. Resistance to ESA is well known in association with infection and other pathology, but  does not always have an identifiable cause.  There are however plausible explanations for the results of the RCTs. ²⁷ Subsequently, only the Besarab study of haemodialysis patients, vulnerable by virtue of  significant  cardiac disease, has been used convincingly to justify conservative dosing with ESAs.²⁸  Even that study was found methodologically insecure.²⁹

A phoenix?

The theoretical principles of the SJUH work became of value again through their adoption as the clinical methodology of PIVOTAL (2019), a study  of the clinical consequences of comparative iron dosing in a UK multi-centre setting.³⁰ The  epidemiology of iron replacement strategies had been impossible to study with the early decision support system, given the limited size of the patient cohorts. In PIVOTAL there was a deliberate change of the Se Ferritin ceiling and novel dosing algorithms were used, which rather disguised the origin in threshold/ceiling theory. However, that was exposed by the  borrowing of the semantics of  PRO- and RE-active clinical intervention. The slightly different, planned, constraints in PIVOTAL obviated comparison with any previous work. The usual limiting values for Se Ferritin in the past were 500 or 800ug/l, rather than the  unevidenced 700ug/l used in PIVOTAL. Haemoglobin outcome values were not controlled in PIVOTAL. A lack of professional insight in the obscured clinical methodology may  have diminished the subsequent applicability and understanding of the study. ³¹

The methodological contribution of the UK was unfortunately not referenced. Only two UK references were given in the main, sponsor-supported, PIVOTAL report, both from the organiser, one of them referring just to the design of the study. It may be relevant to indicate that journal publication is not only an inevitable marketing exercise in ideas, but also of the employed materiel.³² It is unclear why the PIVOTAL organisers did not recognise the UK origin of the model, which might have embellished its reputation, and improved research funding opportunities.

Other specialty applications of threshold/ceiling theory

The basic threshold approach is applicable beyond renal anaemia, as part of a more sophisticated management of whole unit results and RCTs.^15,22,23 It allows calibration of other treatments and variables.¹³ For example,  the dosing advised for a modern calcium mimetic has been determined recently by declaring a ceiling and minimum value of a laboratory variable, iPTH.³³ When any controlled variable, in that case iPTH, has a tendency to rise, a uniform ceiling value makes for the easier clinical management of dosing.

The need for a systematic margin of over-aspiration in targeted activities is manifest in the recommendations for treatment dose (Kt/V urea) in haemodialysis by KDOQI.³⁴ The National Collaborative Dialysis Study demonstrated that it was necessary to ‘aim at’ 1.4 in order to reliably achieve 1.2. The latter value is now called a minimum, although studies continue to demonstrate underachievement.³⁵

The unheralded discontinuation of Rose-Day plotting by the UKRR in 2011, when improved national Hb and URR performance had moved data beyond the linear mid-section of the graphics, discarded the clarity of calibration offered to clinical guideline compliance for other clinical variables.²²

Comment

It is unclear why the threshold-based work has been rather ignored. It was perhaps better understood by pharma representatives than clinicians. Certainly, US clinicians are most familiar, and reluctant to part, with  ‘targeting’ methodologies. There was some concern in the UK that controlling overall unit results might in some way prejudice individual patient management, strengthened no doubt by the later mantras of ‘personalised care’.  That was despite the evidence that the system offered a uniform and fully ascertained platform from which to change dosing for individual patients. It was unfortunate that micro-management of renal anaemia by government payers was attempted at one point in the USA, reinforcing  clinical concern to avoid a possible central NHS control of clinical management (and cost?) based on IT decision support.

It seems likely that many units run semi-structured dosing systems for ESA and iron. Computerisation is not a prerequisite. This was suggested by the readily performed shift in achieved Hbs recorded by the UKRR, when  the  anaemia practice guideline moved from single (>100g/l) to dual  limits (>100<120g/l). The greatest temptation is to use those limits as the thresholds for clinical dose changes, which risks under and over achievement of Hb respectively. It can add to the interest of clinical management to explore the most effective levels of threshold values in each renal unit. In fact, clinical practice guidelines for renal anaemia have come to recommend intervention at some pre-threshold values.

In conventional target-based methodology any declared target value, say of blood pressure, typically returns as the median of the outcome distribution, or a systematic 50% failure. The CHOIR authors later asserted their lower Hb group ‘target’ was an unconventional 11.3g/dl, but it seems more likely that the value was simply the outcome of a late-adopted, structured dosing algorithm. Such algorithms are very difficult to focus predictably on particular outcome values, which calls the published accounts of CHOIR into doubt.

The incentives to develop threshold software, first prompted by whole unit UKRR data displays, were obviated by the reaction to CHOIR and similar studies. However, the fundamentals of human biology suggest that the correction of renal anaemia would come to depend largely on common values. The cyclical excursion in the specialty over several decades from low (>100g/l) to high(<120g/l) Hb guidelines has indeed returned to the limits that were thought desirable in the 1990s. The fashion  of ‘personalised’ clinical management is perhaps best seen as creating the exceptions that confirm the uniformity of practical human biology. As pithily declared by Donald Richardson, the best managed patient is in the best outcome group.

It seems that the technologies of clinical methodology are so pervasive, like IT, that they have not been seen as a significant dimension for research and development.²² Clinical computing and the reporting of the UKRR provoked an interest in a field that is arguably not yet mature in the clinical specialty of nephrology. Currently, it is clinical practice guideline controversies that provide the forum for that maturation.

 

 

REFERENCES

Icebirg Publications (St James’s University Hospital) – in date order

 

1. Jones CH, Richardson D, Ayers S et al. Percentage hypochromic red cells and the response to intravenous iron therapy in anaemic haemodialysis patients. Nephrol Dial Transplant 1998;13(11):2873‑2876.

 

2. Will EJ. Target practice: Editorial.Int J Artif Org 1998;21(8):433‑436).

 

3. Richardson D, Bartlett C, Goutcher E et al. Erythropoietin resistance due to dialysate chloramine: The two‑way traffic of solutes in haemodialysis. Nephrol Dial Transplant 1999;14(11):2625‑2627.

 

4. Richardson D. Bartlett C. Will EJ. Intervention thresholds and ceilings can determine the haemoglobin outcome distribution in a haemodialysis population. Nephrol Dial Transplant 2000;15(12):2007‑2013.

 

5. Richardson D. Bartlett C. Jolly H.  Will EJ. Intravenous iron for CAPD populations: Proactive or reactive strategies? Nephrol Dial Transplant 2001;16(1):115‑119.

 

6. Richardson D. Bartlett C. Will EJ. Optimizing erythropoietin therapy in hemodialysis patients.

Amer J Kid Dis 2001;38(1):109‑117.

 

7. Richardson D, Lindley E, Bartlett C, Will EJ. Biocompatibility and Erythropoiesis: A Randomised Controlled, Single Center Study of Modified Cellulose and Polysulphone Dialysers in a large Hemodialysis Cohort (n=177). J Am Soc Nephrol 2001;12: 240A

 

8. Will E. Targets and targeting: Opinion. Amer J Kid Dis 2001;38(2):411‑414.

 

9. Will EJ. Aiming at Averages: Issue of the Day. J Roy Soc Med 2001;94:617-19.

 

10. Richardson D, Bartlett C, Fullerton L, Bebb C, Will EJ. Dissemination of an anaemia algorithm into a second large haemodialysis (HD) population. Nephrol Dial Transplant 2002;17 (Abstract Suppl 1):230 (T147).

 

11. Richardson D, Lindley EJ, Bartlett C, Will EJ. A randomized, controlled study of the consequences of hemodialysis membrane composition on erythropoietic response. Amer J Kid Dis 2003;42 (3): 551-560.

 

12. Will EJ, Bartlett C, Richardson D. Anemia outcomes in ESRD: Individual or Facility Intervention? Amer J Kid Dis 2003;1123-1124.

 

13. Garthwaite EA, Will EJ, Bartlett C, Richardson D, Newstead CG. Patient-specific prompts in the cholesterol management of renal transplant outpatients: results and analysis of underperformance. Transplantation 2004;78 (7):1042-7.

 

14. Richardson D, Ridley L, Worth DP, Jones CH. A prospective, cohort study of conversion from subcutaneous to intravenous administration of epoetin-a in an iron replete haemodialysis population. Poster presentation at the Renal Association, Aberdeen. April 2004

 

15. Will EJ. Guideline Targets and Thresholds in Hypertension. Rapid response to BHS hypertension guidelines. Br Med J 13 March 2004.

 

16. Tolman C, Richardson D, Bartlett C, Will EJ. Application of computer assisted anaemia management algorithms in haemodialysis patients produces predictable haemoglobin outcomes regardless of the erythropoietic agent or frequency of administration: results of a randomised study. Poster presentation at ERA, Lisbon May 2004. Nephrol Dial Transplant 2004;19 (Suppl 5).

 

17. Tolman C, Richardson D, Bartlett C, Will EJ. A randomised study of weekly subcutaneous Aranesp and Neorecormon in a large, unselected haemodialysis cohort, managed with computer assisted anaemia algorithms. Oral presentation at ERA, Lisbon May 2004. Nephrol Dial Transplant 2004;19 (Suppl 5).

 

18. Tolman C, Richardson D, Bartlett C and Will E. Structured conversion from thrice weekly to weekly erythropoietic regimens using a computerized decision-support system: a randomised clinical study.  J Am Soc Nephrol 2005;16 (5):1463-1470.

 

19. Will EJ. Glucose control: goals, targets and thresholds. 28 April 2006.Response to Watkinson P, Barber VS, Young JD. Editorial: Strict glucose control in the critically ill. Br Med J 2006;332:865–6.

 

20. Robert M West, Katie Harris, Mark S Gilthorpe, Cae Tolman, and Eric J Will. Functional data analysis applied to a randomized controlled trial in hemodialysis patients describes the variability of patient responses in the control of renal anemia. J Am Soc Nephrol 2007;18:2371-2376.

 

21. Will EJ, Richardson D, Tolman C and Bartlett C. Development and exploitation of a clinical decision support system for the management of renal anaemia. Nephrol Dial Transplant 2007; 22 Suppl 4:iv31-iv36.

 

22. Will E. Editorial Review: Intention and Outcome in guideline-based nephrological practice: a suitable space for ‘clinical technology’. Nephrol Dial Transplant 2007;22(11):3110-3114.

 

23. Will EJ. Letter: Clinical Routines in Practice. Br Med J 2008;337(a3102).

 

Other references

 

24. Lines SW, Lindley EJ, Tattersall JE and Wright MJ. A predictive algorithm for the management of anaemia in

haemodialysis patients based on ESA pharmacodynamics: better results for less work. Nephrol Dial Transplant 2012;27:2425–2429. doi: 10.1093/ndt/gfr706

 

25. Lines SW, Carter AM, Dunn EJ, Lindley EJ, Tattersall JE, Wright MJ. A randomized controlled trial evaluating the erythropoiesis stimulating agent sparing potential of a vitamin E-bonded polysulfone dialysis membrane.

Nephrol Dial Transplant. 2014;29(3):649-56. doi:10.1093/ndt/gft481.

 

26. Singh AK, Szczech L, Tang KL, et al. Correction of anemia with epoetin alfa in chronic kidney disease.

N Eng J Med 2006;335:2085–2098.

 

27. Parfrey PS. Should Hemoglobin Targets for Anemic Patients with Chronic Kidney Disease Be Changed? Am J Nephrol 2010;31:565–566. doi:10.1159/000313894

 

28. Besarab A, Kline Bolton W, Brown JK et al. The effects of normal as compared with low hematocrit values in patients with cardiac disease who are receiving hemodialysis and epoetin. N Engl J Med 1998;339:584-90.

 

29. Fishbane S, Wish J. A physician’s persistence uncovers problems in a key nephrology study. Kidney Intl 2012;82(2) :135-. doi:10.1038/ki.2012.122

 

30. Macdougall IC, White C, Anker SD et al. Intravenous Iron in Patients Undergoing Maintenance Hemodialysis

N Engl J Med 2019;380:447-58. doi:10.1056/NEJMoa1810742

 

31. Bhandari S. Re-evaluating national screening for chronic kidney disease in the UK . Rapid Response. BMJ2023;382:e07426532.

 

32. Will EJ. Caveats for scientific publication in the modern market place.

Clin J Am Soc Nephrol 2009;4:1693–1695. doi:10.2215/CJN.06460909

 

33. Koiwa F, Tokunaga S, Asada S et al. Efficacy of Evocalcet in Previously Cinacalcet-Treated Secondary Hyperparathyroidism Patients. Kidney Int Rep. 2021;6(11):2830-2839. doi:10.1016/j.ekir.2021.08.020.

 

34. KDOQI Clinical Practice Guideline for Hemodialysis Adequacy:2015 update. Am.J.Kidney Dis 2015;66:884-930. doi.org/10.1053/j.ajkd.2015.07.015

 

35. Jeon J, Kim GO, Kim BY et al. Effects of Kt/V urea on outcomes according to age in patients on maintenance hemodialysis. Clinical Kidney Journal 2024;17(5) doi.org/10.1093/ckj/sfae116

 

 

 

Last Updated on June 23, 2024 by John Feehally