4.1.1. Clinical introduction [2011]

Why is the haemoglobin level important in patients with CKD? Possible adverse effects of anaemia include reduced oxygen utilisation, increased cardiac output and left ventricular hypertrophy, increased progression of CKD, reduced cognition and concentration, reduced libido and reduced immune responsiveness. How much these adverse effects translate into adverse outcomes such as impaired quality of life, increased hospitalisation, increased cardiovascular events and increased cardiovascular and all-cause mortality has been the subject of debate for several years. What is incontrovertible is that since the introduction of human recombinant erythropoietin for treatment of CKD-related anaemia over 2 decades ago we have had the tools to significantly influence anaemia management. The phenotype of the kidney patient with haemoglobin levels between 5–8 g/dL, rendered massively iron over-loaded and virtually un-transplantable as a result of multiple transfusions, has thankfully become unrecognisable. Attention has shifted from treatment of severe anaemia in dialysis patients to prevention of anaemia non-dialysis and to correction of anaemia to higher levels of haemoglobin.

It is well established that haemoglobin levels fall as kidney function declines but there is significant heterogeneity at each level of kidney dysfunction. Although normal values for haemoglobin in the general population differ by gender this has not been addressed in most study designs of anaemia in kidney disease. Observational data suggest that lower haemoglobin values are associated with increased cardiovascular abnormalities/events, increased hospitalisation, increased mortality, increased transfusion requirements and reduced quality of life. Major criticisms though have been the heterogeneity of such studies and the variation in adjustment for confounders. We do not have randomised controlled trials designed to assess the level of haemoglobin at which we should intervene with treatment but we do have treatment dilemmas. We know from clinical practice that not all patients will necessarily benefit from treatment so at what level of haemoglobin should we consider intervention with anaemia treatment? Should this level differ by age, gender or ethnicity? Should we adopt differing strategies dependent on whether patients are non-dialysis or already receiving renal replacement therapy?

The GDG agreed to address the following question: In patients with chronic kidney disease, what haemoglobin (Hb)/haematocrit (Hct) levels are associated with adverse outcomes and what are the effects of a) age b) gender c) ethnicity?

4.1.2. Methodological introduction

A literature search identified longitudinal,^{133,257,336,340} before and after^{127,202,205,285} and cohort^{60,177,178,186} studies, conducted predominantly in haemodialysis patients.

Four studies^{81,144,170,206} had methodological limitations and were excluded from evidence statements.

Notable aspects of the evidence base were:

• No studies were found which specifically addressed the issues of gender and ethnicity and only one study was identified which stratified the study population according to age²⁰⁵.
• Only two studies included populations over 80 years old^133,178.
• Not all studies reported gender and ethnicity of the participants. Some studies included predominantly male^202,285 or predominantly white participants^60,178 or predominantly male and white participants^81,177. One study included a population that was 67% African American¹³³.
• The number of study participants varied greatly, ranging between 7 and over 60,000.

A comprehensive literature search did not identify any studies that were suitable to address the economic aspects, therefore no health economic evidence statements are given.

4.1.3. Methodological introduction [2011]

The GDG noted a change in terminology for the 2011 update concerning predialysis to nondialysis.

A literature search was undertaken to identify papers published from September 2005 onwards. Eight cohort studies^{113,163,171,175,199,255,335,339} in nondialysis, haemodialysis and transplant patients were included. Studies not meeting the inclusion criteria were excluded.

Notable aspects of the evidence base:

• No studies were found which reported the interaction of age, gender and ethnicity with Hb/Hct levels.
  • One study¹⁶³ included only male patients with subgroup analyses for age and ethnicity. The results were only presented on a forest plot and numerical data were not reported.
• The mean age, where reported, ranged from 51 years³³⁹ to 72 years¹⁷¹; one study¹⁷⁵ reported 29% of the included patients were over 75 years.
• The ethnicity of the patients included in the studies comprised mainly of those classified as white. One study¹⁶³ reported patients with higher Hb levels were likely to be ‘white’.

The outcomes considered in the review are: ·

• Left ventricular hypertrophy
• Hospitalisation
• Mortality
• Composite outcome (all cause mortality, stroke and MI)
• Cardiac events
• Quality of life
• Stroke
• Progression of CKD

4.1.4. Evidence statements [2006, updated 2011]

These evidence statements are grouped by outcome measure per sub-population of anaemia patients.

Hospitalisation

Haemodialysis patients

A cohort (n=66,761), with data stratified into increasing Hct levels and compared with an Hct level of 33 to 35% over a 1-year follow-up period⁶⁰ found the following:

Table 4.1Summary data from study⁶⁰ (Level 2+)

View in own window

Hct (%)	<30	30 to 32	33 to 35 (Ref)	36 to 38	≥39
Hb (g/dl)	<10	10–10.7	11 to 11.7 (Ref)	12 to 12.7	≥13
RR of all-cause hospitalisation	1.42	1.21	1	0.78	0.84
RR of hospitalisation from cardiac causes	1.3	1.17	1	0.75	NS
RR of hospitalisation from infections	1.76	1.3	1	0.82	0.62

RR = relative risk; NS = not significant

In a 2.5-year follow-up study¹⁷⁸, participants (n=50,579) were stratified into increasing Hct levels and compared with patients with the arbitrary reference of Hct 34 to 36% (n=22,192), see Tables 4.2 to 4.5.

Table 4.2

Adjusted relative risk of first hospitalisation due to any cardiac disease (Level 2+).

Table 4.3

Adjusted relative risk of first hospitalisation due to specific cardiac diseases (Level 2+).

Table 4.4

Adjusted relative risk of first hospitalisation for patients with cardiac comorbid conditions (n=45,166) (Level 2+).

Table 4.5

Adjusted relative risk of hospitalisation for patients with Hct 37 to 39% without pre-existing cardiac disease (3-year follow-up) (Level 2+).

There were no new relevant studies identified in the rapid update review reporting the outcome hospitalisation.

Mortality

Nondialysis patients

Evidence statements

There is moderate to high quality evidence^163,175,199 to show that:

• low Hb levels [<11 g/dL] compared to high Hb levels [>13 to ≤14 g/dL] are associated with an increased risk of mortality
• low Hb levels [≥11 to ≤12 g/dL] compared to high Hb levels [>13 to ≤14 g/dL] are associated with an increased risk of mortality
• low Hb levels [>12 g/dL] compared to high Hb levels [≥14 g/dL] are not associated with an increased risk of mortality.

There is uncertainty concerning all of the above results.

There is moderate quality evidence^163,175 to show that a decrement in Hb level of 1 g/dL is associated with an increased risk of mortality.

There is moderate quality evidence³³⁵ to show:

• a decrement in Hb level of 1.5 g/dL is associated with an increased risk of mortality in patients with higher Hb levels [>14.5 g/dL] this decrement is associated with a decreased risk of mortality.

There is low quality evidence¹⁷¹ to show that CHD-mortality is associated with lower Hb quintiles when GFR is estimated using the Cockcroft-Gault method. This effect is not evident when GFR is estimated using the MDRD method.

Evidence report

Three studies^{163,175,199,339} reported the risk for mortality associated with low and high haemoglobin levels. Risk of mortality was assessed over follow-up periods ranging from 16 months¹⁹⁹ to 27 months¹⁷⁵, while overall mortality rates ranged from 0.5% [191/27153]¹⁹⁹ to 29% [245/853]¹⁶³. Mortality rates were stratified according to Hb ranges in one study¹⁶³ [<11 g/dL: 39.0% (68/174); 11.1 to 12 g/dL: 34.2% (74/216); 12.1 to 13 g/dL: 24.9% (50/201); >13 g/dL: 20.2% (53/262)].

An emerging trend suggests that lower Hb levels are associated with an increased risk of mortality compared with higher Hb levels. At higher Hb levels, a significant difference was not observed; however, there is some uncertainty concerning the precision of these effects (figure I.1a to figure I.1c, Appendix I:).

Three studies reported the affect of incremental increases in Hb level on the risk of mortality. The overall mortality rates were: 20% [618/3028]¹⁷⁵; 29% [245/853]¹⁶³; 44.6% [748/1678]³³⁵.

In one study¹⁷⁵ an decrement of 10 g/L [1 g/dL] in Hb level was associated with a significantly increased risk of mortality in patients with: eGFR <15 mL/min [RR 0.91 (95% CI 0.84–0.99); eGFR of 15–29 mL/min [RR 0.86 (95% CI 0.81–0.92)]; eGFR of 30–59 mL/min [RR 0.81 (95% CI 0.71–0.92)] (figure 1.2a, Appendix B).

An increment of 10 g/L [1 g/dL] in Hb level was also associated with a decreased risk in mortality in a second study¹⁶³ [HR 0.86 (95% CI 0.78–0.95)] (figure I.2b, Appendix I:).

A third study³³⁵ reported that an increment of 1.5 g/dL in Hb level was associated with a decreased risk in mortality [HR 0.86 (95% CI 0.79–0.94]. This benefit was increased in patients with Hb levels <14.5 g/dL [HR 0.70 (95% CI 0.63–0.78)]. However, in patients with Hb levels >14.5, an increment of 1.5 g/dL in Hb resulted in an increased risk of mortality [HR 1.31 (95% CI 1.09–1.56)] (figure I.2c, Appendix I:).

A single study¹⁷¹ reported the risk of CHD-related mortality for the lowest Hb quintiles [range: 7.6–14.6], as a continuous variable, compared with patients in higher Hb quintiles using different methods of estimating GFR. GFR estimated with the Cockcroft-Gault method reported an overall mortality rate of 11% [179/1639] and the proportion of patients who died within the groups were as follows: lower quintiles: 41% (74/179); other quintiles: 64% (115/179).

GFR estimated with the MDRD method reported an overall mortality rate of 9% [148/1639] and the proportion of patients who died within the groups were as follows: lower quintiles: 53/148; other quintiles: 95/148.

An increased risk in CHD-mortality associated with lower Hb quintiles was observed when GFR was estimated using the Cockcroft-Gault method (figure I.3, Appendix I:).

This study¹⁷¹ also reported that there was no significant difference in CHD-related deaths in patients with the lowest quintiles of Hb and GFR compared with high Hb and GFR in subgroups for men and women; however, these subgroups included both CKD and non-CKD patients so the results are not presented here.

Haemodialysis patients

Data from a cohort (n=66,761) were stratified into increasing Hct levels and compared with an arbitrary Hct level of 33 to 35% over a 1-year follow-up period⁶⁰:

Table 4.6Adjusted relative risks (Level 2+)

View in own window

Hct (%)	<30	30 to 32	33 to 35 (Ref)	36 to 38	≥39
Hb (g/dl)	<10	10–10.7	11 to 11.7 (Ref)	12 to 12.7	≥13
RR of all-cause mortality	1.74	1.25	1	NS	NS
RR of mortality from cardiac cause	1.57	1.25	1	NS	NS
RR mortality from infections	1.92	1.26	1	NS	NS

NS = not significant

In a 3-year follow-up study¹⁷⁸ participants (n=50,579) were stratified into Hct levels and compared with patients with the arbitrary reference of Hct 34 to 36% (n=22,192):

Table 4.7Adjusted relative risk of mortality due to cardiac diseases¹⁷⁸

View in own window

Hct (%)	34 to 36 (Ref)	37 to 39	≥40
Hb (g/dl)	11.3 to 12 (Ref)	12.3–13	≥13.3
Relative risk	1.00	0.92	0.83
95% CI	N/A	0.87 to 0.98	0.74 to 0.93

Table 4.8Adjusted relative risk of all-cause mortality¹⁷⁸

View in own window

Hct (%)	34 to 36 (Ref)	37 to 39	≥40
Hb (g/dl)	11.3 to 12 (Ref)	12.3–13	≥13.3
Relative risk	1.00	0.92	0.86
95% CI	N/A	0.88 to 0.96	0.80 to 0.93

Table 4.9Adjusted relative risk of mortality for patients with Hct 37 to 39% without pre-existing cardiac disease¹⁷⁸

View in own window

	RR	P value
All-cause death	0.69	0.0002
Any cardiac death	0.69	0.0137

In one study¹³³ (n=309), no association was found between any Hct quartile (<33.4%, ≥33.4 to 35.73%, ≥35.74% to 38.55%, and >38.55%) and survival over 18 months. (Level 3)

In a 4-year study³⁴⁰, renal units with more than 87% of patients achieving target Hct ≥33% (Hb ≥11 g/dl) had a lower mortality rate than those with less than 64% of patients achieving target Hct (p<0.0001). A 10% point increase in the fraction of patients with Hct of more than or equal to 33% (Hb ≥11 g/dl) was found to be associated with a 1.5% decrease in mortality (p=0.003). (Level 3)

A retrospective cohort study with 1-year follow-up¹⁸⁶ (n=75,283) found an increase in the age group associated with higher all-cause and cause-specific mortality. Female patients had better outcomes. When compared with white patients, black patients and other ethnic minority patients had lower all-cause and cause-specific mortality. In the same study¹⁸⁶, mortality data were compared with Hct 30 to <33% (Hb 10 to <11 g/dl)¹⁸⁶, see Table 4.10.

Table 4.10

Adjusted relative risks (Level 2+).

Kidney transplant patients

Evidence statement

There is moderate quality evidence³³⁹ showing there is no significant difference in the risk of mortality in kidney transplant patients with low Hb levels [≤10 g/dL] compared with high Hb levels [>10 to >13 g/dL]. There is some uncertainty in the result.

Evidence report

One moderate quality study³³⁹ examined the association between Hb level and mortality in kidney transplant patients.

Overall mortality rate over a median follow-up period of 8.2 years was 20% [251/825]. The proportion of patients who died within each Hb range was as follows: >10 to 11 g/dL: 31% (28/89); >11 to 12 g/dL: 27% (38/138); >12 to 13 g/dL: 30% (50/167); >13 g/dL: 30% (111/373); ≤10 g/dL : 41% (24/58).

There is uncertainty in the precision around the effect to determine whether Hb levels are associated with risk of mortality (figure I.4, Appendix I:).

4.1.5. Health economic methodological introduction [2011]

No economic studies were included in the 2006 guideline. A literature search was undertaken to identify papers published from September 2005 onwards.

One study¹⁷³ was identified that examined the association between haemoglobin level and cost in nondialysis patients with chronic kidney disease aged 65 years or older who were not receiving treatment for anaemia. This was a retrospective cohort analysis with multivariate regression (covariates: age, gender, GFR, diabetes, hypertension, liver cirrhosis, CAD, MI, LVH). Data was derived from a large US managed care database – this limits the applicability of the results to the guideline. Costs included inpatient and outpatient medical claims and pharmacy dispensing claims.

4.1.6. Health economic evidence statements [2011]

Lefebvre and colleagues¹⁷³ reported that, in CKD patients untreated for anaemia, a haemoglobin level <11 g/dL was associated with an additional monthly cost of £320 (CI: £223, £408) compared to a haemoglobin level >11 g/dL. Every 1g/dL decrease in haemoglobin was associated with a £52 increase in cost (CI: £32–£71).

4.1.7. From evidence to recommendations

Data about the outcome of LVH were presented to the GDG¹⁷⁷. Two studies which demonstrated an association between decreasing left ventricular mass and increasing haematocrit levels^127,257 were based on small sample sizes (n=9 and n=11) and the GDG weighed these studies accordingly in their deliberations.

Two studies were appraised that examined the rate of progression of renal failure but these were excluded as underpowered by the GDG^127,257 and hence, no evidence statements were presented for this outcome.

The GDG noted that the greater hospitalisation rate seen in a study based on registry data⁶⁰ could be a reflection of a sicker population and this may be another reason for the lower Hb level. It was also noted that the lowest haematocrit group required double the amount of EPO to reach this level, and as such, these participants may have a reduced health status.

The study by Moreno et al²⁰⁶ was excluded by the GDG because of a highly selected population (excluding both elderly and ill patients) and a lack of intention to treat analysis. The group agreed to increase the grade of one other study¹⁷⁸ from 3 to 2+ as the study participants had been subdivided according to Hct levels and a multivariate analysis of risk had been performed.

The GDG agreed that the evidence supported an association between decreased haematocrit and increased risk of hospitalisation.

The group felt that the evidence presented on mortality from one study⁶⁰ suggested that there was an increase in mortality between Hct <30 to <33% (Hb levels ~ 1 – 11g/dl) when compared with Hct 33 to 36% (Hb ~ 11–12g/dl). It was noted that this range spans the standard levels quoted in many guidelines. The data presented by two studies^186,340 suggest that an Hb of <11g/dl was the threshold below which there was an increased risk of mortality. However, the GDG noted that these studies may not have accounted for confounding factors such as intercurrent illness. The issue was also raised that there might be a reverse causality and that patients requiring high amounts of epoetin may be sicker and hence more likely to require hospitalisation.

One study¹³³ concluded that the haematocrit level was not a predictor of survival and that other markers of morbidity were more important. The data also suggested that confounding factors may be present that were not taken into account, e.g. infection. This possibility was reflected in the study as the haematocrit levels were corrected for albumin. This study also suggested that men and women require different doses of ESA: women appear to need more ESA than men.

Only one study²⁰² was appraised that evaluated haemodynamic parameters but this was excluded for this outcome by the GDG as it was felt to be underpowered (n=7).

Concerning quality of life in haemodialysis patients(n=57)²⁰², a subgroup analysis of those over and under 60 years of age found a significant increase in quality of life scores associated with higher Hb levels in both age groups.

4.1.8. Recommendation and link to evidence [2011]

1. Consider investigating and managing anaemia in people with CKD if:
   • their Hb level falls to 11 g/dL or less (or 10.5 g/dL or less if younger than 2 years) or,
   • they develop symptoms attributable to anaemia (such as tiredness, shortness of breath, lethargy and palpitations). [new 2011]

4.2.1. Clinical introduction

Data from population studies such as NHANES III in the USA and the NEOERICA study in the UK suggest an increasing prevalence of anaemia with decreasing GFR level. A similar relationship between glomerular filtration rate (GFR) and anaemia has also been demonstrated in population cohorts of people with diabetes³¹⁷. Although anaemia is common in people with diabetes it is also commonly unrecognised and undetected³⁰⁰. The prevalence of anaemia in people with diabetes is increased at all levels of renal function in those with increased proteinuria/albuminuria³¹⁸, and it has been suggested that in people with diabetes, anaemia associated with CKD may occur earlier in the evolution of CKD when compared with people without diabetes. In investigating the evidence base, this section seeks to describe the relationship between GFR and haemoglobin levels and provide guidance for clinicians about the threshold level of GFR below which they should suspect that anaemia is associated with CKD.

4.2.2. Methodological introduction

A literature search identified five studies investigating the association between GFR or creatinine clearance (CCr) with Hb/Hct levels in non-diabetic patients^{20,99,129,155,197} and four studies in diabetic patients^{73,88,316,317}.

Notable aspects of the evidence base were:

• Two studies were not limited to patients with CKD^20,129.
• Two studies were conducted in selected patient populations^155,197 and one study⁹⁹ was conducted in children.
• Patient populations in some studies were not stratified to diabetic and non-diabetic patients and where reported, the percentage of diabetics varied from 5%²⁰ to 28%¹⁵⁵ and to 64.4%¹⁹⁷. All patients with CKD were in the untreated predialysis stage, except for one study where some patients received oral iron (26%) and epoetin (12.8%) to treat their anaemia⁹⁹.
• One study was conducted in people with Type 2 diabetes³¹⁶, and one in people with Type 1 and people with Type 2 diabetes³¹⁷.

A comprehensive literature search did not identify any studies that were suitable to address the economic aspects, therefore no health economic evidence statements are given.

4.2.3. Evidence statements

4.2.4. From evidence to recommendations

The comparison of diabetic and non-diabetic populations was based on a clinical perception that the diabetic population was at risk of developing anaemia of CKD at an earlier stage. The GDG felt that this perception had arisen partly because of the selected patient populations in many of the studies, the cross-sectional nature of the studies, and the lack of standardisation of estimates of renal function used in the various studies.

The current clinical perception of the GDG is that although there was a correlation between diabetes and the anaemia of CKD, the prevalence of anaemia in those with diabetes appeared greater than those without at higher levels of GFR. Within whole population studies there were similar mean haemoglobin levels between those with diabetes and those without diabetes across a range of GFRs.

It was agreed that setting a threshold value of eGFR of 60 ml/min/1.73m2 (the boundary between stage 2 and stage 3 CKD) would be of use in helping clinicians decide whether to consider anaemia of CKD as a cause of the anaemia, although there were some concerns about whether the error around a single measurement would make this a suitable recommendation.

It was felt there was some merit in an empirical statement that supported setting an eGFR of <60 ml/min/1.73m2 which should alert a clinician to consider anaemia of CKD as the cause, and that other causes were likely in patients with a eGFR > 60.

4.2.5. Recommendation

2.

An estimated glomerular filtration rate (eGFR) of <60 ml/min/1.73m2 should trigger investigation into whether anaemia is due to CKD. When the eGFR is ≥ 60 ml/min/1.73m2 the anaemia is more likely to be related to other causes. [D]

4.3.1. Clinical introduction

The purpose of the evidence review in this section was to identify the best combination of tests to determine iron status in patients with CKD.

The aim of determining iron status is to identify which patients need iron supplementation, as well as those who do not. Although absolute iron deficiency may occur in patients with chronic kidney disease we more frequently identify what is termed ‘functional iron deficiency’. Although iron stores may seem adequate when measured by conventional indices of iron status, there may be a lack of ‘freely available iron’ for effective erythropoiesis in the bone marrow.

There is a lack of well-accepted gold standard tests for determining iron deficiency in the setting of CKD. While bone marrow iron stores are often regarded as the best indicator of iron status, this is not universally accepted and taking a bone marrow sample is invasive, relatively time consuming and expensive. The frequent coexisting inflammatory or infective problems in patients with CKD can complicate the interpretation of iron status parameters. For example, serum ferritin is a good marker of storage iron and decreases in iron deficiency states. However, it is also an acute phase reactant, which means it is frequently raised in inflammatory conditions, such as CKD, regardless of the iron status. All the available tests of iron status are subject to similar limitations and detailed discussion is beyond the scope of this guideline. The British Committee for Standards in Haematology is producing a document ‘Evaluation of iron status’, which will deal comprehensively with these issues (although not specifically in the setting of CKD). It is accepted that no single parameter can determine iron status.

In patients without CKD normal serum ferritin levels are over 20 μg/l, but in those with CKD a value of 100 μg/l is considered to be the lower limit of normal to allow for the associated mild inflammatory state. The percentage of hypochromic red cells (HRC) directly reflects the number of red blood cells with suboptimal levels of haemoglobin content (<28 g/dl) and may be determined using certain analysers. HRC <2.5% is normal and HRC >10% indicates definite iron deficiency. Measurement must be on a fresh sample (<4 hours after the blood is withdrawn) because of storage artefact. Reticulocyte haemoglobin content (CHr) may also be measured by certain analysers and is derived from the simultaneous measurement of volume and haemoglobin concentration in reticulocytes. Levels indicating functional iron deficiency depend on the analyser used. Transferrin saturation (TSAT) is a derived value and may be calculated from serum iron × 100 ÷ total iron binding capacity; or serum iron (mg/dL) × 70.9 ÷ serum transferrin (mg/dl). Transferrin levels are also influenced by inflammation and nutrition (correlating with serum albumin levels). A TSAT of <20% suggests iron deficiency.

4.3.2. Methodological introduction

A literature search identified studies which addressed the ability of tests to detect iron deficiency^67,93,147 and the ability of tests to predict the response to intravenous iron supplementation in patients with predefined iron parameters receiving epoetin^{96,97,149,154,184,314}.

Of the six studies looking at the response to intravenous iron, five studies predefined the patient population to whom iron was given as being iron deficient (see Table 4.16). In one study³¹⁴ the response to intravenous iron was used to define the prior iron status. No study addressed the issue of loading with iron prior to epoetin administration.

Table 4.16

Definition of detection of iron deficiency.

4.3.3. Evidence statements

4.3.4. From evidence to recommendations

The group compared the tests based on the sensitivity, specificity and receiver operator characteristics. The group did not use the negative or positive predictive values as they were considered sensitive to demographics and epidemiology and therefore not generalisable.

These iron supplementation studies have dealt with iron deficiency or ‘functional iron deficiency’ (where storage iron may be adequate, but iron utilisation in red cell production is defective). The studies have not addressed the issues of whether iron supplementation could be beneficial in patients having erythropoietin even with apparently normal iron status, or when iron supplementation should be stopped because of a risk of iron overload.

Recticulocyte Hb content and the percentage of hypochromic red cells were also discussed. Neither of these tests are widely available and both are currently under a commercial patent. With respect to recticulocyte Hb content, the GDG felt that although this looked like a sensitive test, the cut-off for this test was a Hb content of less than 26pg. This was considered very low as the normal range is reported to be 31–33pg. The GDG noted that the percentage of hypochromic red cells provided the best sensitivity and specificity from a single test.

In general, the GDG noted that tests for serum ferritin and transferrin saturation were the most widely used but that they had poor sensitivity and specificity. The GDG took note, however, that these tests were both cheap and widely available. It was noted that serum ferritin was the only test addressing iron storage while the other tests reviewed in the evidence assessed iron utilisation. The GDG agreed that no single test was adequate to determine iron status. Serum ferritin showed the best correlation with bone marrow iron scores. Iron deficiency should be ascertained by a combination of serum ferritin (storage iron) and tests of iron utilisation (reticulocyte haemoglobin content, percentage of hypochromic red cells, transferrin saturation, ZPP).

4.3.5. Recommendations

3.

Serum ferritin levels may be used to assess iron deficiency in people with CKD. Because serum ferritin is an acute phase reactant and frequently raised in CKD, the diagnostic cut-off value should be interpreted differently to non-CKD patients. [A(DS)]

4.

Iron deficiency anaemia should be:

• diagnosed in people with stage 5 CKD with a ferritin level of less than 100 μg/l
• considered in people with stage 3 and 4 CKD if the ferritin level is less than 100 μg/l. [D(GPP)]

5.

In people with CKD who have serum ferritin levels greater than 100 μg/l, functional iron deficiency (and hence those patients who are most likely to benefit from intravenous iron therapy) should be defined by:

• percentage of hypochromic red cells >6%, where the test is available or
• transferrin saturation <20%, when the measurement of the percentage of hypochromic red cells is unavailable. [B(DS)]

4.4.1. Clinical introduction

Although anaemia in CKD may develop in response to a wide variety of causes, erythropoeitin (EPO) deficiency is the primary cause of renal anaemia. Predominantly produced by peritubular cells in the kidney, EPO is the hormone responsible for maintaining the proliferation and differentiation of erythroid progenitor cells in the bone marrow. Loss of peritubular cells leads to an inappropriately low level of circulating EPO in the face of anaemia (Figure 4.2).

Figure 4.2

Evolution of anaemia in CKD. (Reproduced with kind permission of Dr Anatole Besarab). EPO = erythropoietin; WHO = World Health Organization.

We know that anaemia develops early in the course of chronic kidney disease. NHANES III found lower levels of kidney function to be associated with lower haemoglobin levels and a higher prevalence and severity of anaemia²⁰. The prevalence of anaemia, defined as haemoglobin levels of less than 12 g/dl in men and less than 11 g/dl in women, increased from 1% at an estimated GFR of 60 ml/min per 1.73 m2, to 9 and 33% at estimated GFRs of 30 and 15 ml/min per 1.73 m2 respectively. Using the same definition of anaemia, it is suggested that in people with diabetes and CKD the prevalence of anaemia in stage 2 and 3 CKD is greater than in those without diabetes⁸⁸. In a study of 5,380 participants from the Kidney Early Evaluation Program, 22% of those with CKD stage 3 and diabetes had anaemia, compared with 7.9% of those with stage 3 CKD alone (p<0.001). In stage 2 CKD 7.5% of those with diabetes were anaemic compared with 5.0% of those without diabetes (p=0.015). In people with diabetes the prevalence of anaemia at all levels of GFR is greater with increasing levels of albuminuria³¹⁶.

When patients with diabetes and CKD are stratified into those more likely to be iron-replete (TSAT>16%) and those less likely to be iron-replete (TSAT<16%) anaemia is associated with a relative lack of EPO response in those with TSAT>16%³¹⁵.

In patients with less advanced CKD there may be some uncertainty about whether or not the anaemia is associated with lack of EPO, and this may be particularly so in transplanted patients in whom immunosuppression may also play a role in suppressing the bone marrow response. In these patients, knowledge of serum EPO levels may be beneficial and the evidence review in this section seeks to address this.

4.4.2. Methodological introduction

One cohort study²⁶⁰, six cross-sectional studies^{10,43,85,91,212,315} and two longitudinal studies, prospective⁵⁰ and retrospective⁶⁴, which examined the association between serum erythropoeitin with Hb levels or renal function, were identified in a literature search.

Notable aspects of the evidence base were:

• The studies comprised selected and unselected participants.
• Of the three studies conducted in people with diabetes, the study populations consisted of people with Type 2 diabetes without nephropathy⁶⁴, selected people with Type 1 diabetes with diabetic nephropathy in the absence of advanced renal failure⁴³, people with Type 1 and 2 diabetes³¹⁵.
• Other causes of anaemia were explicitly ruled out in some studies^{43,50,64,91,260}.
• Where reported, anaemia was defined as <13 g/l for men and <11.5 g/l for women⁶⁴, Hb ≤ 11.5 g/dl for women and 12.0 g/dl for men⁴³, Hb <11 g/dl⁹¹, Hb <12 g/dl for women and Hb <13 g/dl for men³¹⁵.

A comprehensive literature search did not identify any studies that were suitable to address the economic aspects, therefore no health economic evidence statements are given.

4.4.3. Evidence statements

Adults with diabetes

In people with Type 2 diabetes without nephropathy (n=62) a significant negative correlation between serum EPO and Hb levels was found (r2=0.612, p=0.01)⁶⁴. (Level 3)

In contrast to the above finding, a study in people with Type 1 diabetes with diabetic nephropathy (in the absence of advanced renal failure) (n=27), found no significant EPO response to lower Hb levels⁴³. (Level 3)

A cross-sectional study conducted in people with diabetes³¹⁵ found no significant EPO response in anaemic patients (defined as Hb <12 g/dl for women and Hb <13 g/dl for men) with GFR >60 ml/min/1.73m2 or >90 ml/min/1.73m2. (Level 3)

In a subgroup of iron replete diabetic patients (transferrin saturation level >16%), from the above study³¹⁵, serum EPO levels did not change significantly with Hb level as shown below.

Table 4.19

Characteristics in anaemia and raised or normal serum EPO (Level 3).

Children with chronic renal failure

No significant correlation was found between serum EPO and Hb/Hct levels in three studies conducted in children with chronic renal failure (n=7¹⁰; n=10⁸⁵; n=37⁵⁰). (Level 3)

Likewise, no significant correlation was found between serum EPO levels and renal function assessed by means of eGFR (n=37)⁵⁰ or serum creatinine (SCr) (n=30)²¹² in children with chronic renal failure. (Level 3)

The results of a study which investigated Hb and serum EPO levels in children with chronic renal failure and healthy children are shown in Table 4.20.

Table 4.20

Hb and serum EPO in children (Level 3).

4.4.4. From evidence to recommendations

Anaemia is associated with increased EPO levels in individuals without evidence of CKD but the anaemia associated with CKD is characterised by a relative lack of EPO response. However, in the clinical situation routine measurement of EPO levels is of limited value in assessing anaemia.

The GDG reached consensus on a threshold GFR of 40 ml/min, below which anaemia is most likely to be of renal aetiology and measurement of erythropoietin levels will not be required except in exceptional circumstances. At GFR levels between 40 and 60 ml/min, the utility of testing is uncertain from the existing evidence, and a research recommendation is given.

4.4.5. Recommendation

6.

Measurement of erythropoietin levels for the diagnosis or management of anaemia should not be routinely considered for people with anaemia of CKD. [D(GPP)]