Absolute counts, production indices, and the surprisingly difficult problem of grading the marrow response to anemia

A patient has:

Hemoglobin: 3 g/dL
Hematocrit: 14%
MCV: 57 fL
Ferritin: 5 ng/mL
Absolute reticulocyte count: 180 × 10⁹/L

Is the marrow responding appropriately?

For years, I thought this was an easy question.

I have long favored the absolute reticulocyte count, or ARC, over the corrected reticulocyte percentage and reticulocyte production index. My operational threshold has been approximately 120 × 10⁹/L:

ARC >120 × 10⁹/L: appropriate response
ARC <120 × 10⁹/L: inappropriate response

In an earlier Blood Project article, I was even more categorical:

“There are no grades here, just pass-fail.”

If the ARC was above 120 × 10⁹/L, I argued, call it a day.

I have reconsidered that wording.

Not because I have decided that the reticulocyte production index, or RPI, is the better test. I have not.

Rather, this case forced me to recognize that I had conflated two questions:

1. Is there a regenerative erythroid response?
2. Is the magnitude of that response quantitatively appropriate for the severity of the anemia?

An ARC of 180 × 10⁹/L clearly places this patient on the regenerative side of my usual operational threshold.

But when I entered the same data into an RPI calculator, the RPI was below 2, which by conventional criteria is interpreted as an inadequate marrow response.

So the same patient produced two different classifications:

ARC: regenerative

RPI: inadequate

That raised a more fundamental question: are these two measures actually asking the same thing, and if they disagree, which one should we trust?

That question sent me back to the equations, to the original physiology underlying the RPI, and ultimately to a more basic issue:

What exactly are we measuring when we count reticulocytes?

---

What are we actually measuring?

Five Ways to Express the Reticulocyte Response

A mature red blood cell circulates for approximately 120 days. Most of the circulating RBC population therefore represents erythropoiesis that occurred weeks or months earlier.

Reticulocytes provide a much more immediate window into erythroid activity.

A reticulocyte is formed when an orthochromatic erythroblast ejects its nucleus. It retains residual RNA, ribosomes, mitochondria, and other cellular material that disappears as the cell matures. Under steady-state conditions, reticulocytes spend several days maturing in the marrow and approximately another day maturing in the circulation.

That short time window is what makes the reticulocyte count clinically valuable.

The reticulocyte response can be expressed or interpreted in several ways.

Four traditional measures are commonly encountered; a fifth, the bone marrow responsiveness index (BMRI), provides an interesting example of an attempt to severity-adjust the ARC in a specific disease setting.

• Reticulocyte percentage: the raw ratio
• Corrected reticulocyte percentage, or reticulocyte index: a derived adjustment
• Absolute reticulocyte count: the actual circulating concentration
• Reticulocyte production index: a derived model
• Bone marrow responsiveness index (BMRI): a different derived model

They are related mathematically, but they are not identical physiologically.

Reticulocyte percentage: the denominator problem

The crude reticulocyte percentage is:$$Reticulocyte\ \% = \frac{reticulocytes}{total\ RBCs}\times100$$

The denominator may consist of as few as 100 RBCs counted manually on a stained preparation or many thousands of RBCs analyzed by an automated method; either way, the result is the fraction of examined RBCs that are reticulocytes.

Suppose the absolute number of circulating reticulocytes remains unchanged, but the total RBC count falls by half. Because reticulocyte percentage is the number of reticulocytes divided by the total number of RBCs, the reticulocyte percentage approximately doubles, even though there are no more reticulocytes in the blood.

This is why the reticulocyte percentage can be misleading in anemia. As the total RBC count falls, the same absolute number of circulating reticulocytes represents an increasingly large percentage of the remaining RBC population. The more severe the anemia, the greater the potential for this denominator effect to exaggerate the apparent reticulocyte response.

The neutrophil analogy

The problem with percentages becomes obvious if we consider neutrophils.

Consider two patients:

Patient A:
WBC = 0.1 × 10⁹/L
Neutrophils = 99%

Patient B:
WBC = 200 × 10⁹/L
Neutrophils = 2%

If we looked only at percentages, Patient A would appear to have an abundance of neutrophils and Patient B almost none.

But calculate the absolute neutrophil count:

Patient A:
ANC = 0.1 × 0.99 = 0.099 × 10⁹/L

Patient B:
ANC = 200 × 0.02 = 4 × 10⁹/L

The patient with 99% neutrophils is profoundly neutropenic and at increased risk for infection. The patient with only 2% neutrophils has a normal absolute neutrophil count.

The lesson is simple: a percentage cannot be interpreted without its denominator.

The same applies to reticulocytes. The reticulocyte percentage tells us what fraction of circulating RBCs are reticulocytes, not how many reticulocytes are actually present. As the total RBC count falls in anemia, the reticulocyte percentage can rise even if the absolute number of circulating reticulocytes does not.

Corrected reticulocyte percentage

Historically, clinicians corrected the reticulocyte percentage for the degree of anemia:¹

$$Corrected\ reticulocyte\ \% = Retic\ \% \times \frac{patient\ Hct}{normal\ Hct}$$

For example:

Retic = 6%
Patient Hct = 22.5%
Normal Hct = 45%

Then:$$6\times\frac{22.5}{45}=3\%$$

The crude reticulocyte percentage is 6%, but after correction for the anemia it is 3%.

Hemoglobin may also be used instead of hematocrit:

Corrected reticulocyte % = Retic % × (patient Hb / normal Hb)

When hemoglobin and hematocrit track in their usual relationship, the two approaches give similar results. As we will see, they may diverge substantially when red-cell indices are markedly abnormal.

Conventionally, a corrected reticulocyte percentage below about 2% in an anemic patient is interpreted as a low reticulocyte response, whereas a value above about 2–3% is interpreted as a regenerative response. As we will see, whether such thresholds justify the stronger terms “appropriate” and “inappropriate” is a separate question.

But there is a more direct way to solve the denominator problem: determine the absolute number of reticulocytes in the blood, as discussed below.

Return to the Neutrophil Analogy

The analogous maneuver for neutrophils would be:

Corrected neutrophil % = neutrophil % × (patient WBC / normal WBC)

So if the WBC were 2 × 10⁹/L with 50% neutrophils, and we assumed a normal WBC of 5 × 10⁹/L:

Corrected neutrophil % = 50% × (2 / 5) = 20%

But we would never report a “corrected neutrophil percentage.”

We would simply calculate the ANC.

The same logic applies to reticulocytes.

Absolute reticulocyte count

The ARC is:²$$ARC = RBC\ count\times reticulocyte\ fraction$$

If:$$RBC=2.5\times10^{12}/L$$

and:$$Reticulocytes=7.2\%$$

then:$$ARC=2.5\times0.072$$$$=0.18\times10^{12}/L$$

or:$$180\times10^9/L$$

That is an appealing quantity because it answers a direct question:

How many reticulocytes are circulating per unit volume of blood?

For routine anemia classification, I still think this is usually the most useful first number.

But notice the wording.

It tells us how many reticulocytes are circulating, but not whether that degree of reticulocytosis is proportionate to the severity of the anemia.

---

How Do We Know Whether the Response Is Appropriate?

The ARC solves the denominator problem. It tells us how many reticulocytes are actually circulating.

But it leaves a different question unanswered: is that number appropriate for the severity of the anemia?

Consider our patient again:

Hemoglobin = 3 g/dL
ARC = 180 × 10⁹/L

An ARC of 180 × 10⁹/L is above the commonly used operational threshold of about 120 × 10⁹/L and therefore establishes a regenerative response. But would we interpret the same ARC of 180 × 10⁹/L the same way in a patient with a hemoglobin of 10 g/dL?

Probably not.

Ideally, we would have something like:

Expected ARC = f(severity of anemia)

We could then compare the observed ARC with the expected ARC for a given hemoglobin or hematocrit.

But no broadly validated relationship of this kind is used in clinical practice.

Interestingly, the reticulocyte production index does incorporate the severity of anemia into its calculation. But that is not why it was developed. The RPI was designed to address a different problem: the possibility that reticulocytes released prematurely during erythropoietic stress remain recognizable as reticulocytes for longer in the circulation.

As we will see, that distinction matters.

---

What Is the Reticulocyte Production Index?

The 1969 experiment

The intellectual foundation of the RPI comes largely from Robert Hillman’s classic 1969 Journal of Clinical Investigation paper, “Characteristics of marrow production and reticulocyte maturation in normal man in response to anemia.”⁷ These thresholds remain widely reproduced in modern teaching resources.

But it is important to distinguish what the equation does from what we infer from it.

The RPI does not contain a validated function saying:$$Hb=3 \Rightarrow expected\ erythroid\ output=X$$

It adjusts the corrected reticulocyte percentage for the presumed prolonged maturation time of prematurely released reticulocytes.

The designation “appropriate” is an interpretation placed on the resulting number.

How secure is the maturation correction?

This is where things become less comfortable.

The familiar maturation factors are not measurements of reticulocyte residence time in the individual patient.

They are historical approximations.

And the presumed tight relationship between anemia severity and peripheral reticulocyte maturation time has been questioned.

In 1993, Ivor Cavill revisited the issue in an article wonderfully titled “The Rejected Reticulocyte.”⁸ He noted that measured in-vivo maturation times in a small series ranged substantially and argued against assuming a simple fixed relationship between hematocrit and reticulocyte maturation time.

The old Blood Project article already highlighted this concern, as well as later automated data showing only a weak relationship between hemoglobin concentration and measures of reticulocyte maturity.

So two statements can simultaneously be true:

Stress erythropoiesis can release younger reticulocytes that persist longer in the circulating reticulocyte compartment.

And:

The exact maturation correction applicable to an individual patient cannot reliably be inferred from hematocrit using a four-step lookup table.

The first statement supports the concept of the RPI.

The second limits its precision.

Apply the RPI to our patient

Now return to our patient:

Hb = 3 g/dL
Hct = 14%
Reticulocyte percentage = 7.3%
ARC = 180 × 10⁹/L

Using a normal hematocrit of 45%:

Corrected retic % = 7.3% × (14 / 45) = 2.27%

At a hematocrit of 14%, the conventional maturation factor is 2.5:

RPI = 2.27 / 2.5 = 0.91

Importantly, the RPI inherits whatever assumptions are built into the corrected reticulocyte percentage. If hemoglobin rather than hematocrit were used in this patient, the corrected reticulocyte percentage would be 1.46% and the RPI 0.58. Thus, the choice of Hb versus Hct propagates directly into the RPI.

So the same patient gives us two apparently different messages:

ARC = 180 × 10⁹/L → regenerative

RPI = 0.91 → conventionally interpreted as an inadequate response

Neither calculation is wrong. The unresolved question is whether the RPI’s additional corrections justify concluding that the marrow response is inadequate for a hemoglobin of 3 g/dL.

---

Has anyone tried to severity-adjust the ARC directly?

Yes.

And this may be one of the most interesting pieces of the story.

A group studying congenital dyserythropoietic anemia type II developed what they called the bone marrow responsiveness index, or BMRI:⁹

$$BMRI = ARC\times\frac{patient\ Hb}{normal\ Hb}$$

The study was essentially asking:

Can an ARC adjusted for anemia severity distinguish a disease characterized by ineffective erythropoiesis from one characterized by effective erythropoietic compensation?

That is a much narrower question than whether BMRI can determine the adequacy of the marrow response across anemia in general.

Unlike the RPI, there is no maturation correction.

It simply takes the ARC and scales it according to the degree of anemia.

The investigators compared 205 patients with CDA II, a disorder characterized by ineffective erythropoiesis, with 57 patients with hereditary spherocytosis, a model of relatively effective compensatory erythropoiesis.¹⁰

A BMRI below 121 distinguished CDA II from hereditary spherocytosis with approximately 90% sensitivity and 65% specificity.¹¹

That is conceptually fascinating because the BMRI does something closely related to the question we have been circling: it adjusts the absolute reticulocyte count for the severity of anemia without invoking reticulocyte maturation.

But its purpose was narrower. It was developed to distinguish ineffective erythropoiesis in CDA II from the more effective compensatory response seen in hereditary spherocytosis, not to establish a universal standard for how large the reticulocyte response should be at any given hemoglobin concentration.

Apply BMRI to our patient

We can apply the BMRI formula to our patient to see what the index would yield, while recognizing that the published cutoff was derived from a comparison of CDA II and hereditary spherocytosis and has not been validated in severe iron deficiency.

Suppose normal Hb is 14 g/dL.

BMRI = 180 × (3 / 14) = 38.6

This is far below the published cutoff of 121.

We can also turn the calculation around. If we applied the published BMRI cutoff of 121 to this patient, the ARC required to cross that threshold would be:

ARC = 121 × (14 / 3) = 565 × 10⁹/L

That is an extraordinarily high ARC. Rather than proving that this patient’s response is inadequate, the calculation illustrates the danger of extrapolating a disease-specific BMRI cutoff to profound anemia in an entirely different clinical setting.

So now we have:

ARC: regenerative

RPI: conventionally inadequate

BMRI: below the published cutoff

The convergence between RPI and BMRI is interesting because the two indices arrive there differently.

The RPI discounts the observed reticulocyte response partly on the assumption that severe anemia is associated with earlier release and prolonged circulating maturation of reticulocytes.

The BMRI makes no maturation assumption. It simply scales the ARC to the severity of anemia.

Both therefore make an ARC of 180 × 10⁹/L look less impressive in a patient with a hemoglobin of 3 g/dL.

But neither proves that 180 × 10⁹/L is actually an inadequate response in this patient.

The BMRI nevertheless provides an important proof of concept: investigators have explicitly tried to adjust the ARC for anemia severity without invoking reticulocyte maturation.

---

Five Reticulocyte Measures in This Patient

For this patient:

Reticulocyte percentage = 7.3%

Absolute reticulocyte count = 180 × 10⁹/L

Corrected reticulocyte percentage = 2.27% using hematocrit (1.46% if Hb is used)

Reticulocyte production index = 0.91 (0.58 if Hb is used)

Bone marrow responsiveness index = 36, assuming a normal hemoglobin of 15 g/dL

These five numbers are all derived from the same clinical data, but they answer different questions and lead to different impressions of the marrow response.

What is actually missing?

What we do not appear to have in routine clinical hematology is a broadly validated, disease-agnostic relationship that tells us the expected ARC at a given degree of anemia:

$$Expected\ ARC=f(Hb)$$

or:$$Expected\ effective\ RBC\ output=f(Hb)$$

that allows us to say:

Hb 10 → expected ARC range A
Hb 7 → expected ARC range B
Hb 5 → expected ARC range C
Hb 3 → expected ARC range D

under defined physiologic conditions.

That is not the same as saying we know nothing quantitative about erythropoiesis.

Classic ferrokinetic work demonstrates that erythropoietic production can increase several-fold under intense stimulation.

But those are measurements of erythropoietic production, not a validated bedside conversion table between hemoglobin concentration and circulating ARC.

We should be particularly careful not to take a ferrokinetic statement such as “production can rise several-fold,” multiply a normal ARC by the same factor, and declare that result to be the expected ARC.

Doing so assumes that ARC scales directly with production.

The whole rationale for the maturation correction is that this assumption may fail.

The time axis matters

Another reason a simple Hb-to-ARC table would be difficult to construct is time.

Hb 3 reached yesterday is not physiologically equivalent to Hb 3 that developed over several months.

The erythropoietic response takes time to develop.

A reticulocyte count obtained shortly after abrupt hemorrhage or hemolysis may not yet reflect the full marrow response.

Conversely, a patient with longstanding anemia may already have reached a sustained adaptive state.

Thus the same:

Hb = 3
ARC = 180

could mean something very different depending on whether the anemia developed over:

• hours,
• days,
• weeks,
• or months.

Neither ARC nor RPI directly contains that information.

Substrate matters

Our patient also has:$$Ferritin=5\ ng/mL$$

That is not a side issue.

The kidney may be generating a powerful EPO signal.

The erythroid marrow may be biologically intact.

But erythropoiesis cannot proceed normally without iron.

So a patient may simultaneously have:

strong erythropoietic drive

plus

limited ability to translate that drive into hemoglobinized RBC output.

If mechanical hemolysis is also present, the physiology becomes even more interesting:

Hemolysis stimulates erythropoiesis.

Iron deficiency constrains erythropoiesis.

The resulting ARC reflects the net effect of both processes.

Neither ARC nor RPI can disentangle them by itself.

---

What about measuring reticulocyte maturity directly?

The RPI uses the severity of anemia as a surrogate for how immature the circulating reticulocytes are.

Modern analyzers can measure reticulocyte maturity more directly.

Fluorescent dyes bind residual RNA. Younger reticulocytes contain more RNA and therefore generate stronger fluorescence.

This allows calculation of the immature reticulocyte fraction, or IRF.¹²

IRF provides a more direct measure of the maturity distribution of circulating reticulocytes than hematocrit does, although it does not directly measure reticulocyte residence time and cannot simply replace the traditional maturation factor.

If the rationale for dividing the reticulocyte count by 2.5 is that this patient’s blood contains an unusually immature population of reticulocytes, why not measure the maturity of those cells rather than infer it from hematocrit?

In practice, IRF has not replaced the RPI as a universal correction factor. Reticulocyte maturity measures vary among platforms and are less well standardized than the basic reticulocyte count itself.¹³

Still, IRF reminds us how much technology has changed since the RPI was developed.

We are no longer limited to looking at a hematocrit and guessing how immature the reticulocytes must be.

---

Return to the patient

We can now return to the question that started this essay.

Our patient has:

Hb = 3 g/dL
Ferritin = 5 ng/mL
ARC = 180 × 10⁹/L
RPI ≈ 0.9
BMRI ≈ 36

The ARC establishes a regenerative response. The RPI and BMRI both make that response look less impressive when additional adjustments are applied, but neither provides a validated standard for what the reticulocyte response should be at a hemoglobin of 3 g/dL.

Profound iron deficiency also provides a compelling biological reason why an active erythroid response might nevertheless be constrained.

So the most defensible conclusion is:

There is a regenerative erythroid response. Whether its magnitude is quantitatively appropriate for the severity of the anemia remains uncertain.

---

What does the ARC threshold of 120 × 10⁹/L mean, then?

I still find approximately 120 × 10⁹/L extremely useful.

But I would change the language.

I used to say:$$ARC>120\times10^9/L = appropriate\ response$$

I would now say:$$ARC\gtrsim120\times10^9/L = evidence\ of\ a\ regenerative\ response$$

That is an operational branch point.

It is not a physiologic wall.

The exact reference interval varies by laboratory, analyzer, age, and other factors. The 120 threshold should be viewed as a useful clinical heuristic rather than a universal biological constant.

More importantly:

Regenerative does not mean maximally regenerative.

Regenerative does not mean sufficient.

And regenerative does not mean quantitatively appropriate for every severity of anemia.

That is the change in my position.

---

Five Measures, Five Different Questions

Measure	Primary question addressed	Important limitation
Reticulocyte %	What fraction of circulating RBCs are reticulocytes?	Strongly affected by the shrinking RBC denominator in anemia
Corrected reticulocyte %	What is the reticulocyte percentage after correcting for the degree of anemia?	Derived quantity; result depends on how anemia severity is defined (for example, Hb versus Hct)
ARC	Is there a regenerative response?	Measures the circulating reticulocyte concentration; does not establish whether the response is proportionate to anemia severity
RPI	Does the circulating reticulocyte pool accurately reflect red-cell production after accounting for presumed prolonged maturation?	Maturation factor is an historical approximation, not a patient-specific measurement
BMRI	In its studied setting, is the ARC low relative to the severity of anemia?	Validated only in a specific disease comparison; not established as a general anemia index

The interesting point is that there is no single unquestioned gold standard in this table.

---

What I would do clinically

For routine anemia classification, I still start with the ARC.

The first useful question is generally:

Is there evidence of a regenerative reticulocyte response?

An elevated ARC tells me yes.

That pushes me toward:

• blood loss,
• hemolysis,
• recovery from a previous marrow insult,
• or response to therapy.

A low ARC pushes me toward:

• impaired production,
• ineffective erythropoiesis,
• substrate deficiency,
• renal/EPO problems,
• inflammatory suppression,
• or marrow disease.

But in an extreme case, I would stop short of calling an elevated ARC “adequate.”

I would instead ask:

Does the magnitude of the response make biological sense given:

• the severity of anemia,
• the duration of anemia,
• iron availability,
• renal function and EPO response,
• marrow reserve,
• ongoing hemolysis or bleeding,
• inflammation,
• recent transfusion,
• and the trajectory of the reticulocyte count?

The trajectory is particularly useful.

If severe iron deficiency is treated and the ARC subsequently rises substantially, that supports the idea that iron availability had been constraining an active marrow response.

But there is no magic follow-up target such as 300 or 400 × 10⁹/L that must be reached to prove marrow competence.

The direction and clinical context matter more than an invented threshold.

---

So do we need the RPI?

I no longer think the best answer is simply “no.”

The RPI addresses a real physiologic issue:$$circulating\ pool\neq production\ flux$$

when reticulocyte residence time changes.

Hillman’s work deserves more respect than a casual dismissal of the maturation correction would imply.

At the same time, I do not think an RPI calculator converts an uncertain physiologic relationship into a precise measure of what the marrow ought to be doing.

The RPI is best regarded as a semiquantitative model.

Its value may lie in reminding us that the same circulating reticulocyte count need not have the same biological meaning at every degree of anemia.

But its familiar cutoffs should not be mistaken for a validated hemoglobin-specific expected-response curve.

Interestingly, BMRI approaches the problem differently: it explicitly scales the ARC to the severity of anemia without making any assumption about reticulocyte maturation.

Neither approach has emerged as a universal standard.

That may be because the biology simply does not permit a simple universal answer.

---

Bottom line

The crude reticulocyte percentage is particularly vulnerable to misinterpretation in anemia because its denominator, the total circulating RBC population, falls.

The ARC avoids that denominator distortion and remains, in my view, the clearest practical starting point for assessing the reticulocyte response.

An ARC around or above 120 × 10⁹/L supports a regenerative component.

It should not automatically be called an “appropriate” response.

The ARC measures a circulating pool.

The RPI attempts to convert that pool into an estimate of effective production by incorporating red-cell volume and an assumed reticulocyte maturation time.

The traditional maturation correction has a legitimate physiologic rationale, but its patient-specific precision is uncertain.

The RPI also has an interesting mathematical consequence: increasingly severe anemia requires a larger reticulocyte response to achieve the same RPI. That may be clinically sensible, but the maturation factors were not calibrated as a validated Hb-specific expected-response curve.

The BMRI demonstrates that severity-adjusting the ARC without invoking maturation is possible and may be clinically informative in selected populations. But its published cutoff was developed in CDA II versus hereditary spherocytosis and cannot simply be generalized to all anemia.

We therefore have good tools for identifying regeneration.

We have much less certain tools for determining whether the magnitude of regeneration is quantitatively sufficient for a particular patient at a particular hemoglobin.

In the patient with:

Hb 3 g/dL
Ferritin 5 ng/mL
ARC 180 × 10⁹/L

the most defensible conclusion is:

There is a regenerative erythroid response.

Whether that response is quantitatively proportionate to the severity of the anemia cannot be established from the ARC or RPI alone.

---

Closing thought

The reticulocyte count looks like one of the simplest tests in hematology.

It is not.

The percentage measures a ratio.

The ARC measures a pool.

The RPI tries to infer a flux.

The BMRI tries to grade that pool against the severity of anemia.

And none of them directly measures the quantity clinicians often want most:

How much effective erythropoiesis should this particular patient be producing, at this particular moment, with this particular degree and cause of anemia?

Perhaps the mistake was never choosing the wrong index.

Perhaps it was assuming that a single index could answer that question in the first place.

---

Key Takeaways

• Reticulocyte percentage is vulnerable to a denominator problem in anemia. As the circulating RBC count falls, the reticulocyte percentage can rise even if the absolute number of reticulocytes does not.
• The absolute reticulocyte count (ARC) avoids that denominator distortion. For routine anemia classification, it is the clearest practical starting point for deciding whether a regenerative erythroid response is present. An ARC around or above 120 × 10⁹/L supports a regenerative response, but not necessarily an “appropriate” response. The threshold is a useful clinical branch point, not a validated measure of whether marrow output is quantitatively sufficient for the severity of anemia.
• The ARC measures a circulating pool, not production flux. Reticulocyte residence time can therefore affect the relationship between ARC and true erythroid output. But knowing flux is not necessarily required to answer the clinical questions we actually care about: is there a regenerative response, and is that response appropriate for the severity of the anemia? The first can often be addressed with the ARC; the second requires a validated severity-adjusted framework, not necessarily a flux estimate.
• The RPI was developed to correct for presumed prolongation of reticulocyte maturation time, not to determine how much marrow response should be expected at a given hemoglobin. Its primary purpose was to make the circulating reticulocyte pool a better surrogate for production flux in research studies, namely physiologic studies of erythropoiesis.
• The RPI happens to become more stringent as anemia becomes more severe. As hematocrit falls, the corrected reticulocyte percentage decreases and the maturation divisor increases. This gives the RPI a severity-dependent behavior, but that was not the original physiologic question it was designed to answer.
• That severity correction is coarse. Traditional RPI calculations use only a few stepwise hematocrit ranges, with maturation factors of roughly 1, 1.5, 2, and 2.5. Once severe anemia reaches the lowest range, the maturation correction plateaus, so very different degrees of profound anemia may receive the same divisor.
• Neither ARC nor RPI provides a validated expected response for a specific hemoglobin concentration. We do not have a broadly accepted relationship such as expected ARC = f(Hb) that tells us what the marrow “should” be producing at Hb 10, 7, 5, or 3 g/dL.
• Determining whether a response is inadequate is different from explaining why it is inadequate. Iron deficiency, renal dysfunction, inflammation, marrow disease, and other factors may explain a limited erythroid response, but they do not provide a validated standard for determining whether a given ARC is quantitatively appropriate for a given severity of anemia.
• Different reticulocyte measures answer different questions. Reticulocyte percentage measures a ratio, ARC measures a circulating pool, RPI attempts to estimate production flux, and BMRI scales the ARC to anemia severity. None directly tells us how much effective erythropoiesis a particular patient ought to be producing at that moment.