Ferritin Cutoffs and Diagnosis of Iron Deficiency in Primary Care 2024 Jäger et al

[Andy]

Key Points

Question How is the choice of guideline-recommended cutoffs for ferritin associated with the incidence of iron deficiency diagnoses in primary care?

Findings In this cohort study of 255 351 adult primary care patients, ferritin cutoffs of 15, 30, and 45 ng/mL were associated with incidences of iron deficiency diagnoses of 10.9, 29.9, and 48.3 cases per 1000 patient-years, respectively.

Meaning The results of this study provide useful components for the evaluation of ferritin testing in high-resource primary care settings and call for harmonization of guidelines for iron deficiency.


Abstract

Importance Ferritin is often measured by general practitioners, but the association of different cutoffs with the rates of iron deficiency diagnoses, particularly nonanemic iron deficiency, is unknown.

Objective To investigate the association of the ferritin cutoff choice with the incidence of nonanemic and anemic iron deficiency diagnoses in primary care.

Design, Setting, and Participants In this retrospective cohort study, patients 18 years or older with at least 1 consultation with a general practitioner participating in the Family Medicine Research Using Electronic Medical Records (FIRE) project, an electronic medical records database of Swiss primary care, from January 1, 2021, to November 30, 2023, were evaluated.

Exposures Sex, age, clinical patient characteristics, and professional general practitioner characteristics.

Main Outcomes and Measures Incidence of iron deficiency diagnoses (nonanemic and anemic) at ferritin cutoffs of 15, 30, and 45 ng/mL and ferritin testing itself. Time-dependent Cox proportional hazards regression was used to examine associations of patient and general practitioner characteristics with ferritin testing as adjusted hazard ratios (AHRs).

Results The study included 255 351 patients (median [IQR] age, 52 [36-66] years; 52.1% female). Per 1000 patient-years and at ferritin cutoffs of 15, 30, and 45 ng/mL, iron deficiency diagnoses had incidences of 10.9 (95% CI, 10.6-11.2), 29.9 (95% CI, 29.4-30.4), and 48.3 (95% CI, 47.7-48.9) cases, respectively; nonanemic iron deficiency diagnoses had incidences of 4.1 (95% CI, 3.9-4.2), 14.6 (95% CI, 14.3-15.0), and 25.8 (95% CI, 25.3-26.2) cases, respectively; and anemic iron deficiency diagnoses had incidences of 3.5 (95% CI, 3.3-3.7), 6.0 (95% CI, 5.8-6.2), and 7.5 (95% CI, 7.3-7.7) cases, respectively. Ferritin testing showed notable associations with fatigue (AHR, 2.03; 95% CI, 1.95-2.12), anemia (AHR, 1.75; 95% CI, 1.70-1.79), and iron therapy (AHR, 1.50; 95% CI, 1.46-1.54). Ferritin testing was associated with female sex in all age groups, including postmenopausal. Of the patients who received ferritin testing, 72.1% received concomitant hemoglobin testing, and 49.6% received concomitant C-reactive protein testing.

Conclusions and Relevance In this retrospective cohort study of primary care patients, ferritin cutoffs of 30 and 45 ng/mL were associated with a substantially higher incidence of iron deficiency compared with 15 ng/mL. These results provide a basis for health system-level evaluation and benchmarking of ferritin testing in high-resource settings and call for a harmonization of diagnostic criteria for iron deficiency in primary care.

Open access, https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2821953

 

[Arnie Pye]

Isn't the conclusion of this paper equivalent to "water is wet"? Am I missing something?

If the cutoff for the bottom of the range is 15, 30 or 45 ng/mL of ferritin then obviously there will be more people diagnosed with low levels or deficiency when the cutoff is higher.

I have no quibble with the idea that ferritin ranges should be harmonised. I think they should be too.

Incidentally, NICE uses a cutoff of 30ng/mL but many UK labs seem to ignore this, and continue to use levels of 5, 10, and 15 ng/mL for ferritin.

https://cks.nice.org.uk/topics/anaemia-iron-deficiency/

https://cks.nice.org.uk/topics/anaemia-iron-deficiency/diagnosis/investigations/

In the second link, under the heading "Interpreting ferritin levels" it says :

In all people, a serum ferritin level of less than 30 micrograms/L confirms a diagnosis of iron deficiency.

 

[bobbler]

Key Points

Question How is the choice of guideline-recommended cutoffs for ferritin associated with the incidence of iron deficiency diagnoses in primary care?

Findings In this cohort study of 255 351 adult primary care patients, ferritin cutoffs of 15, 30, and 45 ng/mL were associated with incidences of iron deficiency diagnoses of 10.9, 29.9, and 48.3 cases per 1000 patient-years, respectively.

Meaning The results of this study provide useful components for the evaluation of ferritin testing in high-resource primary care settings and call for harmonization of guidelines for iron deficiency.


Abstract

Importance Ferritin is often measured by general practitioners, but the association of different cutoffs with the rates of iron deficiency diagnoses, particularly nonanemic iron deficiency, is unknown.

Objective To investigate the association of the ferritin cutoff choice with the incidence of nonanemic and anemic iron deficiency diagnoses in primary care.

Design, Setting, and Participants In this retrospective cohort study, patients 18 years or older with at least 1 consultation with a general practitioner participating in the Family Medicine Research Using Electronic Medical Records (FIRE) project, an electronic medical records database of Swiss primary care, from January 1, 2021, to November 30, 2023, were evaluated.

Exposures Sex, age, clinical patient characteristics, and professional general practitioner characteristics.

Main Outcomes and Measures Incidence of iron deficiency diagnoses (nonanemic and anemic) at ferritin cutoffs of 15, 30, and 45 ng/mL and ferritin testing itself. Time-dependent Cox proportional hazards regression was used to examine associations of patient and general practitioner characteristics with ferritin testing as adjusted hazard ratios (AHRs).

Results The study included 255 351 patients (median [IQR] age, 52 [36-66] years; 52.1% female). Per 1000 patient-years and at ferritin cutoffs of 15, 30, and 45 ng/mL, iron deficiency diagnoses had incidences of 10.9 (95% CI, 10.6-11.2), 29.9 (95% CI, 29.4-30.4), and 48.3 (95% CI, 47.7-48.9) cases, respectively; nonanemic iron deficiency diagnoses had incidences of 4.1 (95% CI, 3.9-4.2), 14.6 (95% CI, 14.3-15.0), and 25.8 (95% CI, 25.3-26.2) cases, respectively; and anemic iron deficiency diagnoses had incidences of 3.5 (95% CI, 3.3-3.7), 6.0 (95% CI, 5.8-6.2), and 7.5 (95% CI, 7.3-7.7) cases, respectively. Ferritin testing showed notable associations with fatigue (AHR, 2.03; 95% CI, 1.95-2.12), anemia (AHR, 1.75; 95% CI, 1.70-1.79), and iron therapy (AHR, 1.50; 95% CI, 1.46-1.54). Ferritin testing was associated with female sex in all age groups, including postmenopausal. Of the patients who received ferritin testing, 72.1% received concomitant hemoglobin testing, and 49.6% received concomitant C-reactive protein testing.

Conclusions and Relevance In this retrospective cohort study of primary care patients, ferritin cutoffs of 30 and 45 ng/mL were associated with a substantially higher incidence of iron deficiency compared with 15 ng/mL. These results provide a basis for health system-level evaluation and benchmarking of ferritin testing in high-resource settings and call for a harmonization of diagnostic criteria for iron deficiency in primary care.

Open access, https://jamanetwork.com/journals/jamanetworkopen/fullarticle/2821953

Am I reading this right that without the specific numbers the conclusion is just stating a truism that if you make a cut-off higher then you'd get less people. Of course really they aren't changing anything at all. The old chestnut of 'harmonisation' would seem a fair argument if it meant that it wasn't looking at costs but making sure that where it indicated something it was that level being used by all.


The discussion is flagging up a few interesting bits though and I'm still reading it trying to work out how agenda-driven/manifesto-based vs 'review of what would be the right number for different things' it is


For example:

The proportion of ferritin tests without accompanying CRP or hemoglobin tests was surprisingly high because most guidelines explicitly recommend screening for systemic inflammation and anemia in the workup of iron deficiency.4 Hemoglobin and CRP testing was associated with fatigue and with fewer previous primary care visits, suggesting that they were more commonly used in patients in whom a symptom-led search for iron deficiency took place rather than in episodes of care involving routine screening.

and a pretty important note given they suggest their study provides enough information based for health-system level evaluations of ferritin testing is the limitations section, which includes this:

The FIRE database does not allow access to presenting symptoms as documented in the clinical notes, which is an important limitation together with the lack of information on referral. In addition, we had limited access to information on gastrointestinal risk factors, such as celiac disease.

Is the study forgetting that the test is perhaps run as a way in which to check or get further information on indicators on whether there is 'something else' going on, rather than being the 'be all and end all'. ie there might be some where it is 'your ferritin is low, here are supplements' (and I thought that was relatively cheap) but that it is just as likely to rule out certain things or be the start of an indicator for something else?

It seems to mention the good practice on CRP and haemoglobin due to things like inflammation however

 

[bobbler]

Isn't the conclusion of this paper equivalent to "water is wet"? Am I missing something?

If the cutoff for the bottom of the range is 15, 30 or 45 ng/mL of ferritin then obviously there will be more people diagnosed with low levels or deficiency when the cutoff is higher.

I have no quibble with the idea that ferritin ranges should be harmonised. I think they should be too.

Incidentally, NICE uses a cutoff of 30ng/mL but many UK labs seem to ignore this, and continue to use levels of 5, 10, and 15 ng/mL for ferritin.

https://cks.nice.org.uk/topics/anaemia-iron-deficiency/

https://cks.nice.org.uk/topics/anaemia-iron-deficiency/diagnosis/investigations/

In the second link, under the heading "Interpreting ferritin levels" it says :

In all people, a serum ferritin level of less than 30 micrograms/L confirms a diagnosis of iron deficiency.

I thought the following was interesting about how often teh test was offered:

In particular, a Swiss study revealed that 27% of the population received serum ferritin testing in 2018,8 and ferritin ranked among the most frequently ordered laboratory tests in Swiss primary care between 2009 and 2018.9

Which seemed extraordinary. They themselves in the methods note they had a study cohort (ethics had been waived to allow it to be collection of records) of 255 351 patients, "During the study period, 72 817 patients (28.5%) received ferritin testing." the study period was more than a year: "a retrospective cohort study from January 1, 2021, to November 30, 2023". SO I haven't worked out for sure whether some of those patients would have been tested all 3 of those years, or if you can divide that number by 3 - I'd guess it is somewhere in between.

When I looked up the actual reference 9 (Trends and Between-Physician Variation in Laboratory Testing: A Retrospective Longitudinal Study in General Practice - PubMed (nih.gov)) notes it is: "For 15 commonly used test types, we defined specific laboratory testing rates (sLTR) as the percentage of consultations involving corresponding laboratory testing requests. " and that was based on them retrospectively looking at the notes of over 6m consultations from 2009-18.

BUT I still don't know where the 27% figure came from - because the paper mainly just lists changes over that time period or other correlations. I'm very doubtful the 27% is a correct quote though because the introduction states the following, which would make ferritin testing being 27% illogical (if the top 3 most used tests were lower %):

The top three most frequently requested test types among GPs over the study period (Figure 1 and Figure 2) were complete blood count (among-GP median sLTR 15.1%, IQR 11.5–18.5%), C-reactive protein (CRP; among-GP median sLTR 10.4%, IQR 6.8–14.4%) and serum creatinine (among-GP median 7.2%, IQR 5.3–9.1%).

A prominent Swiss primary care guideline discourages iron deficiency screening in the general population, with exceptions for conditions that require special consideration due to increased risk, such as inflammatory bowel disease (IBD), or when evidence suggests a benefit from replete iron stores, such as chronic kidney disease (CKD).11

I'll be honest, the word 'choice' being used sprung out at me here:

Although previous studies have estimated incidences of anemic iron deficiency in primary care,13,14 they have not evaluated the choice of different ferritin cutoffs, and the incidence of nonanemic iron deficiency has not been investigated.

given it follows some claim of it being important because overtreatment can lead to symptoms (particularly in oral preparations) to little benefit and the healthcare costs. I'm unconvinced by this sell - isn't testing ferritin about 2 things, and not one. The first being about what it does on the tin 'is the ferritin level the be all and end all issue' and the second is 'as part of getting to the bottom of'.

I know their issue seems to be about 'not treating/treatment cut-off' not the tests being done themselves apparently but if you were going to do a referral or a treatment and its been a while then you'd probably rule out the ferritin. ANd if you pick 15 how are those who are getting people referred to them going to feel about that?

I don't know how their healthcare system works but the likelihood of someone having ferritin 40 and getting terrible side-effects and still insisting the solution is they will persevere sounds pretty unusual. And at that point if they are getting those side effects still the question is whether the test is sensibly actually being done because the GP is considering other causes and needing to make sure it hasn't dropped prior to looking into those?

 

[Jonathan Edwards]

I agree with others. This is measuring whether or not a piece of string is too long.

I think it reflects a failure to understand that diagnostic cateogories in medicine are always placeholders for questions about what treatments might be useful in a particular case now or later. 'Iron deficiency' is a placeholder for ' the person would benefit from some iron'. Whether or not someone would benefit from iron is dependent on a vast range of factors, none of which correlate absolutely with ferritin. So they have the question back to front.

There is a worrying tendency for primary care physicians to think that this is how you do medicine. It is a clear indication of why the primary care level of understanding of most medical problems is simply not good enough in modern medical science. Primary care should largely be limited to a sorting house and real medical problems should be handled by people with knowledge of the specifics.

 

[Arnie Pye]

If a healthcare system discourages the use of ferritin (iron stores) tests it would be women and girls who suffer for it, since they are the ones who have periods and lose blood every month. And when periods stop there is no automatic improvement in iron levels just because the woman has stopped bleeding every month, although some will improve. And having low ferritin is exhausting, depressing, and painful. It makes life very much harder than it ought to be.

Am I reading this right that without the specific numbers the conclusion is just stating a truism that if you make a cut-off higher then you'd get less people.

You have that the wrong way round.

Let's assume that ferritin levels in a population follow a normal distribution. It's not an unreasonable assumption. If someone has a ferritin level of 14 or less and the reference range is 15 - 200 then a level of 14 is deficient because it is lower than the bottom of the range. If the reference range is 45 - 200 then more people will have a level which is lower than 45 than had a level of 15 or lower.

Haemoglobin which is below range indicates the patient is anaemic. People can be iron deficient long before they are anaemic, but many doctors are quite happy to ignore iron deficiency if the patient is not anaemic which, in my opinion, is cruel.