Deceived by Elevated A1C: Cases of Misdiagnosed Diabetes

A healthy, asymptomatic 19-year-old woman was referred to the diabetes clinic for a second opinion on her diabetes subtype. A recent routine laboratory screening had been performed by her general practitioner (GP) and had revealed an elevated A1C, which was confirmed at the diabetes outpatient clinic. Table 1 summarizes representative glycemic measures. However, a nonfasting plasma glucose level measured on the same day was only 86.4 mg/dL (4.8 mmol/L).

The patient had a normal body weight with a BMI of 19.5 kg/m², and her family history of diabetes was unknown, as she had been adopted from China. Physical examination was normal and without insulin resistance (IR) stigmata such as acanthosis nigricans, acne, or central adiposity. Polycystic ovary syndrome was not suspected because she reported having regular periods and had no complaints about hirsutism.

Because GAD65 and IA-2 autoantibodies were negative, a MODY screening was performed (specifications are provided in the Supplementary Material). A variant of unknown significance (VUS) in the CEL gene (c.1719_1720ins33) was detected. Pathogenic variants in the CEL gene are related to MODY8/CEL-MODY (11). After this, insulin treatment was initiated for a clinical diagnosis of early antibody-negative type 1 diabetes.

Nevertheless, the patient’s A1C results remained elevated despite insulin treatment, and persistently normal blood glucose levels were obtained by self-monitoring of blood glucose (SMBG). Therefore, continuous glucose monitoring (CGM) was carried out. CGM data revealed normoglycemia, with a mean glucose of 100.8 mg/dL (5.6 mmol/L) and time in range (70.2–140.4 mg/dL [3.9–7.8 mmol/L]) >99%. CGM was repeated after cessation of insulin treatment, with similar results. Next, a confirmatory oral glucose tolerance test (OGTT) was performed (75 g glucose monohydrate dissolved in 225 mL water) and showed a slightly impaired fasting glucose (IFG) tolerance. Simultaneously, point-of-care testing (POCT) for A1C became available in the clinic and showed an A1C of 5.3% (34 mmol/mol). Hence, the diabetes diagnosis was rejected, and insulin treatment was permanently ceased.

A hemoglobin variant was suspected, but the patient had normal hemoglobin levels with normocytic erythrocytes, and the HPLC-based hemoglobin fraction analysis showed no signs of obvious interfering variants. However, A1C was also cross-examined by both HPLC and CE on the same sample, and owing to the observation of a substantial discrepancy, the HBA1/HBA2 and HBB genes were sequenced. These gene analyses are not widely available but can be requested in laboratories that specialize in diagnosing hemoglobinopathies. The sequencing revealed that the patient was a heterozygous carrier of the Hb Phnom Penh variant, which has previously been reported to cause falsely elevated A1C when analyzed with HPLC (12,13). The prevalence and distribution of this variant is unknown, but it has been detected in individuals from several Southeast Asian countries (13–16).

The patient was discharged from the clinic and informed that IFG is associated with increased risk for diabetes later in life. The impact of the VUS in the CEL gene was not investigated further after retraction of the diabetes diagnosis.

A 41-year-old Danish woman was referred to the clinic for review of her diabetes classification. Four years before the referral, diabetes was diagnosed incidentally from routine blood testing. Her elevated A1C was confirmed by a repeated measurement and followed by measurement of GAD65 antibodies (<14 kIU/L [not elevated]) and fasting C-peptide (579 pmol/L [reference values for a female >10 years of age 370–1,470 pmol/L]). Representative glycemic measures are shown in Table 1. Metformin treatment was initiated by her GP, and a decrease in A1C was observed, with stable quarterly values of 6.0–6.5% (42–48 mmol/mol) over the next 4 years.

The patient had a BMI of 23 kg/m² at the time of diagnosis and was healthy apart from chronic tension headaches and mild depression. Her father (on insulin), paternal grandmother (on insulin), and two paternal aunts (on metformin) were diagnosed with diabetes between the ages of 30 and 50 years. All affected relatives were of normal weight or mildly overweight. The patient never had hyperglycemic symptoms, and SMBG had always been within the normal range. The physical examination was without clinical signs of IR. She was nulligravida; hence, there was no history of gestational diabetes.

Surprisingly, at her first visit to the clinic, her POCT A1C was 5.1% (32 mmol/mol). Based on this result, metformin treatment was paused. An OGTT was performed, and she had a fasting plasma glucose (FPG) of 108.0 mg/dL (6.0 mmol/L) and a 2-hour plasma glucose of 154.8 mg/dL (8.6 mmol/L), and such results were in accordance with IFG and impaired glucose tolerance (IGT) but not diabetes. Because of the combination of normal A1C, mild hyperglycemia, and family history of early-onset diabetes, MODY screening was conducted with normal sequencing of relevant genes (see Supplementary Material for specifications). Treatment with metformin was terminated, and an OGTT was repeated 1 year later showing a 2-hour plasma glucose of 133.2 mg/dL (7.4 mmol/L).

The patient showed no obvious biochemical changes suggestive of changed erythrocyte turnover from blood samples adjacent to her referral (i.e., she had stable total hemoglobin, normocytic erythrocytes with normal reticulocyte count, and no elevated lactate dehydrogenase or bilirubin). The chromatograms from the A1C measurement by HPLC were carefully reviewed and revealed subtle signs of an interfering agent. At the next follow-up in the outpatient clinic, the patient reported long-term daily use of up to 2–3 g of acetylsalicylic acid because of her headaches. Acetylsalicylic acid is well known as a possible interfering agent in A1C measurements performed by HPLC methods, with significant increases in A1C (by 0.2% [2 mmol/mol]) after short-term in vitro exposure to concentrations corresponding to maximum plasma concentration when used as analgesics (corresponding to an intake of 3–4 g/day) (8). However, no information is currently available on the in vivo effect of long-term intake of acetylsalicylic acid in analgetic doses on A1C levels measured by HPLC. After the introduction of a CE-based A1C measurement method in the laboratory, a cross-examination was performed. A1C measured by HPLC was 6.4% (46 mmol/mol), whereas A1C measured by CE was 5.5% (37 mmol/mol).

The patient was informed that the previous diagnosis of diabetes was incorrect, but because of possible IFG and IGT and a positive family history of diabetes, she was advised to have annual testing for diabetes (using either A1C measured by CE or an OGTT).

A 37-year-old Afghan woman with diabetes was referred to the clinic for review of diabetes subtype and treatment before a scheduled fertility treatment. Her first A1C ≥6.5% (48 mmol/mol) had been observed 6 years before the referral, and repeated A1C measurements in the intervening period were stable at 6.6–6.7% (49–50 mmol/mol), as shown in Table 1. Metformin treatment had been attempted several times by her GP without obvious improvement in glycemic control, and the treatment was ceased because of gastrointestinal side effects.

The patient was known to have multiple psychiatric diagnoses, long-term recurrent migraines, a tendency to have recurrent cystitis, and intermittent unspecified neurological symptoms. Her father and sister had type 2 diabetes, and diabetes had also been suspected previously in two younger brothers.

At her first visit to the outpatient clinic, she reported hyperglycemic symptoms (i.e., fatigue, increased thirst, dryness of the mouth, and polyuria) and abdominal discomfort. Physical examination was without clinical signs of IR, and her BMI was 26.4 kg/m². Fasting C-peptide was 901 pmol/L (reference values for a female >10 years of age 370–1,470 pmol/L), and a concomitant plasma glucose was 91.8 mg/dL (5.1 mmol/L). She was negative for GAD65 and IA-2 autoantibodies. MODY screening was performed and revealed a VUS in the HNF1B gene (c.Ser367Gly) (see Supplementary Material for specifications). Pathogenic mutations in the HNF1B gene result in HNF1B-MODY, a syndromic form of diabetes that often includes malformations of the kidneys and in some cases also genital malformations. Because the patient was confirmed to not have diabetes, had normal anatomy of the kidneys and genital organs (based on a prior MRI scan), and had given birth to a son 20 years earlier, we did not investigate further whether the detected variant was pathogenic (17).

Her first appointment in the outpatient clinic coincided with an A1C method change from HPLC to CE at the hospital laboratory, and a surprisingly low A1C of 4.9% (30 mmol/mol) was found. Interestingly, this level of A1C corresponded nicely with multiple random in- hospital plasma glucose measurements undertaken during the previous 10 years (from 79.2 to 111.6 mg/dL [4.4–6.2 mmol/L]). Again, a falsely elevated A1C by HPLC because of a possible silent hemoglobin variant was suspected.

Parameters related to erythrocyte turnover, as well as hemoglobin fractions, were normal except for a markedly reduced hemoglobin A₂ (HbA₂) fraction. Gene sequencing of HBA1/HBA2 and HBB genes was performed by Sanger sequencing without any abnormalities. Missense mutations of the HBD gene have earlier been reported to cause erroneous A1C results in patients with initial negative screening for hemoglobinopathies and are commonly associated with a reduced HbA₂ fraction (18). Therefore, a genetic variant of the delta-globin (HBD gene [part of HbA₂]) was considered the most likely cause for the interference observed with HPLC. Unfortunately, analysis of the HBD gene was not available at our facility, but the two younger brothers had similarly reduced HbA₂ fractions. Consequently, the diagnosis of diabetes was rejected, and the patient was informed that her risk of developing diabetes in the future was equal to that in the general population.

In this article, we have presented three cases of misdiagnosed diabetes caused by falsely elevated A1C related to analytical interference. The three patients were identified in a university hospital diabetes outpatient clinic with ∼4,000 patients and were identified within this group of patients between 2019 and 2022.

The risk of erroneous A1C results in the presence of hemoglobin variants is well known and was described in the first recommendations from the ADA and WHO on the use of A1C as a diagnostic marker of diabetes (1,2). However, for individual clinicians, it is a relatively rare phenomenon and, thus, easily missed. Nevertheless, it is crucial for patients that diabetes diagnoses are correct in every case.

A confirmatory A1C measurement is recommended when asymptomatic patients are diagnosed with diabetes. A 2009 international expert report on A1C recommended that the confirmatory test uses the same parameter (i.e., FPG, OGTT, or A1C) as the original test because heterogeneity has been demonstrated among the populations diagnosed with diabetes by the three different methods (2). In many instances, such an approach will result in repeated A1C measurements by the same method. The current ADA guidelines (3) state that A1C interference should be considered when marked discrepancy between A1C and plasma glucose or SMBG results is observed. However, many patients do not routinely perform SMBG (3).

We recommend that SMBG- or CGM-based indicators of glycemic burden should be measured before initiating insulin therapy, which, in retrospect, was not performed sufficiently in the first case presented here.

We have extracted a number of characteristics from the presented cases that may alert clinicians to the (rare) possibility that an initial diabetes diagnosis is incorrect (Table 2). Furthermore, we propose an approach for evaluating these patients to conclusively confirm or reject the diagnosis of diabetes, including analyzing A1C with different methods on the same blood sample and using OGTT, CGM, and genetic testing for hemoglobin variants (Table 2).

In addition to the investigations listed in Table 2, fructosamine is an additional parameter that may be relevant to measure in cases of suspected erroneous A1C. However, fructosamine is not included as a diagnostic criteria of diabetes (3), and it can be difficult to translate into actual glycemic burden and risk for diabetes complications. Hence, fructosamine may be useful for confirming a suspicion of a spurious A1C result in combination with other methods and as being a supplementary analysis in those with hemoglobin disorders.

Of note, various analytical interferences may occur with any method used for the measurement of A1C, but in many cases, such interferences may be discovered by cross-examination with two or more methods. Furthermore, it is well described that there may be discrepancies among the different diagnostic parameters (i.e., A1C, OGTT, and FPG), which can introduce uncertainty regarding the diagnosis (19). Hence, both the biological differences and analytical variation must be taken into account when A1C results by several means of measurement are compared, and repeated measurements may be necessary before a final conclusion can be made.

Furthermore, interfering hemoglobin variants may be difficult to detect, and it is important for clinicians to work closely together with the local laboratory in suspected cases. In addition, additional medical history of possible exogenous substances that can interact with hemoglobin is essential (as in case 2), and in other cases, an extra blood sample for genetic analysis of hemoglobin variants is required before the laboratory can reach a conclusion about the underlying interference.

If analytical interference in a specific A1C assay is confirmed, A1C measured with a different and unaffected method may still be used to monitor glycemic control in the same patient. However, care should be taken if a diabetes diagnosis is made by use of an alternative A1C assay in a patient for whom analytical interference with another method has already been observed; in some cases, it may not be feasible to use A1C for diagnosis at all.

In this article, we have only discussed the risk of misdiagnosis resulting from falsely high A1C results. However, falsely low A1C measurements may also occur, and in such cases, the interference will cause underdiagnosing or suboptimal treatment of diabetes. An example of this could be elevated fetal hemoglobin, which has been demonstrated to introduce falsely low values by some assays.

Finally, we would like to emphasize that, for most patients, A1C is still a reliable, inexpensive, and precise biomarker for the diagnosis of diabetes.

The authors thank the three patients described in cases 1–3 for contributing to this manuscript by providing informed consent regarding the publication of their anonymized medical history.

The salary of K.M.L. was provided by the Steno Diabetes Center Aarhus, which is partially funded by an unrestricted donation from the Novo Nordisk Foundation.

No potential conflicts of interest relevant to this article were reported.

K.M.L. wrote the manuscript. J.S. handled patient inclusion and clinical monitoring. A.W.-L. and A.A. performed specialized biochemical testing, including gene sequencing and its interpretation. All authors critically reviewed the manuscript and approved the final version. J.S. and A.A. are the guarantors of this work and, as such, had full access to all of the data presented and take responsibility for the integrity of the data and the accuracy of the review.