Patient result monitoring of HbA_1c shows small seasonal variations and steady decrease over more than 10 years

Introduction

It is important that patient results are as accurate as possible. Thus, there is a need for quality control of the test results. The test results are usually monitored by internal controls usually applying the Westgard rules and external quality assurance (EQA) programs [1, 2]. The internal quality control materials need to have a long shelf life which means that they have been treated to ensure that they are stable over time, and they may also contain non-human additives [3]. This could lead to commutability problems [4, 5]. The problem for EQA programs is that there is a need for large sample volumes of stable materials that can be distributed to the laboratories [6]. The larger EQA programs have hundreds of participants, and each participant will receive 0.5–1 mL of sample which adds up to several hundred mL. In most cases it is not possible to obtain and ship such large volumes of fresh genuine human samples. The EQA providers thus often rely on outdated surplus human materials from routine testing. To increase the stability, the samples are often frozen or freeze dried and stabilizers are added. The problem for the EQA organizations and the laboratories is that the preparation may be commutable with some reagents/instruments but not all [2, 7]. The lack of commutability may lead to higher or lower assay results for some of the reagent-instrument combinations.

Even if internal quality controls and EQA samples will continue to be the basis for quality control of laboratory test results, there is a need for other ways to monitor our methods [8]. One such possibility is the use of continuous monitoring of the patient results [3, 9]. Current laboratory information systems are able to extract patient median/mean values for a specific analyte.

If there are sufficiently large number of samples analyzed, the center part of the test result distribution is usually stable over time providing that the analyte has a stable calibration. Median and mean values are usually more stable than e.g., the 2.5 and 97.5 percentiles. If the patient results show an upward or downward trend or shift, this should be further investigated regardless if the internal quality controls or EQA samples show the same pattern. Recently, the Swedish EQA organization Equalis started a new program for patient median values. The participants report regularly their patient median results for some of the high-volume analytes. This allows for interlaboratory comparisons.

Hemoglobin A1c (HbA_1c) was used as a model for patient median value monitoring over time at our laboratory. There are several different methods available for measuring HbA_1c such as boronate affinity chromatography, capillary zone electrophoresis (CZE), enzymatic assays, immunoassays, and ion exchange high-performance liquid chromatography (HPLC) [10, 11]. A previous study has shown good agreements between the clinically used HbA_1c methods in Sweden [11]. There are also national quality targets for the analyte to ensure a low method variation [12]. HbA_1c is a high-volume assay which assures a more accurate determination of patient median values. In 2010, the Swedish laboratories changed from a mono S calibration of HbA_1c to the IFCC calibration in mmol/mol. Changes from one calibration to another is often accompanied with an increased risk of calibration problems. HbA_1c results were collected from the start of reporting of IFCC calibrated test results by the laboratory and data was collected up to August 2022.

The aim of the study was to document any changes over time and to investigate possible seasonal variation on HbA_1c results. The median value is the most frequently recommended variable to monitor. The aim of the study was also to compare the changes in median values with changes in mean and quartile values.

Materials and methods

Results

Changes in HbA_1c results over time

The population consisted of 335,894 females (46.5 %) and 386,659 males (53.5 %). The median age was 65.6 years (interquartile range 53.1–74.5 years, total range 0.02–109 years). All HbA_1c test results were sorted according to year from 2011 to 2021 (Figure 1 and Table 1). There was a slight steady decrease over time. From 2011 to 2021 the decrease was for mean values 9 %, median 10.2 %, lower quartile 7.9 %, upper quartile 11.2 %, 0.10 percentile 6.6 % and 0.90 percentile 9.3 %. The upper quartile and 0.90 percentile decreased over the study period by 7 mmol/mol (2.8 %) while the 0.10 percentile decreased by 2 mmol/mol (2.3 %) and the lower quartile decreased by 3 mmol/mol (2.4 %) (Table 1). No clear shift in HbA_1c levels was observed due to the shift in methods. Mean, median, lower, and upper quartiles and 0.10 percentile and 0.90 percentile showed an even change over time. Max and min values showed a higher variation between years (Table 1).

Figure 1:

HbA_1c values over time. The data is presented as lower quartile, median, mean, and upper quartile for each year 2011–2021.

Table 1:

Number of HbA_1c results reported per year and mean, median, min, max, lower, and upper quartiles and 0.10 and 0.90 percentiles for each year.

	Valid n	Mean	Median	Min	Max	Lower quartile	Upper quartile	Percentile 0.10	Percentile 0.90
2011	39,996	52	47.9	14	182	39.1	60.8	35.3	75
2012	43,956	51.6	47.3	11	159.5	38.6	60.9	34.8	75
2013	49,633	51.5	46.7	7.6	184	38	60.7	34.5	75.6
2014	53,105	50.4	45	4	189.9	37.1	59.8	33.5	75
2015	60,297	49.9	45	8	233.4	37	59	33.1	73.4
2016	66,009	48.9	44	13.2	186	36	57.6	32.2	72.9
2017	72,045	48	43	12.5	185	36	56	32.2	71
2018	73,889	47.7	43	12	195	36	55	33	69
2019	76,856	48	43	11	181	37	55	33	69
2020	70,850	47.2	43	3.7	191	36	54	33	68
2021	75,663	47.3	43	0	180	36	54	33	68
Difference		5	5		0	3	7	2	7

Correlation between number of tests per month and mean HbA_1c results

There was a correlation between the number of tests per month and the mean HbA_1c value (Spearman R²=0.56) (Figure 2). The mean HbA_1c decreased when the number of tests decreased.

Figure 2:

Correlations between the number of tests per month and mean HbA_1c value.

Discussion

This study demonstrates a practical approach to add patient data into the traditional laboratory QC program of internal and external control materials.

Monitoring laboratory methods using patient medians offers a more robust and clinically relevant approach, addressing the unique characteristics of patient populations and contributing to the overall quality and reliability of laboratory testing. Patient medians capture the central tendency of a patient population and is usually recommended as the primary measure for patient results as medians are more resistant to outliers than e.g., patient means (Figure 3).

Figure 3:

Mean HbA_1c values in the Swedish National Diabetes Register 2010–2022.

In this study the variation in patient HbA_1c results was investigated. The Swedish National Diabetes Register (NDR) is a quality register containing patient information provided by healthcare providers nation-wide and is used as a tool for continuous quality assessment for diabetes care in Sweden [13]. NDR uses HbA_1c as a quality marker for the diabetes care in Sweden [14, 15]. This registration allows the comparison between national trends for HbA_1c with trends observed for the patient HbA_1c values in this study. The parameters studied were mean, median, lower quartile, upper quartile values, 0.10 percentile, 0.90 percentile and min and max values. The inter-month coefficient of variation for the four studied parameters were similar, indicating that it apart from median HbA_1c values it is also possible to use the other parameters for assay monitoring. The addition of lower and upper quartiles increases the HbA_1c range that can be monitored to approximately 36–54 mmol/mol (5.4–7.1 %). Including 0.1, and 0.9 percentiles increased the monitoring range even further to 33–68 mmol/mol (5.2–8.4 %). No sudden shift in HbA_1c levels was observed when changing from immunological detection using Cobas c501 to capillary electrophoretic detection using Capillarys 3. This is in agreement with the internal method validation performed before moving the method to the new platform and previous method comparison [10, 11].

Normally, patients sampled during holiday periods are sicker than the individuals sampled during regular months. In this study the opposite was found with a positive correlation between the number of tests per month and HbA_1c values. Our interpretation of these findings is that the regular monitoring of diabetes patients is reduced during holiday periods e.g. July and the Christmas period. As HbA_1c mainly is used for long-time monitoring of diabetes patients the testing can easily be postponed until after the vacations. In contrast, acute testing of patients with potential diabetes will remain constant thus contributing to a lower median HbA_1c level during the vacation period in Sweden.

When comparing HbA_1c results from the beginning of the study period with the last part of the study period a decline in HbA_1c values was detected regardless of which of the monitoring parameters that was used. The decline is similar to the decline observed in the Swedish diabetes registry [16]. Controlling for sex and Type 1 diabetes duration, mean HbA_1c decreased between 2011 and 2017 in all age cohorts in Sweden (p<0.001) [16]. Thus, the decline in our patient results is not due to a method deviation but reflects the Swedish general trend towards reduced HbA_1c values nationally which is due to earlier detection and more effective treatment of patients with diabetes. The number of diabetes patients and especially type II diabetes patients are increasing worldwide [17]. Lowering HbA_1c levels is often a primary objective in diabetes management because it correlates with reduced risk of long-term complications associated with diabetes. Advancements in diabetes care, including the development of new medications, glucose monitoring technologies, improved insulin formulations, and better education and support for patients, have contributed to better glycemic control for many individuals. Additionally, lifestyle modifications such as dietary changes, regular physical activity, and weight management can also lead to improvements in HbA_1c levels. Regular monitoring, personalized treatment plans, and ongoing support from healthcare providers are crucial for optimizing diabetes care and achieving target HbA_1c levels. Early detection and treatment will reduce chronic damage due to diabetes. The increasing number of test results per year and the tendency that the change over time was less pronounced for the 0.1 percentile and lower quartile supports the interpretation that the reduction was due to earlier detection of diabetes patient or more efficient treatment of diabetes patients leading to reduced HbA_1c values especially in the upper HbA_1c ranges.

Patient median monitoring can help in early detection of methodological issues that might affect test accuracy and precision. Timely identification allows for corrective actions to be taken before widespread issues occur. The patient median results are used for internal QC control. If results from the external quality assurance program indicates a problem, we compare the results with the patient median before we take any actions. We also use patient medians to verify that batch changes of reagents have not caused deviations in test results. The patient median can also be used to verify that there has not been any shift in calibration over time in clinical studies that are ongoing for several months/years.

Monitoring laboratory methods with patient medians helps in identifying and correcting systematic errors that may affect diagnostic accuracy. This ensures that test results are reliable and can be confidently used by healthcare professionals for patient diagnosis and treatment decisions. A limitation of the use of patient medians is that it is based on the assumption that the sampled patient groups are stable. If the number of assays per year is limited a shift in the use of a specific assay could influence the patient median values. For instance, creatinine is mainly used as a GFR marker in very large volumes. The patient median seems to be very stable. In contrast, the GFR marker cystatin C is sometime used in primary care exclusively for very old patients where creatinine is considered by the GPs as less reliable. This will create a bias when comparing patient medians with other hospitals. Also, changes over time e.g. reduced HbA_1c median values during July as shown in this study could cause interpretation problems (Table 2).

Conclusions

In conclusion, monitoring laboratory methods using patient medians, means, quartiles and 0.1 and 0.9 percentiles offers a more robust and clinically relevant approach to method monitoring. The data shows that the HbA_1c has a small seasonal variation and that the HbA_1c method has had a stable calibration between 2010 and 2022. The decline over time is similar to the decrease observed in the Swedish National Diabetes Register.