Introduction

Serum creatinine (Crea) and cystatin C (Cys C) are commonly used biomarkers for the assessment of glomerular filtration rate (GFR), with elevated GFR levels usually indicating impaired renal filtration function1,2.

Crea is the end product of creatine and creatine phosphate metabolism and is primarily generated in muscle cells at a constant rate3,4. Its low molecular weight and lack of binding to albumin allow for free filtration through the glomeruli, making renal excretion the primary route of elimination. Therefore, elevated serum Crea usually reflects impaired kidney clearance and potential renal dysfunction5. Cys C is produced by all nucleated cells and is widely distributed throughout human tissues and bodily fluids. Its small molecular size (13.3 kDa) and positive charge at physiological pH enable free glomerular filtration6. Following filtration, over 99% of Cys C is reabsorbed and degraded in the proximal tubules. Its constant production, free filtration, and nearly complete renal catabolism make Cys C an ideal biomarker for GFR7,8.

Both serum Crea and Cys C are commonly measured in health checkups to assess renal function9,10,11. However, based on existing reference intervals (RIs), discrepancies between the results of these two biomarkers are frequently observed in practice. For example, individuals may have elevated serum Crea levels but normal or even low Cys C levels, and other individuals may exhibit normal serum Crea but elevated Cys C levels. These inconsistencies can cause confusion for both clinicians and individuals undergoing physical examinations. Although both markers reflect the GFR, interpreting the results becomes challenging when one marker is outside the RIs but the other is not. This raises the question of whether these individuals actually have kidney damage.

The above discrepancies may be due to the use of inappropriate RIs, highlighting the need to investigate whether the currently used RIs are suitable. In this study, using our hospital’s laboratory information system (LIS) database, we analyzed test data from 36,836 individuals who underwent health checkups at our hospital. We initially found that, although the results for serum Cys C and Crea were positively correlated, the strength of this correlation varied across different age groups. Furthermore, we found that the results for Crea and Cys C were inconsistent for some individuals. Younger people were more likely to have elevated Crea but normal Cys C concentrations whereas older individuals tended to show elevated Cys C but normal Crea concentrations. These findings suggested that age and sex may be related to the discrepancies observed between these two biomarkers. Our further analysis showed that both biomarkers increased with age, with Cys C exhibiting a particularly strong age dependency. We observed significant sex differences in Cys C and Crea levels across the entire study population. Therefore, we established new age- and sex-partitioned RIs using the direct method (2.5th–97.5th percentiles), in accordance with the CLSI EP28 guidelines12. These updated RIs better reflect physiological variations across different ages and sexes and improve the agreement between Cys C and Crea measurements.

Results

Study population characteristics

After excluding 21,868 cases according to the criteria outlined in Fig. 1, a total of 14,968 unique individuals were included. Of these, 69% were female individuals with a median age of 37 years (range 13–91), and 4642 were male individuals with a median age of 39 years (range 16–88). The non-normal age distribution is shown in Supplementary Fig. 1.

Fig. 1
figure 1

Flowchart of participant selection. RIs for UA, urea, ALT, and AST are presented in Supplementary Table 1. Abbreviations: ALT, alanine aminotransferase; AST, aspartate aminotransferase; UA, uric acid; HbA1c, glycated hemoglobin; HDL, high-density lipoprotein; LDL, low-density lipoprotein; Glu, glucose.

Full size image

Consistency of result interpretation using existing RIs across different age and sex groups

The age- and sex-partitioned RIs are used for Crea in our laboratory, and therefore subsequent analyses were stratified accordingly. The results indicated that among individuals over age 60 years, the linear regression relationship between Cys C and Crea was more pronounced, as reflected by higher r² values (Fig. 2). Among female participants aged 20–60 years, 148 had either Crea or Cys C levels exceeding the upper reference interval (URI); however, only eight had both biomarkers exceeding the URI, accounting for just 5.4%. For female individuals older than 60 years, 37 had either Crea or Cys C levels above the URI; notably, none had both biomarkers exceeding the URI. In male participants aged 20–60 years, 173 had either Crea or Cys C exceeding the URI, with only 15 exhibiting both biomarkers exceeding the URI, representing only 8.6%. Similarly, in men aged over 60 years, 81 had either Crea or Cys C levels exceeding the URI, but none had both biomarkers exceeding the URI (Table 1).

The percentage of individuals under 60 years old exceeding the Crea URI was significantly higher than the percentage of those over 60 years old. There may be two explanations for this. The first one is that younger individuals generally have greater muscle mass, which leads to higher Crea production in comparison with older individuals. Another reason is that the RIs for younger people were narrower than those of older individuals, making the former more likely to have results that exceed the URI. The situation for Cys C was the opposite; older people were more likely to exceed the URI, and the percentage of men exceeding the URI was significantly higher than that of women (Table 1). These results indicate that the currently used RIs lack consistency across different age and sex groups and cannot accurately reflect the characteristics of the population.

Table 1 Distribution of inconsistent Crea and Cys C results across sex and age groups based on old and updated ris.
Full size table
Fig. 2
figure 2

Distribution and linear regression analysis of Crea and Cys C across different sexes and ages. The red and black dashed lines represent the old RIs for Cys C and Crea, respectively. The solid black line represents the regression line. The regression equation and squared correlation coefficient (r²) are also displayed.

Full size image

Dependence of serum Cys C and Crea on age and sex

The above results indicated that age and sex are important factors influencing the interpretation of Cys C and Crea results, especially in cases where discrepancies between the two biomarkers are observed. As expected, among individuals with abnormal Cys C or Crea results, younger individuals were more likely to have Crea exceeding the URIs while maintaining normal Cys C levels whereas older individuals were more inclined to have elevated Cys C levels but normal Crea. This trend was consistent across both sexes (Fig. 3). We first investigated the influence of age and found that both serum Crea and Cys C increased with age. Notably, Cys C exhibited a stronger correlation with age, as evidenced by a significantly higher r² value (Fig. 3). This could explain why older people and men were more likely to have Cys C levels exceeding the fixed URI.

To further evaluate the effects of sex and age and determine whether sex- and age-partitioned RIs are warranted, the SDRs across sex groups and age strata (< 40, 41–60, and ≥ 61 years) were calculated. As shown in the first two rows of Table 2, both age and sex had a significant impact on Cys C levels (all SDRs > 0.4), indicating the need for age- and sex-partitioned RIs. For Crea, sex demonstrated a stronger effect (SDR > 0.4) than age, which showed a minimal impact (SDR < 0.4). Subsequent sex-stratified analysis revealed that in male individuals, age exerted a notable effect on Cys C but had a minimal influence on Crea. This pattern was similarly observed in female participants. Finally, within each age subgroup, sex continued to show clinically meaningful effects on both markers, with all SDRs exceeding 0.4 (Table 2).

Fig. 3
figure 3

(A, B) Age comparison between individuals with abnormal Crea and Cys C results. The groups “HCrea NCys C” and “HCys C NCrea” refer to individuals with Crea results exceeding the URIs but normal Cys C levels and those with Cys C results exceeding the URI but normal Crea levels, respectively. (C, D) Scatter plots illustrating the relationship between Crea, Cys C levels, and age. The red dashed lines in (C) indicate the RI for Cys C. The solid black lines depicted in (C) and (D) represent the regression lines.

Full size image
Table 2 SDR in analyzing demographic effects on Cys C and Crea levels.
Full size table

Updated age- and sex-specific RIs for Crea and Cys C

The observed age- and sex-dependency of Cys C and Crea highlights that inappropriate non-age- and sex-partitioned RIs may contribute to the discrepancies described previously. Currently, the RIs for Cys C in our laboratory are not stratified by age or sex. Furthermore, although the RIs for Crea are partitioned by age, they are limited to two broad groups (age ≤ 60 and > 60 years), indicating a need for more refined stratification. Therefore, we established continuous age- and sex-partitioned RIs for both of these using quantile regression analysis, leveraging the large sample size of our dataset. The URIs at each age, separated by sex, are provided in Supplementary Table 2, with corresponding regression lines illustrated in Fig. 4. For Cys C, the URI initially decreased slightly with age, then began to rise after approximately age 40 years. By contrast, the age-related effect on Crea is less pronounced, with values remaining relatively stable across age groups, which aligns with the findings using SDR analysis (Table 2). Because continuous regression-based RIs are challenging to integrate into the LIS, and considering age-related fluctuations of the upper URIs (shown in Fig. 4) as well as the SDR values for Cys C and Crea, we used 10-year age bins for Cys C and four clinically relevant age strata (< 18, 18–50, 51–70, ≥ 71 years) for Crea to calculate nonparametric URIs (Table 3). Notably, the sample sizes for both the age groups ≤ 18 years (n < 10) and > 71 years fell below the CLSI EP28-recommended minimum of 120 participants. Consequently, the derived RIs for these age strata, particularly for the cohort aged ≤ 18 years, should be interpreted with caution owing to limited representativeness.

Based on the updated RIs, differences in the percentage of abnormal Cys C and Crea results across age and sex groups were significantly reduced (Table 1). This demonstrates that the new RIs exhibited better consistency across different ages and sexes, allowing for a more accurate reflection of the population characteristics. Notably, the percentage of individuals with both abnormal Cys C and Crea results also increased, suggesting that appropriate RIs could reduce the inconsistency between these two biomarkers in reflecting GFR (Table 1).

Fig. 4
figure 4

Partitioned upper and lower RIs for serum Crea and Cys C. RIs are indicated by the red lines, and individual data points are represented by gray circles.

Full size image
Table 3 Age- and sex-partitioned RIs for serum Crea and Cys C.
Full size table

Discussion

Skeletal muscle is the primary source of Crea; therefore, owing to lower muscle mass and metabolism, elderly individuals produce less Crea, leading to lower serum Crea concentrations. However, most studies, including ours, show that serum Crea actually increases with age13. A possible explanation for this is that whereas Crea production decreases with age due to reduced muscle mass, aging also results in a natural decline in GFR14,15. The reduced GFR decreases the Crea clearance, leading to an increase in serum Crea. Thus, the combined effects of decreased production and reduced clearance cause a slight increase in serum Crea with age. Sex- and age-partitioned RIs are widely used to assess levels of Crea. Notably, such RIs divide age into broad intervals, which may not adequately capture the full impact of age on serum Crea. For example, we found that the percentage of individuals under age 60 years who exceeded the URIs was significantly higher than that of individuals over age 60 years, based on the RIs used in our laboratory. This suggests that current RIs may potentially lead to an overestimation of GFR in younger populations or underestimation in older people.

Cys C has been considered an ideal biomarker for evaluating GFR, and its serum concentration is nearly entirely dependent on kidney function. Although many studies have reported that age influences serum Cys C concentration13,16,17 both age-partitioned RIs and non-age-partitioned RIs are currently used in clinical practice, such as the non-age-partitioned 0.51–1.09 mg/L value used in our laboratory and the age-partitioned RIs (18–49 years: 0.63–1.03 mg/L, ≥ 50 years: 0.67–1.21 mg/L) used by Mayo clinic laboratories). Similarly, our study showed that Cys C is significantly correlated with age, which was even more pronounced than the correlation with Crea. This suggests that using fixed RIs may lead to the underestimation or overestimation of GFR levels in certain populations. In fact, all nucleated cells in the human body produce Cys C at a constant rate, and its production remains stable with aging. The question therefore arises as to why its serum level increases with age. This is likely due to the natural decline in GFR that occurs with aging14,15. Although Cys C production remains relatively constant, the body’s ability to excrete it decreases, resulting in elevated serum concentrations with increased age. This also explains why Cys C is more strongly correlated with age than Crea.

Regarding the influence of sex on Cys C levels, some studies have found no significant difference between male and female individuals9,18. However, our findings revealed higher serum Cys C concentrations in male than in female participants, which is consistent with most published studies13,16,19,20,21,22. Notably, the influence of sex on serum Cys C levels has been inconsistently reported across studies. Although most studies, including ours, have observed higher Cys C concentrations in male individuals, some studies have reported higher levels in female participants17,23. These discrepancies may be attributed to differences in study populations and the measurement systems used in each study. Additionally, our study showed that the influence of sex persisted across the entire age range whereas many previous studies did not find a significant influence of sex among older individuals (age > 50 or > 60 years)13,17,19,21. One possible explanation is that earlier studies had limited sample sizes and included relatively few elderly individuals, which may have prevented the complete capture of sex-related differences in this age group. For example, most of these studies had a total sample size of fewer than 1000 participants. Compared with those studies, our study has several key strengths. (1) By leveraging a modern hospital LIS and China’s large population base, we screened 14,968 qualified individuals from the health checkup population. Most age subgroups met the CLSI-recommended minimum sample size of ≥ 120 for RI development. By contrast, earlier studies typically included smaller cohorts, making it difficult to establish reliable age- and sex-partitioned RIs. Even in studies that attempted age stratification, the subgroup sample sizes often failed to meet the CLSI EP28-A3c guidelines13,22. (2) Compared with similar recent studies that also used advanced LIS to screen data from health check-up populations16,17 our study applied more stringent and comprehensive inclusion and exclusion criteria. Specifically, we excluded individuals with abnormal results for liver function, blood glucose, blood lipids, and certain renal biomarkers (urea and uric acid), based on real laboratory test results rather than self-reported health status. As a result, only 14,968 participants (40%) out of 36,836 self-reported healthy individuals were included, based on rigorous filtering using eight objective laboratory test results. Contrarily, for example, the study by Ming Ji et al. included 10,640 participants (72%) from 14,736 self-reported healthy individuals, with exclusions based on fewer laboratory results and less stringent thresholds (only 4096 excluded)16. (3) In addition to establishing age- and sex-partitioned RIs for Cys C and Crea, we also evaluated the consistency between these two markers in assessing renal function among individuals who underwent health check-ups. This analysis may have practical value for clinicians, offering guidance on how to interpret discordant results between these biomarkers.

Notably, although our updated RIs improved the consistency between Cys C or Crea in renal function assessment, a substantial proportion of individuals still exhibited discordant results, with elevated levels of either Cys C or Crea alone. This discrepancy likely stems from the fundamental biochemical differences in these markers, especially molecular size. Evidence shows that Crea and Cys C reflect different aspects of kidney injury owing to their physicochemical differences. Crea (0.113 kDa) mainly indicates the glomerular filtration of small molecules and Cys C (13 kDa), a mid-sized protein, more accurately reflects the filtration of medium-sized proteins and peptides (10–30 kDa)24. This divergence is exemplified in conditions like shrunken pore syndrome where the glomerular filtration barrier selectively restricts mid-sized molecules (e.g., Cys C) while remaining relatively permeable to small solutes (e.g., Crea). Consequently, patients with shrunken pore syndrome often show markedly elevated Cys C levels with near-normal Crea24.

To better interpret Crea and Cys C results, clinicians need to understand the limitations of both biomarkers in assessing kidney function. The complex metabolism of serum Crea introduces several limitations affecting its accuracy as a GFR marker5. These include dependence on liver function and skeletal muscle mass25unmeasurable and saturable proximal tubular secretion26 dietary influences27 and extrarenal clearance pathways. For example, serum Crea may overestimate GFR in individuals with reduced muscle mass due to advanced age or chronic conditions such as liver failure or cancer. Conversely, Crea may underestimate GFR in people with greater muscle mass, such as athletes or bodybuilders, as well as those consuming diets rich in red meat28. By contrast, Cys C offers a more accurate estimate of GFR, especially in patients with muscle loss or chronic diseases29. Although Cys C levels in serum, urine, and other body fluids are not influenced by factors affecting Crea, they may vary owing to other pathological conditions. Elevated Cys C levels have been observed in HIV infection, asthma, hyperthyroidism, and corticosteroid treatment whereas lower levels have been associated with cancer, abdominal aortic aneurysm, neurological inflammatory disorders, and cyclosporine treatment30. Therefore, clinicians should consider these factors when interpreting Crea and Cys C results.

Our study has some limitations. First, unlike some other studies, we did not include information on factors such as body mass index, smoking and drinking habits, and other relevant variables. Therefore, we were unable to evaluate their influence on Cys C levels13,16,21,22. Second, there was an imbalance in the number of male and female participants (69% vs. 31%), although the male sample size (n = 4642) was still sufficient for the analysis. Third, although our study provides comprehensive RIs for most age groups, the statistical power for older (> 70 years) and pediatric (< 18 years) populations was limited owing to small sample sizes (n < 120 and n < 10, respectively). Additionally, we assumed that after excluding individuals with abnormal liver function (ALT, AST), blood sugar (Glu, HbA1c), blood lipids (LDL, HDL), and renal indicators (urea, UA), the remaining population was relatively healthy and suitable as a reference group for Crea and Cys C. Although most participants could be considered healthy under these criteria, we cannot completely rule out the possibility that some may have subtle renal impairment despite normal test results. However, we believe that the proportion of such cases is very small and unlikely to significantly affect our conclusions.

Methods

Study population

We collected test results for serum Crea, urea, uric acid (UA), Cys C, alanine aminotransferase (ALT), aspartate aminotransferase (AST), high-density lipoprotein cholesterol (HDL), low-density lipoprotein cholesterol (LDL), blood glucose (Glu), and glycated hemoglobin (HbA1c) as well as age and sex information from 36,836 individuals who underwent physical examinations at Huashan Hospital, affiliated with Fudan University (Shanghai, China), between June 1, 2021 and August 30, 2024. All participants self-reported no history of major chronic diseases, including hypertension, diabetes mellitus, or dyslipidemia. All data were retrieved from the LIS and patients’ privacy was strictly protected during data use for research. Owing to the retrospective nature of the study, the Ethics Committee of Huashan Hospital, Fudan University waived the need for informed consent and the study was approved by the Ethics Committee of Huashan Hospital, Fudan University (No. KY2023-505). This study was conducted in accordance with the Declaration of Helsinki.

Instruments and measurements

Fasting blood samples were taken from the cubital vein, with participants in the seated position, between 8.00 am and 8.00 pm. All blood samples were analyzed in the clinical laboratory of Huashan Hospital. Serum Crea levels were determined using a creatininase method (Hitachi LABOSPECT 008 AS, Hitachi High-Tech Corporation, Tokyo, Japan) with an analytical range of 7–8840 µmol/L. Serum Cys C levels were measured with a Latex-Enhanced Immunoturbidimetric Assay (Hitachi LABOSPECT 008 AS) with an analytical range of 0.25–7.80 mg/L. In our laboratory, the age- and sex-partitioned RIs for Crea are as follows: for male individuals aged 20–60 years, 57–97 µmol/L; for male individuals over 60 years, 57–111 µmol/L; for female individuals aged 20–60 years, 41–73 µmol/L; and for female individuals over 60 years, 41–88 µmol/L. These intervals were derived from the reagent instructions, based on data from 524 healthy individuals, including 262 female individuals, and was previously verified by our laboratory. The RI of Cys C is 0.51–1.09 mg/L, which is not specific to sex or age and was developed using data from 124 healthy individuals.

Participants’ blood Glu, UA, urea, ALT, AST, HDL, and LDL were all measured using a Hitachi 7600 Automatic Biochemical Analyzer and Hitachi LABOSPECT 008 AS (Hitachi High-Tech Corporation, Tokyo, Japan). HbA1c was measured using a MEDCONN Q8000 (MEDCONN, Shanghai, China). All measurements were performed after successful daily quality control at two concentration levels and were conducted by professional clinical laboratory personnel. All information on measurements of those biomarkers, including the measurement principle, instruments, reagents, calibrators, quality controls, RIs, and linear ranges, is provided in Supplementary Table 1.

Data screening

To identify the reference population for the subsequent investigation of serum Crea and Cys C, we excluded individuals with abnormal results for blood Glu, HbA1c, HDL, LDL, UA, urea, AST, and ALT. Abnormal results were defined as values outside the RIs currently used by our laboratory. Ultimately, 14,968 individuals with normal blood lipid biomarkers, liver function biomarkers, Glu, HbA1c, UA, and urea levels were included, of which 69% were female individuals. The data screening process is illustrated in Fig. 1.

Statistical analysis

For the included individuals, ordinary linear regression was applied to evaluate the correlation between Cys C and Crea. We calculated and compared the percentage of individuals in the study population with Cys C above the old RI, Crea above the old RI, and both above the updated RI. To determine whether there was a significant difference in age between individuals with abnormal Crea or Cys C results, non-parametric group comparisons were performed using the Mann–Whitney U-test. To evaluate the need to establish sex- and age-partitioned RIs, we calculated standard deviation ratios (SDRs) using one-way analysis of variance based on the method proposed by Ichihara et al.31,32 which has been widely adopted in related studies33. The SDR is defined as the ratio of between-group to within-group SD. This serves as a quantitative measure of how much grouping factors, such as sex or age, contribute to the overall variation observed in test results. This metric helps to determine whether differences between groups are large enough to exceed the natural biological variation among individuals and the variability introduced by measurement error.

Briefly, the between-sex variation (SD sex) was calculated as the SD of the mean values for each sex group relative to the grand mean of the combined population. To account for variability within individuals, the between-individual variation (SD bet−indiv) adjusted for sex effects was calculated as the average of the within-sex SDs, specifically the SDs calculated separately within male and female groups. Finally, the SDR was computed as the ratio of SD sex to SD bet−indiv, reflecting the relative importance of sex as a grouping factor in test result variation:

$$\:SDR=\frac{{SD}_{sex\:\left(or\:age\right)}}{{SD}_{bet-indiv}}$$

A similar approach was used to calculate the SDRs of age group, with age stratified as ≤ 40, 41–60, and ≥ 61 years. According to the relevant literature, we adopted an SDR threshold of 0.4, which was used to establish the variation-partitioned RIs in Asian populations31,34. On the basis of significant correlations and SDRs observed among age, sex, and levels of Cys C and Crea, age- and sex-partitioned RIs were established using quantile regression analysis. The direct method (2.5th–97.5th percentiles) was applied in accordance with CLSI EP28 guidelines12. Owing to the challenges of integrating continuous regression-based RIs into LIS, we further established age-binned intervals: 10-year intervals for Cys C and four age intervals for Crea (≤ 18, 18–50, 51–70, ≥ 71 years). The distribution of abnormal results across different sexes and age groups was recalculated and compared using updated RIs. All analyses were performed using Microsoft Excel 2016 (Microsoft Corporation, Redmond, WA, USA), GraphPad Prism 9.5 (GraphPad Software, San Diego, CA, USA), and MedCalc 18.11.6-64-bit (Mariakerke, Belgium).