Clinical and biochemical correlates of diabetic nephropathy: a retrospective cross-sectional analysis in an Algerian cohort design, retrospective, observational, cross-sectional study

¹CancerLab Research Laboratory, Faculty of Medicine, University of Abou-Bekr Belkaid, Tlemcen, Algeria, ²Department of Biochemistry, Tidjani Damerdji University Hospital Center, Tlemcen, Algeria, ³Department of Epidemiology and Preventive Medicine, Tidjani Damerdji University Hospital Center, Tlemcen, Algeria, ⁴National Institute of Advanced Training of Midwives, Tlemcen, Algeria

^&Corresponding author
Houssam Boulenouar, CancerLab Research Laboratory, Faculty of Medicine, University of Abou-Bekr Belkaid, Tlemcen, Algeria

Introduction    

Diabetic nephropathy (DN) is one of the most consequential microvascular complications of diabetes mellitus. It develops through a fairly predictable structural sequence - glomerular hypertrophy, mesangial expansion, glomerulosclerosis, and eventually tubulointerstitial fibrosis - and remains the single largest driver of chronic kidney disease (CKD) worldwide [1-3]. Estimates place its lifetime prevalence at 20-40% among people with diabetes [4-6], and once renal function begins to decline, cardiovascular risk rises well before dialysis becomes necessary [2].

Microalbuminuria has long served as the standard non-invasive marker for detecting and tracking DN [7]. The classic trajectory runs from normoalbuminuria to microalbuminuria and then macroalbuminuria, culminating in overt nephropathy and, for a subset of patients, end-stage renal disease [8]. In one well-characterised cohort, Saulnier et al. found that roughly half of diabetic patients remained normoalbuminuric while the rest had already progressed to micro- or macroalbuminuria [9], a distribution that illustrates how much of the at-risk population can be missed if albuminuria is checked too late or too infrequently.

Albuminuria alone does not capture the full metabolic picture. Beyond glycaemic indices such as HbA1c, a growing body of work has profiled additional early biomarkers of DN, from routine laboratory panels to newer tissue-specific candidates [10]. Considerable research is also directed at tissue-specific markers - urinary proteomic panels, interleukins, tubular injury markers such as NGAL and KIM-1 [11,12] - and Umanath and Lewis have summarised how these fit within the broader management framework of glycaemic control, blood-pressure control, and renin-angiotensin-system blockade that anchors current DN therapy [13]. In everyday clinical practice, however, particularly where laboratory resources are limited, it is the routine biochemical panel that continues to carry most of the diagnostic weight, an approach consistent with the cross-disciplinary biomarker framework proposed by Hsu et al. for the CKD Biomarkers Consortium [14], and with reviews emphasising urinary proteomics as a monitoring tool of growing but still largely investigational value [12].

The estimated glomerular filtration rate remains the reference measure of overall kidney function. Following the National Kidney Foundation's K/DOQI framework, eGFR is used to classify CKD into five stages that guide both prognosis and management [15]. In adults, eGFR is most often derived from serum creatinine using validated equations such as the modification of diet in renal disease (MDRD) study equation [16], which incorporates age, sex, and body size, or the older Cockcroft-Gault formula, which instead estimates creatinine clearance directly from age and body weight [17].

Against this background, we set out to characterise the cross-sectional relationships among routine blood markers in three clinically distinct groups seen in Tlemcen, Algeria: patients with generic nephrotic syndrome, patients with uncomplicated diabetes, and patients with confirmed diabetic nephropathy. The nephrotic group was included by design. These patients have heavy proteinuric kidney disease without diabetes and therefore serve as a positive control for advanced proteinuric renal injury: a marker that is disturbed in both nephrotic and diabetic nephropathy patients is probably tracking kidney damage as such, whereas a marker that departs from normal only in diabetic nephropathy is more likely to reflect the diabetic milieu itself. This comparison gives the contrasts between the two diabetic groups a reference point they would otherwise lack. We aimed to describe how these markers differ across groups and to map the pattern of correlations among them, to inform continuous, low-cost monitoring strategies suited to this regional setting.

Methods     

The overall analytical approach follows the cross-disciplinary biomarker framework outlined by Hsu et al. for the CKD Biomarkers Consortium [14].

Study design, setting, and period: this retrospective, observational, cross-sectional study was carried out at two sites in Tlemcen, Algeria: the Nephrology Department of the Tidjani Damerdji University Hospital Center (HUC Tlemcen) and the Agadir Medical Center. Data collection spanned ten months, from March 1 to December 31, 2021. Because the study relied on a defined set of exclusion criteria intended to remove patients with acute confounding conditions (see below), the final analytic cohort of 990 patients represents a selected, rather than exhaustive, sample of the patients seen at these centres during that window.

Study population and sampling: patients were drawn from the general nephrology and diabetes clinic population of Tlemcen. The recruitment pathway and allocation to study arms are summarised in the flowchart (Figure 1). Three groups were defined a priori: uncomplicated diabetic patients (confirmed diabetes mellitus per World Health Organization criteria [18], without documented renal complications); diabetic nephropathy patients (confirmed diabetes with a documented, sustained decline in renal function attributable to diabetic kidney disease); and nephrotic patients (followed in the nephrology service for clinical nephrotic syndrome of any cause, included as a non-diabetic comparator group for renal impairment, that is, as a positive control for advanced proteinuric kidney disease, as explained in the introduction). Patients were excluded if they were pregnant, newly diagnosed with diabetes during the study window (precluding assessment of disease duration), had undergone renal transplantation or already required dialysis, followed an atypical dietary regimen or had an altered muscle mass likely to bias creatinine-based estimates, or had experienced recent major physical trauma. The archived registers did not record the number of patients screened, or the number removed under each exclusion criterion, in a form that could be retrieved for this analysis. Figure 1 therefore lists the exclusion criteria without counts, and the size of the source population from which the 990 analysed patients were drawn cannot be reported.

Data collection: data were gathered through two complementary routes. A structured questionnaire captured demographic and anthropometric information - age, sex, weight, height, marital status, diabetes duration, comorbidities, and family history of diabetes - directly from participants. This was cross-checked and supplemented against the archived medical records held by the nephrology department and the Agadir centre, which also provided historical laboratory values.

Biochemical analyses: blood was drawn by venipuncture in the morning after an overnight fast, from the arm contralateral to any intravenous access to avoid haemodilution artefacts. Samples were processed at the Central Laboratory of Biochemistry and the Biochemistry Laboratory of the Nephrology Department, HUC Tlemcen, within two hours of collection; residual samples were refrigerated at 2-8°C for 5-7 days for quality-control verification. Five biochemical panels were analysed: diabetes profile (fasting glycaemia, HbA1c); lipid profile (triglycerides, total cholesterol); renal profile (urea, creatinine, creatinine clearance, eGFR); ionic profile (sodium, potassium, chloride); and bone/mineral profile (calcium, phosphorus, uric acid). Assays were run on a Siemens Dimension EXL 200 chemistry analyser, a Bio-Rad D-10 HbA1c system (HPLC), and a Siemens RAPIDPoint 500 blood-gas and electrolyte analyser.

Statistical analysis: body mass index (BMI) was calculated as weight in kilograms divided by height in metres squared and classified according to WHO categories [19]. Estimated glomerular filtration rate was estimated with the MDRD equation [16], which was the equation used by the hospital laboratories to report eGFR routinely throughout the study period. Analyses were performed in IBM SPSS Statistics version 23.0 (IBM Corp., Armonk, NY, USA). Continuous variables are reported as mean ± standard deviation or median (interquartile range), depending on distribution, assessed with the Shapiro-Wilk and Kolmogorov-Smirnov tests; variance homogeneity was checked with Levene's test. Between-group comparisons of categorical variables used the Pearson chi-square test. For continuous variables, the independent-samples t-test or one-way ANOVA was used when normality and variance-homogeneity assumptions held; otherwise, the Mann-Whitney U and Kruskal-Wallis tests were applied. Bivariate associations were assessed with Pearson's correlation coefficient, classified as weak (0.00 ≤ |r| < 0.25), moderate (0.25 ≤ |r| < 0.75), or strong (|r| ≥ 0.75). Statistical significance was set at α = 0.05.

Because this manuscript revision was undertaken without renewed access to the patient-level dataset, we supplemented the originally reported tests with effect-size estimates (Cohen's d, eta-squared, Cramér's V) computed directly and transparently from the published group-level statistics (means, standard deviations, and counts underlying Table 1, Table 2, and Table 3). These summary-level analyses quantify the magnitude of the group differences reported in the results and should be read as complementary to, rather than a substitute for, analyses performed on the original patient-level data. An earlier draft of this manuscript referred to a multivariate logistic regression that adjusted continuous and categorical variables for age; because no such regression output was ever generated or verifiable against the underlying data, that claim has been removed from this version rather than retained or reconstructed. As a consequence, none of the group comparisons or correlation coefficients reported below are adjusted for age, diabetes duration, or any other potential confounder.

Ethical considerations: the study was conducted in accordance with the Declaration of Helsinki and approved by the Institutional Ethics Committee of the Tidjani Damerdji University Hospital Center, Tlemcen, Algeria. Given its retrospective, observational design and reliance on de-identified, pre-existing records, the requirement for prospective informed consent was waived by the ethics committee. All records were anonymised before analysis.

Results     

Study population and demographic characteristics: patient allocation across the three clinical arms is summarised in the study flowchart (Figure 1). A total of 990 patients met the inclusion criteria: 253 nephrotic patients (25.6%), 517 patients with uncomplicated diabetes (52.2%), and 220 patients with diabetic nephropathy (22.2%). The overall cohort showed a female predominance - 570 women (57.6%) versus 420 men (42.4%), a ratio of 1.35. Full demographic data are given in Table 1. Age differed significantly across the three groups (p < 0.001): patients with diabetic nephropathy were the oldest on average (65.9 ± 12.1 years), followed by uncomplicated diabetic patients (62.0 ± 11.6 years) and nephrotic patients (53.7 ± 19.5 years). Body composition also differed markedly by group (p < 0.001): three-quarters of nephrotic patients (75.9%) fell within the normal BMI range, whereas both diabetic cohorts skewed toward overweight and obese categories (mean BMI 28.7 ± 5.1 kg/m² in uncomplicated diabetes and 27.5 ± 6.5 kg/m² in diabetic nephropathy). A positive family history of diabetes was reported by 47.8% of the uncomplicated diabetic group and 40.9% of the diabetic nephropathy group (p < 0.001). Comorbidities were common across all three groups but disproportionately so in diabetic nephropathy (90.9% versus 57.7% in uncomplicated diabetes and 59.7% in the nephrotic group; p < 0.001). Diabetes duration was the strongest discriminator between the two diabetic groups: patients with nephropathy had lived with diabetes for 14.0 ± 8.5 years on average, compared with 9.8 ± 7.4 years in the uncomplicated group (p < 0.001, Cohen's d = -0.54).

Clinical and biochemical profile: Table 2 summarises the biochemical comparisons across groups. Fasting glycaemia, blood urea, creatinine, creatinine clearance, eGFR, albumin, and uric acid all differed strongly across the three groups (all p < 0.001). Hemoglobin A1c (p = 0.006), triglycerides (p = 0.012), total cholesterol (p = 0.016), and phosphorus (p = 0.013) also differed significantly, while sodium, potassium, chloride, and calcium did not reach significance. Renal staging by eGFR diverged sharply between groups (p < 0.001, Cramér's V = 0.49). Among uncomplicated diabetic patients, 90.7% remained in CKD stage I-II, with the remainder in stage III; none had progressed to stage IV or V. The nephrotic group showed intermediate attrition, with 43.9% in stage I-II and 27.3% already in stage V. Renal deterioration was most advanced in the diabetic nephropathy group: only 13.6% remained in stage I-II, while 45.9% had reached stage V, end-stage renal disease.

Effect sizes: diabetes versus diabetic nephropathy: comparing the two diabetic groups directly, effect sizes were largest for the renal-function markers: eGFR (Cohen's d = 2.50), creatinine clearance (d = 1.95), and albumin (d = 1.71) all showed very large between-group differences, followed by creatinine (d = 1.88) and uric acid (d = 1.23). Effects were moderate for phosphorus (d = 0.73) and calcium (d = 0.48), and small-to-moderate for triglycerides (d = -0.59), total cholesterol (d = -0.26), and fasting glycaemia (d = 0.19); HbA1c did not differ meaningfully between the two diabetic groups once compared directly (d = 0.08, p = 0.28), suggesting that chronic glycaemic control alone does not clearly separate patients who have already developed nephropathy from those who have not.

Bivariate correlations between blood markers: the full matrix of pairwise Pearson correlations reaching significance (p < 0.05) within each of the three groups is given in Table 3. In nephrotic patients, urea correlated positively with creatinine (r = 0.422), phosphorus (r = 0.609), and uric acid (r = 0.641), and negatively with albumin (r = -0.460) and sodium (r = -0.519); the most striking association was a near-inverse relationship between sodium and phosphorus (r = -0.957, p = 0.001), consistent with tubular regulatory failure in advanced renal impairment. Sodium and chloride correlated strongly and positively (r = 0.721), as expected physiologically, while calcium correlated negatively with total cholesterol (r = -0.703) and creatinine (r = -0.611). In uncomplicated diabetic patients, urea rose with creatinine (r = 0.566, p < 0.001) and fell with both creatinine clearance (r = -0.477) and eGFR (r = -0.625); fasting glycaemia correlated with HbA1c as expected (r = 0.665, p < 0.001), and BMI correlated positively with creatinine clearance (r = 0.541, p < 0.001), a relationship that likely reflects the hyperfiltration seen in early, less advanced diabetic kidney disease rather than a protective effect of higher body mass. In patients with diabetic nephropathy, urea again tracked creatinine and uric acid positively while correlating negatively with creatinine clearance, eGFR, sodium, and albumin; the strongest association in this group was between triglycerides and total cholesterol (r = 0.821, p < 0.001), reflecting the dyslipidaemia that commonly accompanies advanced diabetic microvascular disease. Age correlated negatively with HbA1c (r = -0.640) and calcium (r = -0.601) and positively with uric acid (r = 0.530) in this group, a pattern consistent with older nephropathy patients presenting with more advanced, uric-acid-elevating renal impairment rather than with better glycaemic control per se.

Discussion     

The global rise in diabetes prevalence continues to drive a parallel increase in chronic kidney disease [20]. Diabetic nephropathy typically begins silently and, over years of suboptimal metabolic control, becomes a leading indication for dialysis and renal transplantation in many health systems; roughly 20-40% of people with diabetes will develop some degree of nephropathy over the course of their disease.

Our cohort adds a regional data point to this picture. Age emerged as an important correlate of nephropathy, with the diabetic-nephropathy group markedly older on average than the other two - a pattern that sits somewhat at odds with the youth-onset ESRD risk described by Pavkov et al. in Pima Indian populations with early type 2 diabetes [21], and is more consistent with a slower, cumulative form of renal injury tied to disease duration [22]. Diabetes duration itself was the clearest discriminator between the uncomplicated and complicated diabetic groups (14.0 versus 9.8 years, d = -0.54), reinforcing chronicity as a central driver of nephropathy risk - a relationship also tracked longitudinally in the UK Prospective Diabetes Study (UKPDS 64) [23] - and in line with the broader management perspective set out by Umanath and Lewis [13] and the consensus recommendations of Tuttle et al. [24].

Both diabetic cohorts were, on average, overweight. Glycaemic control, measured by HbA1c, was similarly poor in both diabetic groups (8.3% and 8.2%) and did not differ meaningfully between them once compared directly (d = 0.08). This suggests that, at least in this cross-sectional sample, HbA1c alone does not clearly separate patients who already have nephropathy from those who do not, even though chronically elevated HbA1c is well established as a driver of microvascular complications generally [25]. Two explanations seem plausible for this apparent paradox. The first is a matter of timing. A single HbA1c value reflects glycaemia over the preceding two to three months, whereas nephropathy results from exposure accumulated over many years. Long-term follow-up of the UKPDS and DCCT/EDIC cohorts showed that the renal benefit of earlier intensive control persisted well after HbA1c levels in the two treatment arms had converged, the phenomenon described as a legacy effect or metabolic memory [26,27]. By the same reasoning, patients in our nephropathy group may carry the imprint of past hyperglycaemia that their current HbA1c no longer shows. The second explanation is analytical. In advanced CKD, and especially in stage IV-V disease, which accounted for almost two-thirds of our nephropathy group, shortened erythrocyte survival, renal anaemia and its treatment with iron or erythropoiesis-stimulating agents, and carbamylation of haemoglobin with some assay methods can all bias HbA1c, usually downwards [28]. Hemoglobin A1c may therefore understate true glycaemia precisely in the patients with the most severe renal disease, which would narrow any difference between the two diabetic groups. Our data cannot distinguish between these two mechanisms; markers less dependent on red-cell turnover, such as glycated albumin or fructosamine, would help to do so in future work. The absence of a Type 2 diabetes-specific stratification of our cohort - an approach that has, in other settings, uncovered biologically distinct diabetes subgroups with different complication profiles [29] - is a limitation we return to below.

The renal markers told a more consistent story. Blood urea and creatinine were both markedly elevated in the diabetic-nephropathy group relative to uncomplicated diabetes, and the effect sizes for eGFR (d = 2.50), creatinine clearance (d = 1.95), and albumin (d = 1.71) were the largest observed in the entire panel - considerably larger than the corresponding effects for the lipid or glycaemic markers, and broadly consistent with the worldwide risk-factor picture described by Gheith et al. [30]. The strong inverse sodium-phosphorus relationship in nephrotic patients (r = -0.957) and the tight triglyceride-cholesterol coupling in diabetic nephropathy (r = 0.821) point to two distinct downstream disturbances - tubular electrolyte handling in the first case, atherogenic dyslipidaemia in the second - that likely deserve separate clinical attention rather than being read as a single composite signal.

The lipid profile of the nephropathy group also matches the dyslipidaemia usually described in advanced CKD. Compared with uncomplicated diabetes, triglycerides were clearly higher (1.70 versus 1.30 g/L), whereas total cholesterol barely differed (1.60 versus 1.51 g/L). This is the pattern expected in uraemia, in which reduced lipoprotein lipase and hepatic lipase activity slows the clearance of triglyceride-rich lipoproteins, so that very-low-density and intermediate-density lipoprotein remnants, which carry cholesterol as well as triglycerides, accumulate in the circulation [31]. When a larger share of plasma cholesterol is carried in these remnant particles, triglycerides and total cholesterol tend to rise and fall together, which offers a reasonable explanation for the correlation of 0.821 observed here. Read in this way, our finding reflects a well-documented pathophysiological pattern rather than a regional peculiarity, although we did not measure HDL or LDL cholesterol or lipoprotein subfractions and therefore cannot confirm the mechanism directly.

We used the MDRD equation for eGFR, in keeping with the original study protocol and, more importantly, with local practice: during the 2021 study window, the biochemistry laboratories of HUC Tlemcen routinely reported eGFR with the MDRD equation. Since this was a retrospective analysis of real-world records, we kept the values that clinicians had actually used to stage their patients rather than recalculating them with another equation, which would in any case have required patient-level creatinine data that were not available for this revision. More recent equations such as CKD-EPI are generally considered more accurate at higher filtration rates, and head-to-head comparisons suggest the choice of equation can meaningfully shift staging classification in diabetic populations [32]. This choice may partly explain why our diabetic-nephropathy cohort's mean eGFR (29.7 mL/min/1.73m²) reads lower than the comparatively preserved values reported by Araki et al. in an earlier-stage cohort (approximately 88 mL/min/1.73m²) [1]. The Cockcroft-Gault formula [17], an older alternative that estimates creatinine clearance rather than a standardised eGFR, was not used here but remains in clinical use in some settings and could be explored in future comparative work. The parallel rise in uric acid in advanced nephropathy mirrors findings from Kim et al. in a Korean veteran cohort [33], consistent with impaired renal handling of purine metabolites as filtration declines. Beyond the biochemical panel itself, management of diabetes in the context of declining renal function also carries specific pharmacological and nutritional considerations that are well summarised elsewhere [25], and that go beyond the scope of the present cross-sectional analysis.

Limitations: several limitations bear on these findings. The cohort was drawn from two centres in a single region over a ten-month window, limiting generalisability. The analysis is cross-sectional and can describe association but not causality. Every comparison and correlation reported here is unadjusted. Without the patient-level data, we could not fit multivariable models, and two confounders are of particular concern: patients in the nephropathy group were older than those with uncomplicated diabetes (65.9 versus 62.0 years) and had lived with diabetes for longer (14.0 versus 9.8 years). Both factors are independently associated with declining renal function, so part of the differences we attribute to nephropathy in eGFR, creatinine, uric acid, or phosphorus, for instance, may reflect age and diabetes duration rather than nephropathy itself. The same caution applies to the within-group correlations, none of which were adjusted for age. In addition, the number of patients excluded at each step could not be retrieved from the archived registers, which limits any assessment of selection bias. Because this revision was carried out without renewed access to the underlying patient-level dataset, the effect sizes presented above were derived from published group summary statistics rather than the original records and should be interpreted as indicative pending confirmation against the raw data. Prospective, multicentre studies with patient-level datasets, adequately powered for multivariable modelling with adjustment for age, sex, and diabetes duration, are needed to determine which of these biomarkers carry independent predictive value for progression to end-stage renal disease.

Conclusion     

In this Algerian cohort, renal deterioration was already advanced by the time diabetic nephropathy was clinically recognised: nearly half of the affected patients (45.9%) had reached stage V CKD, compared with almost none of the uncomplicated diabetic group. Routine glycaemic, lipid, and renal markers differed clearly between groups, and the largest effects were concentrated in the direct filtration measures - eGFR, creatinine clearance, and albumin - rather than in glycaemic control per se. These findings support the systematic use of an integrated, low-cost biochemical panel alongside albuminuria screening in primary and diabetes-clinic care, particularly in settings where more specialised biomarkers are not readily available.

Competing interests     

The authors declare no competing interests.

Authors' contributions     

Conceptualization: Moustafa Sellah. Methodology: Moustafa Sellah, Houssam Boulenouar, Derbali Regagba. Validation: Moustafa Sellah, Houssam Boulenouar, Mohammed Benaouda, Derbali Regagba. Formal analysis: Siham Zouhiri, Nacera Djelti, Moustafa Sellah. Investigation: Siham Zouhiri, Nacera Djelti, Moustafa Sellah, Mohammed Benaouda. Resources: Moustafa Sellah. Data curation: Siham Zouhiri, Nacera Djelti, Moustafa Sellah, Houssam Boulenouar, Derbali Regagba. Writing - original draft: Moustafa Sellah, Houssam Boulenouar. Writing - review and editing: Houssam Boulenouar, Moustafa Sellah. Visualization: Houssam Boulenouar, Moustafa Sellah. Supervision: Moustafa Sellah, Houssam Boulenouar. Project administration: Moustafa Sellah. All authors have read and approved the final version of this manuscript.

Acknowledgments     

The authors thank the nephrology and biochemistry laboratory staff of the Tidjani Damerdji University Hospital Center and the Agadir Medical Center, Tlemcen, for their assistance with data and sample collection.

Tables and figure     

Table 1: demographic and anthropometric characteristics by study group (N = 990)

Table 2: clinical and biochemical characteristics, and CKD staging by eGFR, by study group

Table 3: bivariate Pearson correlations between blood markers, by study group (pairs significant at p < 0.05)

Figure 1: study flowchart showing patient recruitment, allocation to the nephrotic, diabetic, and diabetic-nephropathy groups, and the sequence of biochemical analysis and statistical processing

References