Does Diabetes Alter CSF Total Protein Levels? A Retrospective Cohort Study

Abstract

Background and Purpose:

Elevation of total protein level in cerebrospinal fluid (CSF-TP) in diabetic patients is often disregarded by clinicians. However, existing studies on the topic have significant limitations, and therefore we aimed to explore the relationship between diabetes and CSF-TP in a large database of CSF samples.

Methods:

Retrospective review of all diagnostic lumbar punctures at the Ottawa Hospital between 1996-2016. Patients were excluded if they had elevated CSF cell counts, or a condition known to elevate CSF-TP. Multivariate linear regression modeling considered the effects of age, sex, and diabetes.

Results:

Among 6124 patients (746 with diabetes, 5378 without), mean CSF-TP did not differ significantly between groups (0.39 and 0.35 mmol/L, p = 0.2). When controlled for age and sex, there was no significant effect of diabetes on CSF-TP and no significant correlation between mean serum glucose and CSF-TP (R² = 0.12).

Conclusions:

CSF-TP did not differ significantly between diabetic and non-diabetic groups, once the influence of age and sex was controlled. Elevated CSF-TP should be regarded as pathologic, even in the setting of diabetes.

Introduction

An elevated cerebrospinal fluid total protein (CSF-TP) level may be an important laboratory clue for the detection of infectious or autoimmune diseases of both the central and peripheral nervous systems, as well as tumors of the neuraxis. Clinicians have however tended to question the diagnostic significance of raised CSF-TP in patients with diabetes, based on the assumption that their CSF-TP can be expected to commonly exceed the reference range for the non-diabetic population.^1-3

Most of the literature on CSF-TP in diabetic subjects is of questionable quality. It suffers from major limitations relating to small sample size, lack of age and sex-dependent reference values, outdated laboratory methods and limited diagnostic ascertainment of conditions expected to elevate CSF-TP.^4-9 Moreover, diabetic microvascular complications (polyneuropathy, retinopathy) are often lumped with macrovascular complications (stroke, peripheral vascular disease) under the umbrella of “neurological complications.”

We have recently published age-dependent reference limits for CSF-TP from a large cohort of patients who underwent diagnostic lumbar puncture at the Ottawa Hospital. ¹⁰ In the current study, we analyzed the impact of a criterion-based diagnosis of diabetes on mean CSF-TP as well as the correlation of CSF-TP with the most contemporaneous glycemic level.

Methods

We conducted a retrospective review of patients who underwent diagnostic lumbar puncture at the Ottawa Hospital between 1996-2016. Research ethics board (REB) approval was obtained from the The Ottawa Hospital Research Institute. Data was extracted from The Ottawa Hospital Data Warehouse, as described in a previous publication. ¹⁰

Our initial data set consisted of 19,591 cerebrospinal fluid (CSF) samples acquired through diagnostic lumbar puncture. Cases were first excluded for the following criteria: incomplete laboratory data (n = 3,546), white blood cell count > 5 × 10⁶ / L (n = 4,906), red blood cell count > 50 × 10⁶ / L (n = 2,324), CSF glucose < 2.5 mmol/L (n = 177), age <18 (n = 197), lack of diagnostically conclusive clinical documentation (n = 100), duplicate samples (n = 456), and statistical outliers by the Tukey Method (n = 296). Based on individual chart review, we excluded cases (n = 1,465) that met criteria for at least one of 60 conditions considered as likely cause of elevated CSF-TP (such as multiple sclerosis, acute inflammatory demyelinating polyneuropathy, and CNS infections). ¹⁰ Diabetes was however not an exclusion criterion by itself.

The diagnosis of diabetes was established in our administrative health database based on relevant ICD-10 code and one instance of random serum glucose >11.1 mmol/L. Relevant ICD-10 codes included: E10.*, E11.*, E13.*, E14.* and O24.5*-O24.8*.

For each patient we searched the database for blood glucose values obtained around the time of lumbar puncture. If several blood glucose results were on file, we used the mean glycemia value. Where only a single glucose value was available, the mean was identical to that value.

Multivariate linear regression modeling was conducted, using age, sex, and diabetes as independent variables, and CSF-TP as the dependent variable. Modeling techniques accounted for interactions between co-variates. All statistical analysis was performed using R (R Foundation for statistical computing, Vienna, Austria). ¹¹ Plots of mean glucose vs. age and mean CSF-TP vs. age were generated to confirm the known effects of age, in this data subset. We also generated boxplots of mean CSF-TP according to diabetic status, and linear regression modeling of CSF-TP versus mean serum glucose.

Results

Our final cohort consisted of 6124 patients, of whom 746 had diabetes, and 5378 did not. The diabetic patients were older (mean age 55.7 vs 43.9 years, p < 0.001), although no gender difference was observed (62.9% vs 59.4% female, p = 0.06). Mean CSF-TP did not differ significantly between groups (0.39 and 0.35 gl/L, p = 0.2). Using linear regression modeling accounting for interactions between variables, both age (p < 0.001) and sex (p = 0.007) showed significant effect on mean CSF-TP. However, as demonstrated in Figure 1A, there was no significant effect of diabetes on CSF-TP (p = 0.20) when controlled for age and sex. For mean serum glucose values, 395 patients had more than one measurement (range 1-161 values, median 4). The median interval between lumbar puncture and serum glucose determination was 3 days. In Figure 1B, we explored the relationship between mean serum glucose and CSF-TP (separated according to diabetic and non-diabetic patients) and demonstrated no significant correlation (R² = 0.12). We found a similar lack of correlation when we performed the same analysis using the single glucose value closest to the date of lumbar puncture.

Figure 1.

A, Boxplot of CSF-TP (CSF total protein) versus diabetic status. B, Plot of CSF-TP (CSF total protein) versus mean glucose, stratified by diabetes status.

Discussion

Elevations in CSF-TP result primarily from excess intrathecal immunoglobulin production or abnormal blood-CSF permeability, less frequently from impaired reabsorption. CSF is an ultrafiltrate of plasma, isolated from the general circulation by barriers at the level of both the neuraxis and the spinal nerve roots. Tight junctions between capillary endothelial cells play a key role in limiting the passage of large protein molecules. It would certainly be valid to postulate that the well documented ultrastructural capillary abnormalities of diabetes might affect the blood-CSF barrier. ⁹

In the current study, we analyzed the largest assembled cohort of patients with both diabetes and CSF-TP reference data in the literature. Our diabetic population consisted of patients who met criteria for diabetes and included a combination of those with and without polyneuropathy (including classical diabetic polyneuropathy, and more atypical forms of neuropathy such as radiculoplexus neuropathy or CIDP). Based on our findings, CSF-TP did not differ significantly between diabetic and non-diabetic groups, once the influence of age and sex was controlled.

Our data conflicts with the findings from a 2008 study by Kobessho et al, ⁷ who reported CSF examination in 16 diabetic patients with neurological disorders not expected to elevate CSF-TP (spinocerebellar ataxia, myasthenia gravis, mononeuropathies). Their study suggested a linear relationship between duration of diabetes and CSF-TP. However, the findings in Kobessho et al must be interpreted with caution, as the study did not account for patient age, and sample size was extremely small. ⁷ Interestingly, the Kobessho et al study did not find a relationship between length dependent polyneuropathy findings (absent reflexes, reduced sensation) and CSF-TP elevation. ⁷ Our findings were more in line with a 2002 study by Lozeron et al, ¹² wherein CSF analysis was performed for a series of 100 consecutive diabetic patients presenting with some form of neuropathy. The study showed that elevated CSF-TP was mainly present in patients with either radiculoplexus neuropathies or CIDP, and that mean CSF-TP was significantly lower in subgroup with distal symmetrical polyneuropathy (0.57 vs 1.08 g/L, with no patients in the distal symmetrical polyneuropathy group having a value above 1.0 g/L).

The findings in Figure 1B lend support to the notion that serum glucose should not influence CSF-TP, both in diabetic and non-diabetic patients. In both groups, inspection of the plots reveals that a correlation line cannot be fit (R² = 0.12).

Our study did have several limitations worth noting. First, this was a retrospective study and depended on accurate ICD-10 coding to confirm diabetic status. However, we attempted to refine the definition by mandating elevated random glucose (>11.1 mmol/L) in addition to an appropriate ICD-10 code. Second, we did not examine the reason for lumbar puncture. Third, we did not account for the duration and severity of diabetes in this population, as we did not have access to detailed clinical data. Fourth, we did not separately analyze the subgroup of patients with radiculoplexus neuropathies, or CIDP, in association with diabetes. These disorders, in both diabetic or non-diabetic individuals, likely feature prominent blood-nerve barrier alterations, and might be associated with higher CSF-TP values. Our study of the correlation between blood glucose and CSF-TP can only be considered an exploratory analysis, as individual blood glucose values are subject to wide diurnal and treatment-related fluctuations. Hemoglobin A1C measurements would have provided a more reliable longitudinal measure of glycemia but were too infrequently available in the first half of the 20-year database.

Despite the above limitations, our study should help disprove the myth that diabetes alone might be a valid justification for ignoring a finding of elevated CSF-TP. Therefore, we would recommend that even in the presence of diabetes, clinicians should regard high CSF-TP as potentially pathologic, provided that results are interpreted with age and sex adjusted reference values.