Diagnostic Test Accuracy of Urine C-peptide Creatinine Ratio for the Correct Identification of the Type of Diabetes: A Systematic Review

Objective: To examine the accuracy of urine c-peptide creatinine ratio (UCPCR) for identifying the type of diabetes in appropriate clinical settings. Design: Systematic review of test accuracy studies on patients with different forms of diabetes. Data sources: Medline, Embase and Cochrane library databases from 1 January 2000 to 15 November 2020. Eligibility criteria: Studies reporting the use of UCPCR for diagnosing patients with type 1 diabetes mellitus (T1DM), type 2 diabetes mellitus (T2DM) and monogenic forms of diabetes (categorized as maturity-onset diabetes of the young [MODY]). Study selection and data synthesis: Two reviewers independently assessed articles for inclusion and assessed the methodological quality of the studies using the Quality Assessment of Diagnostic Accuracy Studies-2 tool, with input from a third reviewer to reach consensus when there was a dispute. Meta-analysis was performed with the studies reporting complete data to derive the pooled sensitivity, specificity and diagnostic odds ratio (DOR), and narrative synthesis only for those with incomplete data. Results: Nine studies with 4,488 patients were included in the qualitative synthesis, while only four of these (915 patients) had complete data and were included in the quantitative synthesis. All the studies had moderate risk of bias and applicability concerns. Meta-analysis of three studies (n=130) revealed sensitivity, specificity and DOR of 84.4% (95% confidence interval [CI] 68.1–93.2%), 91.6% (82.8–96.1%) and 59.9 (32.8–106.0), respectively, for diagnosing T1DM using a UCPCR cut-off of <0.2 nmol/mmol. For participants with T2DM (three studies; n=739), UCPCR >0.2 nmol/mmol was associated with sensitivity, specificity and DOR of 92.8% (84.2–96.9%), 81.6% (61.3–92.5%) and 56.9 (31.3–103.5), respectively. For patients with MODY in the appropriate clinical setting, a UCPCR cut-off of >0.2 nmol/mmol showed sensitivity, specificity and DOR of 85.2% (73.1–92.4%), 98.0% (92.4–99.5%) and 281.8 (57.5–1,379.7), respectively. Conclusions: Based on studies with moderate risk of bias and applicability concerns, UCPCR confers moderate to high sensitivity, specificity, and DOR for correctly identifying T1DM, T2DM and monogenic diabetes in appropriate clinical settings. Large multinational studies with multi-ethnic participation among different age groups are necessary before this test can be routinely used in clinical practice. Study registration: Protocol was registered as PROSPERO CRD42017060633.

Identifying the type of diabetes correctly can be difficult, especially in adults, because of the heterogeneity in the clinical presentation.
However, it is important to accurately diagnose the type of diabetes for its clinical, prognostic, therapeutic and psychosocial implications. The clinical characteristics, diagnostic work-up, therapies and complications of each form of diabetes are different, and therefore, physicians should accurately characterize diabetes after diagnosis.
Proinsulin is formed by the human pancreatic β cells, which is cleaved into insulin and C-peptide in equimolar quantities. 1 Therefore, plasma levels of C-peptide can be used as a surrogate marker of endogenous insulin secretory capacity. 2 Type 1 diabetes mellitus (T1DM) results from autoimmune destruction of the β cells, leading to absolute deficiency of insulin within a few years of disease onset, resulting in very low or unmeasurable plasma insulin and C-peptide levels. Conversely, in type 2 diabetes mellitus (T2DM), insulin deficiency is relative rather than absolute. Monogenic diabetes is clinically heterogeneous, but is typically an autosomal-dominant, non-insulin-dependent diabetes lacking autoantibodies, also known as maturity-onset diabetes of the young (MODY). 3 In all types except T1DM, plasma C-peptide will be detectable.
Therefore, assessing plasma C-peptide levels can be useful in both identifying diabetes subtypes and planning management strategies. 3

touchREVIEWS in Endocrinology
However, there are practical difficulties in measuring plasma C-peptide levels, mainly the requirement that specimens be kept in ice because of the short biological half-life (30 minutes). 3 As such, collection and transport of samples between outpatient clinics, home settings and testing laboratories is challenging.
C-peptide is largely metabolized by the kidney through glomerular filtration and tubular uptake, with only 5% of the total amount produced excreted in the urine. 4 Urinary C-peptide remains stable at room temperature for up to 72 hours when preserved in boric acid, and can be easily transported to testing laboratories, even from remote settings. 3 Therefore, urine C-peptide becomes an attractive and noninvasive alternative for testing β-cell function and insulin production capacity. To avoid having to collect urine for 24 hours to estimate β-cell reserve, spot urine C-peptide creatinine ratio (UCPCR) is an easy alternative option. 5

Methods
We performed the study as per the guidelines specified in the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA). 11 We searched the Medline, EMBASE and Cochrane Library databases to first-degree relative of that parent; lack of characteristics of T1DM (lack of islet-cell antibodies, low or no insulin requirement 5 years after diagnosis ± serum C-peptide levels >200 pmol/L); and lack of classical features of T2DM (marked obesity or acanthosis nigricans). 13 The remainder were considered to be diagnosed with T2DM.

Data extraction and quality assessment
Two independent investigators (JMP and BS) independently assessed all the titles and abstracts to identify studies eligible for inclusion in the review. The same authors then reviewed the full texts of these studies for eligibility to be included in the DTA study, and extracted data from the eligible studies using a standardized data-extraction form. The form included the following characteristics of each trial: first author's name; year of publication; study population characteristics, including sample size, geographical location, mean age and sex; and diagnostic criteria, including screening and confirmatory tests for the type of diabetes. Differences between reviewers were resolved by discussion with the third reviewer (APA).
The risk of bias and applicability of the identified studies were assessed by two independent reviewers (JMP and CJF) using the modified Quality Assessment of Diagnostic Accuracy Studies-2 (QUADAS-2) criteria, for patient selection, performance of the index test, performance of the reference test, and flow and timing. 14 Conflicts were resolved by consensus between the two reviewers and, when necessary, with additional input from a third reviewer (APA).

Statistical analysis and data synthesis
In our quantitative analysis, we included only studies reporting full data  15 Summary sensitivity, specificity, +LRs, −LRs and diagnostic odds ratios (DORs) also were derived using a bivariate random-effects model.
To visualize heterogeneity of diagnostic accuracy among the included studies, we plotted sensitivity and specificity of our index tests on coupled forest plots using RevMan version 5.4 (Cochrane Training, London, UK). 16 When meta analysis was appropriate (given the number of studies and extent of clinical heterogeneity), we pooled results from the included studies. Because our random-effects meta analysis was performed for a single threshold, we chose a bivariate model for binary results to determine summary estimates of sensitivity and specificity with 95% confidence and prediction regions. 17 All meta-analyses were performed using MetaDTA version 2.01, an online application that uses statistical software R and the existing packages Shiny and lme4. [18][19][20] Results

Study characteristics
Our search on 15 November 2020 identified 1,389 citations in all three databases and, after removing duplicates, 994 titles were screened for eligibility for inclusion. The study flow diagram is shown in Figure 1.
Of these, four studies reported the role of UCPCR among 836 patients with T2DM; 18,21-23 seven studies studied that in 3,395 patients with T1DM; 9,10,18,24-27 and six studies studied the role of UCPCR among 257 patients with MODY (Table 1). 9,10,[21][22][23][24][25][26][27][28] The assessment of risk of bias and applicability via QUADAS-2 is presented in Figure 2 and Figure 3. 9,10,21,22,24-28 Overall, the risk of bias was scored as 'low' or 'concern' in 100% of studies for the index test, 'unclear' in 42% for the reference standard test, and 'high' or 'concern' in 55% for flow and timing and 80% in the patient selection domain. The applicability of studies was scored as 'low concern' in 100% for patient selection and index test, and 'unclear' in 35% for reference standard test.
Only four studies reported (or authors shared) the full data for appropriate quantitative synthesis; therefore, complete meta-analysis could be performed. 21   Diagnostic value of urine C-peptide creatinine ratio for diagnosing type 1 diabetes mellitus  Figure 4). 21 Full data on the sensitivity, specificity, positive predictive value (PPV) and negative predictive value (NPV) were not available from four studies, 9,25,27,28 and the study by Besser et al. in 2013 used a UCPCR cut-off of <0.7 nmol/mol; therefore, we could not include these studies in the meta-analysis. 10 The available data on these studies are depicted in    A sensitivity analysis for studies with patients with T1DM or T2DM is shown in Table 2. 21,22,24,26 The study by Hope et al. reported the use of UCPCR for identifying insulin deficiency among patients with T2DM, but no other studies reported comparable data in this review and therefore could not be meta-analysed. 24 Diagnostic value of urine C-peptide creatinine ratio for diagnosing maturity-onset diabetes of the young  Figure 6). 22