Factors potentiating risk of higher lactate level associated with metformin therapy: a cross-sectional study in an ambulatory care setting

This was a retrospective, cross-sectional study based on routinely collected medical records (secondary data analysis). Data for this study were collected as per the approval from the Institutional Review Board (IRB) at Mohammed Al-Mana College for Medical Sciences (SR/RP/79, Approval Date February 17, 2022) and the approval from the Qatif Central Hospital’s IRB (QCH-SRECO 19-2022, Approval Date June 8, 2022). Data were collected between the period of August 1, 2022 and February 20, 2023 from the hospital information technology department. All participants were deidentified and the waiver of consent collection was approved by IRB as data were collected retrospectively.

The inclusion criteria were patients aged 40 years or older who attended an outpatient endocrinology and internal medicine clinic and were diagnosed with T2DM between January 2020 and December 2021, with a completed medical record. The study cohort was limited to patients visiting outpatient endocrinology and internal medicine clinics to mitigate confounding associated with acute metabolic abnormalities. The set of 40 years as an inclusion criterion was selected, as older age is associated with a higher prevalence of multiple comorbidities, which could complicate T2DM management and influence lactate levels.^9,13

Patients aged 40 years old and younger, attending surgical or other medical care clinics, or an emergency department, and not diagnosed with T2DM were excluded. Consequently, patients with acute critical illness, severe infection, sepsis, or hypoxic respiratory conditions throughout the duration of the study were not to be included. For patients with multiple visits during the study period, only data from their first reported visit were included.

This retrospective study utilized routinely collected clinical data; thus, a formal a priori sample size calculation was not conducted. Instead, all eligible patients who met the inclusion criteria during the study period were incorporated to enhance statistical power and precision of estimates.

Demographic data, comorbidities, and recent laboratory findings were extracted from the medical electronic record, while the prescribed medications were verified from the pharmacy electronic records to ensure accurate representation of patient profiles. Comorbidities were identified and coded as per the International Classification of Diseases, 10th revision, 2016 (ICD-10).¹⁴ Comorbidities identified in the literature to potentiate the risk of MALA were collected and documented as heart failure, anemia, thyroid diseases, liver diseases, and renal diseases.¹¹ The Charlson Comorbidities Index (CCI) was calculated to predict 1-year mortality risk, and the creatinine clearance (CrCl) was estimated using the Cockcroft-Gault Equation.^15,16 Significant laboratory tests linked to evaluate and assess elevated lactate level were identified as creatinine kinase (to determine level of muscle damage), random blood glucose and glycosylated hemoglobin (to evaluate level of T2DM management), hemoglobin (to assess level of anemia), lactate dehydrogenase, and lactate level (Table S1). Lactate concentration was analyzed as a continuous variable across all statistical models. In this study, the term “elevated lactate” represents a relative or marginal increase in comparison to the reference group, rather than values exceeding pathological thresholds.

Blood lactate levels were determined based on standard laboratory tests documented in the hospital’s electronic medical record system. Lactate testing was conducted as a component of routine outpatient clinical assessment or metabolic evaluation at the treating physician’s discretion, rather than following a study-mandated procedure. All measurements were derived from venous blood samples analyzed in the hospital’s central laboratory utilizing standardized automated enzymatic techniques.

Lactate levels for metformin users aligned with durations of active metformin therapy as recorded in pharmacy prescription data. The retrospective approach limited standardization of the precise timing relative to the most recent metformin dose; however, all patients included had active metformin prescriptions during lactate testing.

Serum creatinine levels used for calculating CrCl were obtained from laboratory tests performed during the same outpatient visit as the lactate measurement or within a maximum of 7 days. For patients with multiple creatinine measurements throughout this period, the value nearest to the lactate test date was used. Metformin exposure and daily dosage were confirmed as active during laboratory testing via pharmacy prescription records.

The medications data, including long-term regular, short-term, as-needed prescriptions, and supplements, were collected and coded as per the Anatomical Therapeutic Chemical (ATC) classification system.¹⁷ From this list, the number of prescribed medications (NPM) was counted and documented. Furthermore, medications identified in the literature to potentiate the risk of MALA were documented as renin-angiotensin-aldosterone system inhibitors (RAASIs), diuretics, beta blockers (BBs), NSAIDs, antiepileptics, and systemic corticosteroid intake.

Records with missing values for the main outcome, blood lactate level, were excluded from the analysis. Analyses of other variables used only complete cases. The amount of missing data for the covariates was low, with less than 5% missing for each variable. Therefore, no imputation methods were applied.

Several potential sources of bias were considered. Selection bias was reduced by including all eligible patients who met the predefined criteria during the study period and by using electronic medical records instead of self-reported data. Information bias was reduced by extracting demographic, clinical, laboratory, and medication data in a standardized format from validated hospital electronic systems and pharmacy records. Misclassification bias was minimized by coding comorbidities with ICD-10 codes and medications using the ATC classification system.

In the current study, patients were classified based on their age into either middle-aged patients (<65 years) or older patients (⩾65 years) to assess the prevalence and pattern difference between these two groups in terms of metformin prescribing and prevalence of factors potentiating the risk of elevated lactate level. Furthermore, patients were classified based on the total daily metformin dose into low-dose users (⩽1000 mg/day), medium-dose users (1001–1500 mg/day), and high-dose users (⩾1501 mg/day) to assess the dose-dependent relationship with lactate level.

Demographic variables, comorbidities, and medication use were summarized using mean ± standard deviation (SD) for parametric continuous variables, median with interquartile range (IQR) for nonparametric continuous variables, and number with frequency (%) for categorical variables. For comparisons of continuous variables, the Student t test and the Mann–Whitney U test were used for parametric and nonparametric data, respectively. The Chi-square test was used to compare the frequency of categorical variables between groups. The pattern of metformin prescribing (difference in doses) and the difference between different metformin dosing groups were assessed using the analysis of variance test.

Multivariable linear regression was used to evaluate factors associated with higher lactate levels (continuous outcome) in metformin users. Variable selection was determined by (i) clinical plausibility and existing research on metformin-induced lactate elevation and (ii) statistical associations identified in univariable studies. The base adjustment set comprised age, sex, body weight, CrCl, CCI, and daily metformin dosage, as these variables were known or expected to influence lactate production/clearance and metformin exposure. All covariates were simultaneously incorporated into the final model (entry approach) to derive adjusted unstandardized regression coefficients (β) with 95% confidence intervals. Multicollinearity was evaluated by the variance inflation factor (VIF), where VIF values beyond 10 indicate concerning collinearity. To examine potential effect modification by age group, stratified multivariable analyses among older (⩾65 years) and middle-aged (<65 years) metformin users were conducted, using the same predetermined adjustment method.

Statistical analysis was performed using the SPSS statistical package version 26, whereas graphs were produced using GraphPad Prism software.

The study’s findings were reported following the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) standards, which are used for cross-sectional studies.¹⁸

While this study has several strengths; yet, it is important to acknowledge specific limitations. First, the cross-sectional design restricts the ability to establish causality between metformin use and lactate elevation. However, regression models were applied to identify associated factors to overcome this limitation. This study is based on data collected from a single center, which may limit the generalizability to broader populations with different demographic and clinical characteristics. However, the large sample size with the diverse characteristics supports generalizability to the Saudi Arabian and Middle Eastern populations. Because the study was conducted in an outpatient setting, which diminishes the probability of severe acute illness, important clinical data regarding transitory illnesses, inflammatory states, or subclinical hypoxic conditions were often inaccessible and therefore could not be incorporated into the regression models. Residual confounding from acute physiological stressors cannot be entirely ruled out and may have affected lactate levels in a limited subset of patients.²⁹ Lactate levels were assessed at a single time point, limiting the ability to evaluate longitudinal changes or fluctuations in lactate metabolism. Additionally, lactate measurements were obtained during routine clinical care rather than at standardized time points relative to medication dosing, which may introduce variability in observed concentrations. Although CrCl and lactate levels were being measured during the same clinical period, the retrospective design of the study limited accurate synchronization with the precise timing of metformin administration. Consequently, transient variations in renal function or lactate associated with recent administration could not be entirely documented.

Lastly, as a cross-sectional study, lactate measurements were obtained during routine outpatient visits and were not associated with documented acute clinical symptoms, hospitalizations, or hemodynamic instability at the time of testing. Therefore, no direct short-term clinical effects attributable to lactate elevation were detected in our cohort.