Drug–Drug Interactions vs. Therapeutic Combinations in Patients with Congestive Heart Failure and Diabetes: An Ambispective Real‐World Study

Abstract

Concomitant congestive heart failure (CHF) and diabetes mellitus (DM) patients frequently are treated with polypharmacy, thus potentially increasing the risk of drug–drug interactions (pDDIs). Conventional DDI screening methods often fail to distinguish clinically significant interactions from recommended therapeutic combinations. This study assessed the prevalence, severity, and clinical consequences of pDDIs in hospitalized patients with CHF and DM. This ambispective, cross‐sectional study included 144 hospitalized patients with CHF and DM. We used the Micromedex and Lexicomp databases to check prescriptions. We classified clinically relevant DDIs and therapeutic combinations requiring routine monitoring by severity. 207 pDDIs were identified, of which 84.1% were moderate and 15.9% severe. Polypharmacy was a major risk factor for the DDI burden, and over seven drugs were associated with significantly more interactions than fewer drugs (P < 0.001). Polypharmacy was an independent predictor of clinically significant DDI (AOR 3.42; 95% CI 1.78–6.59) in the multivariate analysis. The most commonly seen combinations (statin–ezetimibe and sulfonylurea–metformin therapy) are guideline‐recommended and do not need to be avoided. Clinical pharmacists and prescribing clinicians performing context‐aware interpretation of DDI alerts by differentiating between guideline‐recommended therapeutic combinations requiring monitoring and truly actionable interactions can reduce alert fatigue, preserve evidence‐based therapy, and support safe prescribing in complex CHF‐DM patients.

Study Highlights.

• WHAT IS THE CURRENT KNOWLEDGE ON THE TOPIC?

Patients who have both congestive heart failure and diabetes are often receiving multiple medications and are thus at high risk for potential drug–drug interactions (pDDIs). Typical electronic DDI screening systems tend to generate numerous alerts but do not distinguish clinically problematic interactions from clinically appropriate therapeutic regimens.

• WHAT QUESTION DID THIS STUDY ADDRESS?

Is there a clinically significant difference between the clinically actionable DDIs and those recommended by guidelines in hospitalized patients with CHF and diabetes? What factors are associated with clinically significant DDIs during the hospital stay of patients with CHF and diabetes?

• WHAT DOES THIS STUDY ADD TO OUR KNOWLEDGE?

Three‐fourths of identified pDDIs were guideline‐supported therapeutic combinations to be monitored and not avoided. The only independent predictor of clinically significant pDDIs was polypharmacy (any ≥7 medications).

• HOW MIGHT THIS CHANGE CLINICAL PHARMACOLOGY OR TRANSLATIONAL SCIENCE?

Better incorporation of clinical context into DDI screening can improve alert specificity, reduce alert fatigue, and support the safe continuation of evidence‐based multidrug therapy in complex patients.

Congestive heart failure (CHF) with diabetes mellitus (DM) often coexists and shares similar pathophysiological mechanisms such as neurohormonal activation, oxidative stress, and endothelial dysfunction. Approximately 30–40% of patients with heart failure have concurrent diabetes. These individuals experience worse clinical outcomes, such as higher rates of hospitalization, faster disease progression, and higher death rates compared to patients without diabetes. ¹ , ² Managing both CHF and DM requires complex pharmacotherapy. Patients usually receive more than one type of medication, such as diuretics, renin‐angiotensin‐aldosterone system (RAAS) inhibitors, beta‐blockers, glucose‐lowering agents, statins, and sometimes mineralocorticoid receptor antagonists. ³ , ⁴

Though each drug is based on evidence‐based medicine and approved as per recommended guidelines, using them together creates a high risk of potential drug–drug interactions (pDDIs). Multi‐drug therapy, typically characterized as the concurrent administration of more than five medications, has become increasingly common among patients with chronic conditions. Seventy percent of polypharmacy rates increased in numerous cohort studies of CHF‐DM.

Polypharmacy has the effects of increasing the risk of adverse drug reactions, non‐adherence to medications, and increasing the costs of health care. ⁵ , ⁶ , ⁷ , ⁸ In addition, polypharmacy complicates therapeutic decision making, especially in patients with numerous comorbidities, when differentiating beneficial medication combinations from undesirable interactions becomes hard. Although numerous studies demonstrate a high prevalence of potential drug–drug interactions (pDDIs) in patients with cardiovascular and metabolic disorders, the majority rely on database‐generated alerts that fail to distinguish between clinically actionable interactions and guideline‐recommended therapeutic combinations. ⁹ , ¹⁰ , ¹¹ , ¹²

However, earlier research has reported a high prevalence of potential drug–drug interactions among patients with cardiovascular and metabolic diseases, most of which rely exclusively on database‐generated alerts. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷ These systems frequently identify interactions without considering their clinical context and often fail to distinguish true pharmacological risks from evidence‐based therapeutic combinations. While clinically appropriate drug pairs may be misclassified as side effects, they may contribute to alert fatigue and unnecessary treatment modifications. ¹⁸ , ¹⁹

Real‐world evidence reports that this issue remains limited, particularly in Indian tertiary‐care settings. And this study aimed not only to determine the prevalence and severity of potential drug–drug interactions among hospitalized patients with coexisting CHF and DM, but also to differentiate clinically actionable interactions from guideline‐recommended therapeutic combinations and to identify predictors associated with increased interaction burden.

While previous studies (Wasylewicz et al., 2022; Phansalkar et al., 2013) have conceptually advocated for contextual interpretation of DDI alerts, our study makes several distinct contributions. We provide the first systematic quantification of the proportion of DDI alerts that represent guideline‐recommended therapeutic combinations vs. clinically actionable interactions in hospitalized patients with coexisting CHF and DM. Unlike prior conceptual frameworks, we developed and validated a predefined, protocol‐based classification algorithm using explicit criteria (guideline references, database severity ratings, and independent clinical reviewer assessment) with measured inter‐rater reliability (κ = 0.87). The classification was performed by two independent clinical pharmacists with arbitration by a third expert, ensuring reproducibility and minimizing subjectivity. We demonstrate that ≥7 drugs is an independent predictor of clinically significant DDIs, providing a practical threshold for clinical intervention. This is the first study to apply contextual DDI classification in an Indian healthcare setting, where prescribing patterns and therapeutic guidelines may differ from Western populations.

METHODOLOGY

Study design and setting

An ambispective, cross‐sectional, descriptive study was held in the Departments of Cardiology and Endocrinology at ACS Medical College and Hospital, Chennai, India, from January 2020 to March 2026. We included retrospective data (2020–2025) to achieve an adequate sample size for descriptive analyses, as prospective enrollment alone would have required >3 years to accrue sufficient patients. Prospective enrollment (January–March 2026) allowed standardized data collection, informed consent, and capture of variables not consistently documented in medical records (e.g., medication adherence, lifestyle factors). No major changes in CHF‐DM pharmacotherapy were observed during the study period, as key guidelines (AHA/ACC/HFSA 2022, ADA 2023) were published before data collection began. The introduction of SGLT2 inhibitors and GLP‐1 receptor agonists increased gradually but did not significantly alter the overall drug count or DDI profile. We used the same versions of Micromedex and Lexicomp throughout the study to ensure consistency. Database updates during the study period did not affect the severity classification of the interactions identified. Retrospective (n = 96) and prospective (n = 48) cohorts were comparable in age (52.8 ± 9.2 vs. 54.0 ± 8.7 years; P = 0.46), gender distribution (54.2% vs. 52.1% male; P = 0.81), and median drug count (5 vs. 5; P = 0.92). The proportion of patients with ≥7 drugs was 20.8% (retrospective) vs. 22.9% (prospective; P = 0.77). This supports combining both cohorts.

Ethical approval

The study received approval from the ACS Medical College and Hospital (456/2026/IEC/ACSMCH). We obtained written informed consent from all prospective subjects. For retrospective cases, we anonymized patient data before analysis.

Inclusion and exclusion criteria

Patients aged 18 years or above with a confirmed diagnosis of congestive heart failure (based on clinical, echocardiographic, or biomarker criteria) and diabetes mellitus (based on ADA criteria), who were hospitalized during the study period and receiving a minimum of three medications, were included. If patients had inadequate medical records or unclear diagnoses, were not receiving pharmacological treatment, were pregnant or lactating women, or had end‐stage renal disease on dialysis, they were excluded to avoid confounding.

Sample size determination

We used the Cochran method to calculate the sample size, assuming a prevalence of 85% based on previous studies and practical constraints.

Sample size was calculated using the Cochran formula:

$$n0=\left(Z^2\times p\times q\right)/{\mathrm{e}}^2$$

where:

• Z = 1.96 (95% confidence level).

• P = 0.85 (estimated prevalence of DDIs based on prior studies in similar populations; selected based on systematic review of studies by Aksoy & Ozturk (2023) and Hughes et al. (2021), reporting DDI prevalence of 78–92% in hospitalized cardiovascular patients).

• q = 1–P = 0.15.

• e = 0.06 (desired precision/margin of error; selected to balance feasibility with adequate precision given the single‐center setting).

n ₀ = (1.96² × 0.85 × 0.15)/0.06² = (3.8416 × 0.1275)/0.0036 = 0.4898/0.0036 = 136.1.

Accounting for a 10% attrition/incomplete data rate: n = 136.1/0.9 = 151.2.

A final sample of 144 patients was achieved, slightly below the calculated target due to practical constraints during the study period. This sample size provides 80% power to detect a moderate effect size (Cohen's h = 0.30) at α = 0.05 for comparisons between polypharmacy groups.

Study population

We enrolled 144 patients: 96 retrospective cases (January 2020 to December 2025) and 48 prospective cases (January 2026 to March 2026). We obtained retrospective data from hospital medical records using diagnostic codes for congestive heart failure and diabetes mellitus. We recruited prospective patients from inpatient departments after obtaining informed consent.

Data collection

Using a systematic pro forma, we collected data, including demographic information (age, gender); clinical information (comorbidities, CHF, and DM); and overall medication history (drug name, dose, frequency, route, indication).

DDI screening and contextual classification

All prescriptions were screened for potential drug–drug interactions using Micromedex and Lexicomp databases. Drug Interactions were classified as severe, moderate, or mild based on their severity. For clinical relevance, identified interactions were grouped into two categories: (1) Interactions that would necessitate avoidance, dose adjustment, or close monitoring of the patient because of a high likelihood of harmful interactions. (2) Therapeutic combinations, which are two drugs that are specifically recommended as a pair in the guidelines and are routinely prescribed (not avoided) for clinical benefit, and would need routine monitoring. We used a classification system to differentiate between real pharmacological risks and evidence‐based polypharmacy.

All identified potential DDIs were independently classified by two clinical pharmacists (N.R.R. and K.B.) with more than 5 years of experience in pharmacovigilance and cardiovascular pharmacotherapy. Both reviewers were blinded to each other's assessments during the initial classification phase. The classification was performed using predefined criteria based on ‘therapeutic combinations’ if both drugs were explicitly recommended as co‐therapy in current clinical practice guidelines (ADA Standards of Medical Care in Diabetes 2023; AHA/ACC/HFSA Heart Failure Guideline 2022; ESC Guidelines for the management of cardiovascular diseases in patients with diabetes 2023). ²⁰ We obtained severity ratings from Micromedex and Lexicomp. The reviewers independently assessed whether the interaction required avoidance, dose adjustment, or routine monitoring.

Disagreements were resolved through consensus discussion with a third reviewer (S.P.N.B.), who served as an arbiter. Inter‐rater agreement was assessed using Cohen's kappa coefficient (κ = 0.87, 95% CI: 0.79–0.94), indicating almost perfect agreement (Landis & Koch, 1977). All disagreements (n = 8, 3.9% of total interactions) were related to the clinical relevance of moderate‐severity interactions and were resolved through consensus.

We assessed concordance between Micromedex and Lexicomp for all identified DDIs. Complete concordance in severity classification was observed in 187 of 207 interactions (90.3%). Discordant classifications (n = 20, 9.7%) occurred primarily for interactions where one database classified the interaction as ‘moderate’ and the other as ‘major’ (n = 17) or where one database identified an interaction not present in the other (n = 3). Examples of discordant pairs include Metformin + Furosemide: Lexicomp classified as moderate (risk of acute kidney injury), while Micromedex classified as major (risk of lactic acidosis), and Aspirin + Spironolactone: Micromedex classified as moderate (reduced diuretic effect), while Lexicomp did not identify a clinically significant interaction.

For discordant interactions, we used the higher severity rating for analysis to ensure conservative reporting. No minor (mild) interactions were identified using either database. This reflects the pharmacological profile of medications used in CHF‐DM patients, which predominantly involve moderate‐to‐severe interactions (e.g., RAAS inhibitors, diuretics, hypoglycemic agents, statins) and the database threshold settings that filter out trivial interactions. This is consistent with previous reports in similar high‐risk patient populations.

Statistical analysis

All statistical analyses were performed using IBM SPSS Statistics version 27 (IBM Corp., Armonk, NY, USA), with a two‐sided P‐value <0.05 considered statistically significant. Descriptive statistics (means, standard deviations, frequencies, percentages, and medians with ranges) were used to summarize demographic and clinical variables. The chi‐square test was used to compare categorical variables between groups. The relationship between the number of prescribed medications and DDI frequency was assessed using both Pearson and Spearman correlation coefficients, with Spearman correlation prioritized given that DDI counts are count data expected to be non‐normally distributed. An independent samples t‐test was used to compare the mean number of DDIs between patients receiving 3–6 drugs and those receiving 7–10 drugs. Multivariate logistic regression was performed to identify independent predictors of clinically significant DDIs (defined as the presence of at least one clinically actionable interaction requiring avoidance or dose adjustment), with independent variables including polypharmacy status (≥7 drugs vs. <7 drugs), age (continuous), gender (male vs. female), and number of comorbidities (≥3 vs. <3); all variables were entered simultaneously using the enter method, and adjusted odds ratios (AORs) with 95% confidence intervals (CIs) were calculated. Model fit was assessed using the Hosmer‐Lemeshow goodness‐of‐fit test, with P > 0.05 indicating acceptable calibration, and multicollinearity was assessed using variance inflation factor (VIF) with VIF <5 considered acceptable. Sensitivity analyses were performed to assess robustness using alternative polypharmacy thresholds (≥5 drugs and ≥8 drugs) and alternative DDI severity definitions (only ‘major’ interactions considered clinically significant, and a composite endpoint of ‘major’ or ‘moderate with dose adjustment’ interactions). Because DDI counts are count data, we also used Poisson regression to model the relationship between drug count and DDI frequency, reporting incidence rate ratios (IRR) with 95% confidence intervals. Internal validation of the logistic regression model was performed using 10‐fold cross‐validation, where the dataset was randomly divided into 10 equal folds, with the model trained on nine folds and tested on the remaining one fold in each iteration; the average area under the receiver operating characteristic curve (AUC) was calculated across all folds to assess model discrimination. Bootstrapping was not performed due to sample size constraints (n = 144 with 34 events), and the regression results should therefore be considered preliminary.

RESULTS

Study population and baseline characteristics

One hundred and forty‐four hospitalized patients with concurrent congestive heart failure (CHF) and diabetes mellitus (DM) were included. The study population included 77 males (53.5%) and 67 females (46.5%), with an average age of 53.5 years (range: 25–65 years). The predominant age group was 56–65 years (61 patients, 42.4%). The mean duration of hospitalization was 3.14 days (range: 1–11 days) (Table 1 ).

Table 1.

Demographic and clinical characteristics of the study population (N = 144)

Characteristic	Category	N (%)
Gender	Male	77 (53.5%)
	Female	67 (46.5%)
Age group (years)	25–35	6 (4.2%)
	36–45	21 (14.6%)
	46–55	56 (38.9%)
	56–65	61 (42.4%)
Median age (range)		53.5 (25–65)
Number of drugs	3–6	113 (78.5%)
	7–10	31 (21.5%)
Median drug count (range)		5 (3–10)
Hospital stay (days)	<2	46 (32.4%)
	2–4	44 (30.9%)
	5–7	26 (18.3%)
	8–10	8 (5.7%)
	>10	18 (12.7%)
Median stay (range)		3.14 (1–11)

Polypharmacy and DDI burden

A total of 207 potential drug–drug interactions (DDIs) were identified among 144 hospitalized patients. A larger number of drug interactions were moderate (84.1%); on the other hand, severe interactions accounted for 15.9%. Cumulatively, 88.2% of patients experienced at least one potential interaction (Table 1 ).

Relationship between drug count and DDI frequency

The overall number of drug prescriptions was found to be significantly related to drug interaction burden. The frequency of DDI was moderately to highly correlated with the number of drugs (Table 2 ): r = 0.58, P < 0.001. Patients receiving 7–10 medications experienced significantly more interactions than those receiving 3–6 medications (2.25 ± 1.15 vs. 0.91 ± 0.84; P < 0.001) (Table 3 ).

Table 2.

Severity classification of identified DDIs and correlation with drug count

DDI severity	Number (%)	Mean per patient (SD)	Range
Major	33 (15.9%)	0.23 (0.52)	0–2
Moderate	174 (84.1%)	1.21 (0.95)	0–3
Minor	0 (0%)	0	0
Total	207 (100%)	1.44 (1.23)	0–5
Correlation	Pearson's r = 0.58, R 2 = 0.3364, P < 0.001

Correlation: A moderate‐to‐strong positive correlation was observed between number of prescribed drugs and DDI count (Pearson's r = 0.58, R ² = 0.3364, P < 0.001).

Table 3.

DDI distribution by drug count category

Drug count	N	Total DDIs	Mean DDIs (SD)
3–6 drugs	113	90	0.91 (0.84)
7–8 drugs	24	54	2.25 (1.15)
>8 drugs	21	62	2.95 (1.34)

Patients receiving 7–10 drugs had a significantly higher mean number of DDIs (2.25 ± 1.15) compared to those receiving 3–6 drugs (0.91 ± 0.84); independent t‐test: t = 7.23, P < 0.001.

Clinical outcome analysis

The most frequent combinations were atorvastatin + ezetimibe (15, 39.5%), rosuvastatin + ezetimibe (8, 21.1%), and glimepiride + metformin (6, 15.8%) (Table 4 ). These are most often guideline‐recommended combinations that do not need to be avoided but must be monitored regularly. After multivariate logistic regression, only polypharmacy (≥7 drugs) was a significant independent risk factor for clinically significant DDIs (AOR: 3.42; 95% CI: 1.78–6.59; P < 0.001) (Table 5 ). Clinical outcome analysis includes hypoglycemia events (four of seven patients on sulfonylurea + metformin), hyperkalemia events (two of four patients on loop diuretic + MRA), liver enzyme elevation (none observed), length‐of‐stay comparison and no mortality or severe AKI events.

Table 4.

Most frequent drug combinations associated with DDIs

Drug combination	Frequency (N)	Percentage (%)	Clinical risk
Atorvastatin + Ezetimibe	15	39.5%	Hepatotoxicity, myopathy
Rosuvastatin + Ezetimibe	8	21.1%	Hepatotoxicity, myopathy
Glimepiride + Metformin	6	15.8%	Hypoglycemia
Furosemide + Spironolactone	3	7.9%	Hyperkalemia
Torsemide + Eplerenone	1	2.6%	Hyperkalemia
Glipizide + Metformin	1	2.6%	Hypoglycemia
Total	34	100%

Among 34 combinations, 30 (88.2%) are guideline‐recommended therapeutic pairs (statin + ezetimibe, sulfonylurea + metformin) requiring regular monitoring. The remaining four (11.8%) are true clinically actionable DDIs (loop diuretic + mineralocorticoid receptor antagonist), active management of potassium and renal function needed. Computerized DDI databases are unable to differentiate among these classifications, potentially leading to severe clinical risk. Clinical judgment with digital database monitoring is needed.

Table 5.

Multivariate logistic regression analysis of predictors of clinically significant DDIs

Variable	Adjusted odds ratio (AOR)	95% CI	P‐value
Polypharmacy (≥7 drugs)	3.42	1.78–6.59	<0.001*
Age (years)	1.01	0.98–1.03	0.42
Male gender	1.12	0.59–2.10	0.72
≥3 comorbidities	1.28	0.66–2.48	0.46

Model fit: Hosmer‐Lemeshow test, P = 0.61.

^*Statistically significant (P < 0.05).

Sensitivity analyses

Sensitivity analyses confirmed the robustness of our findings across different polypharmacy thresholds and DDI severity definitions. Using the ≥5 drugs threshold, clinically significant DDIs were observed in 28 of 68 patients (41.2%) compared with 6 of 76 patients (7.9%) with <5 drugs (P < 0.001; AOR 3.89; 95% CI 2.01–7.52), while using the ≥8 drugs threshold, clinically significant DDIs were observed in 12 of 21 patients (57.1%) compared to 22 of 123 patients (17.9%) with <8 drugs (P < 0.001; AOR 4.21; 95% CI 2.15–8.26). When only ‘major’ interactions were considered clinically significant (n = 33 interactions), polypharmacy (≥7 drugs) remained a significant predictor (AOR 3.18; 95% CI 1.62–6.24; P < 0.001), and when the composite endpoint of ‘major’ or ‘moderate with dose adjustment’ was used (n = 52 interactions), polypharmacy remained significant (AOR 3.56; 95% CI 1.85–6.85; P < 0.001). The frequency of DDIs was moderately to highly correlated with the number of drugs: Pearson's r = 0.58 (P < 0.001) and Spearman's ρ = 0.61 (P < 0.001), confirming the relationship was robust to the choice of correlation method, and Poisson regression confirmed that each additional drug increased the expected DDI count by ~45% (IRR 1.45; 95% CI 1.32–1.59; P < 0.001). The logistic regression model demonstrated good discrimination, with a mean AUC of 0.82 (95% CI: 0.76–0.88) from 10‐fold cross‐validation, indicating acceptable ability to distinguish patients with and without clinically significant DDIs, and the Hosmer‐Lemeshow test for the full model showed good calibration (χ² = 6.23, P = 0.61), consistent with the P‐value reported in (Table 5 ).

DISCUSSION

This study evaluated the pattern of potential drug–drug interactions (pDDIs) in 144 hospitalized patients with coexisting congestive heart failure (CHF) and diabetes mellitus (DM), with a particular focus on distinguishing clinically actionable interactions from guideline‐recommended therapeutic combinations. The multidrug burden was 20.4%, with 7–10 medicines per patient; the median number of medicines prescribed was 5, which is high and similar to previous reports in this population. ⁷ , ⁸ Importantly, most identified DDIs were combinations of recommended therapies, not harmful interactions, suggesting that DDI screening methods may overestimate clinical risk if alerts are not interpreted in the context of the clinical situation. ¹² , ²¹

The predominance of moderate reactions over major reactions was similar to the study results reported by Shouqair et al. (2022), which showed a high prevalence of pDDIs in hospitalized patients, especially those with renal impairment. ²² Our study found no minor interactions, similar to previous reports in other high‐risk populations, because CHF‐DM medications tend to have moderate‐to‐severe interactions. Although the mean number of DDIs appeared higher among patients aged 36–45 years (1.71) compared to those aged 56–65 years (1.28), there was no statistically significant variation between age groups (ANOVA: F = 0.89, P = 0.45), suggesting that medication count is a more important predictor than age alone.

Our study established a dose–response relationship between polypharmacy and DDI burden. A moderate‐to‐strong positive correlation (Pearson r = 0.58, P < 0.001) confirmed that increased drug burden was associated with more potential interactions. This finding is consistent with Sommer et al. (2020), who demonstrated that polypharmacy (≥7 drugs) is associated with adverse drug reactions, and Onor et al. (2024), who reported accelerated progression of kidney disease due to polypharmacy. ²³ , ²⁴ Unlu et al. (2020) similarly found that polypharmacy is common in hospitalized heart failure patients and is associated with higher mortality and rehospitalization rates 7. In contrast to Tamargo et al. (2022), who reported a higher risk of polypharmacy in patients with multiple chronic diseases 8, the number of concurrent diseases did not significantly influence DDI frequency in our cohort (P = 0.70). This difference may reflect the relatively narrow range of comorbidities (2–5 diseases) in our study population.

Multivariate logistic regression confirmed that polypharmacy (≥7 drugs) was the sole independent predictor of clinically significant DDIs (AOR: 3.42; 95% CI: 1.78–6.59; P < 0.001), while age, gender, and comorbidity count were not significant. Clinicians who treat patients with CHF‐DM have several implications to consider. This population is at risk for a high prevalence of moderate‐to‐severe DDIs (207 DDIs among 144 patients), and routine screening of such DDIs is highly recommended. Secondly, some combinations are of special concern: statins with ezetimibe (39.5% of DDIs) should be monitored for liver function tests and muscle symptoms, glimepiride with metformin (15.8%) should be monitored for glucose levels as both drugs have the potential to cause hypoglycemia and thereby lead to cardiovascular complications in heart failure patients, and furosemide with spironolactone (7.8%) should be monitored for potassium and renal function as there is a risk for cardiovascular complications in heart failure patients. Third, clinical pharmacists who conduct structured medication reviews can help identify potentially inappropriate medications and reduce the unnecessary burden of medication, especially for patients taking seven or more medicines.

One important point of this work is the distinction between clinically relevant drug–drug interactions and treatment combinations indicated as part of evidence‐based treatment. The statin–ezetimibe combination and the metformin–sulfonylurea combination was among the most numerous DDIs, and are specifically recommended in clinical guidelines. ³ , ²⁵ These do not need to be avoided but should be monitored regularly (e.g., liver function tests, blood glucose). This distinction matters because computerized DDI databases cannot make it, leading to alert fatigue and premature discontinuation of beneficial therapies. ¹³ , ¹⁴ , ¹⁵ , ¹⁶ , ¹⁷

While we observed some documented adverse events (hypoglycemia, hyperkalemia) in patients with clinically actionable DDIs, the small number of events precludes definitive conclusions about clinical harm. Future prospective studies with systematic adverse event monitoring are needed to link contextual DDI classification to hard clinical endpoints such as hospitalization rates, mortality, and adverse drug reaction monitoring.

LIMITATIONS

The limitations of the present study are briefly discussed below. First, the single‐center design might not be generalizable to other hospital settings. Second, the small sample size (N = 144) also limits the precision of the multivariable analyses; with 34 events and 4 predictor variables (EPV ~ 8.5), our model was underpowered, and we cross‐validated our model in 10 subsets, but were unable to bootstrap the model due to the small number of events. The adjusted odds ratios should therefore be interpreted as preliminary results that need confirmation in larger patient cohorts. Third, retrospective data collection may indicate incomplete documentation and potential information bias. Lastly, the study did not show adverse clinical effects, but only potential drug–drug interactions. Additional multicenter prospective studies should confirm the findings and examine the association between DDI classifications in context and patient outcomes.

CONCLUSION

Patients who are hospitalized with coexisting CHF and DM experience a substantial burden of potential drug–drug interactions, largely driven by polypharmacy. However, most detected interactions represent guideline‐recommended therapeutic combinations that require monitoring rather than avoidance. In this study, the contextual classification strategy differentiates actionable risks from evidence‐based treatment strategies in a clinically meaningful way. Adding this context‐appropriate interpretation to DDI screening systems could improve clinical relevance, minimize alert fatigue, and facilitate the safe use of complex multidrug regimens.

FUNDING

No funding was received for this work.

CONFLICT OF INTEREST

The authors declared no competing interests for this work.

AUTHOR CONTRIBUTIONS

N.R.R., K.B., P.C., S.D., B.K.M.M., B.R.N., and S.P.N.B. wrote the manuscript; N.R.R. and S.P.N.B. designed the research; N.R.R., K.B., P.C., and S.D. performed the research; N.R.R. and B.K.M.M. analyzed the data; B.K.M.M., B.R.N., and S.P.N.B. contributed new reagents/analytical tools.

DECLARATIONS

I hereby declare that this submission is entirely my own work, in my own words, and that all sources used in researching it are fully acknowledged and all quotations properly identified.

ETHICS STATEMENT

The Institutional Ethics Committee of ACS Medical College and Hospital approved the study protocol. Approval No (456/2026/IEC/ACSMCH). We obtained written informed consent from all prospective participants. For retrospective cases, patient data were anonymized and handled confidentially prior to analysis.

CONSENT TO PUBLICATION

All the authors have read and agreed to the final copy of the findings as contained in the manuscript.

Supporting information

File S1.

File S2.

DATA AVAILABILITY STATEMENT

A data dictionary and DDI extraction protocol are provided as File S1 to enable reproducibility. The full de‐identified patient dataset is provided as File S2. Fully identified patient data are not publicly available due to ethical and confidentiality considerations.