Pakistan Journal of Medical & Health Sciences

Predictive Role of C-Reactive Protein Ferritin D-Dimer and Interleukin-6 in Determining COVID-19 Severity and Outcomes

TANVEER AHMAD — 2026-02-28

Background: The unpredictable clinical course of COVID-19 has led to an urgent need for early prognostic biomarkers. Among the most studied markers are C‑reactive protein (CRP), ferritin, D‑dimer, and interleukin‑6 (IL‑6), which may reflect underlying inflammation, coagulopathy, and cytokine response.

Aim: To evaluate the predictive role of CRP, ferritin, D‑dimer, and IL‑6 levels in determining the severity and clinical outcomes of COVID‑19 among hospitalized patients at Lady Reading Hospital, Peshawar.

Methodology: A descriptive cross‑sectional study was conducted at Lady Reading Hospital, Peshawar, from August to December 2024. A total of 200 RT‑PCR‑confirmed COVID‑19 patients were enrolled. Serum CRP, ferritin, D‑dimer, and IL‑6 levels were measured on admission. Patients were classified into mild, moderate, and severe categories based on WHO guidelines. Outcomes such as ICU admission, need for mechanical ventilation, and mortality were recorded.

Results: Elevated levels of CRP, D‑dimer, and IL‑6 were significantly associated with severe disease (p<0.001). Ferritin levels were also higher in severe cases but showed moderate predictive value (p=0.042). IL‑6 and D‑dimer had the strongest correlation with ICU admission and mortality.

Conclusion: CRP, D‑dimer, and IL‑6 are valuable biomarkers for early prediction of disease severity and poor outcomes in COVID‑19. Routine measurement of these markers at Lady Reading Hospital may guide clinical decision‑making and resource allocation.

Diagnostic Role of Contrast-Enhanced Ultrasound (CEUS) in Characterizing Indeterminate Liver Lesions

DR JODAD — 2026-02-28

Background: Indeterminate lesions of the liver are commonly detected in abdominal imaging and in most instances they are cause of diagnostic ambiguity when traditional ultrasound, CT, or MRI fails to clearly define them as benign or malignant. Contrast-Enhanced Ultrasound (CEUS) has become a promising, real-time imaging modality, which does not expose ionizing radiation or any exposure of nephrotoxic contrast to visualize the microvascular perfusion. It is especially useful in the assessment of hepatic lesion evaluation because of its capacity to show characteristic enhancement and a wash-out effect.

Objectives: The objective of the study was to assess the diagnostic utility of CEUS as a tool to characterize indeterminate liver lesions as well as compare patterns of its improvement with ultimate diagnosis made using CT/MRI, histopathology, or imaging follow-up.

Methods: The study was a cross-sectional diagnostic accuracy study, which was carried out among tertiary care hospitals in Punjab, Pakistan between June 2024 and May 2025. There were 80 adult patients with unspecified hepatic lesions identified after the first grayscale imaging who were enrolled. Secondary generation microbubble contrast agent was used to perform the CEUS and a lesion was evaluated at the arterial, portal venous and late stages. Behavioral enhancements, wash-in behavior, wash-out behavior and perfusion were studied. Contrast-enhanced CT/MRI, biopsy or 6-month follow-up imaging were used to confirm the final diagnoses.

Results: CEUS correctly detected 47 out of 52 benign lesions and 26 out of 28 malignant lesions and had a diagnostic accuracy of 91.2 with high sensitivity (92.8) and specificity (90.4). Typical patterns in persistent or isoenhancing late phase were observed in benign lesions (hemangiomas and focal nodular hyperplasia). Malignant lesions such as hepatocellular carcinoma and metastases showed hyperenhancement of the arteries and then early or late wash-out.

Conclusion: CEUS is an effective, safe, cost effective modality, which can distinguish between benign and malignant liver lesions. It can play a crucial role in eliminating unnecessary biopsies and supplement CT/MRI in routine liver examination, since its application in diagnostic algorithms can greatly decrease the number of unnecessary biopsies.

AI-Driven Biomarker Discovery: Promise vs. Reality in Translational Medicine

NAVEED SHUJA — 2026-02-28

Artificial intelligence (AI) has rapidly become one of the most influential forces shaping modern translational medicine¹. With the explosion of multi-omics data, electronic health records, imaging archives, and high-throughput biological assays, AI promises an unprecedented ability to decode complex biological systems and uncover clinically actionable biomarkers². In principle, these tools can accelerate early diagnosis, improve risk stratification, personalize therapy, and reduce the time and cost of drug discovery³. Yet, despite remarkable progress, the actual translation of AI-derived biomarkers into routine clinical practice remains limited⁴. This gap between promise and reality reflects multiple scientific, technical, ethical, and regulatory challenges that must be addressed for AI to fulfil its potential⁵.

The promise of AI-driven biomarker discovery is rooted in its power to process vast amounts of heterogeneous data genomic, epigenomic, transcriptomic, proteomic, metabolomic, clinical, and imaging far beyond the capacity of traditional biostatistical approaches⁶. Machine learning and deep learning algorithms can detect subtle, nonlinear patterns that signal disease onset long before clinical symptoms arise⁷. Moreover, network-based models can unravel interactions between genes, proteins, pathways, and metabolites that drive pathophysiology, offering mechanistic insights that inform new therapeutic targets⁸. This capability is particularly transformative in complex diseases such as diabetes, cancer, metabolic syndrome, neurodegeneration, and cardiovascular conditions, where multilevel dysregulation makes single-marker approaches insufficient⁹.

At a population level, AI can democratize biomarker discovery¹⁰. Open-access repositories such as NCBI Gene Expression Omnibus (GEO), ArrayExpress, and precision-medicine datasets generated through federated learning platforms allow researchers in developing countries including Pakistan to engage in cutting-edge biomarker science without requiring expensive laboratory infrastructure¹¹. AI platforms can even harmonize noisy, multi-institutional datasets and overcome batch effects, making them invaluable for studying diseases with high heterogeneity across populations¹².

However, the reality is more complicated. A striking proportion of AI-identified biomarkers fail to advance beyond computational discovery due to poor reproducibility, lack of external validation, or inability to demonstrate added predictive value over established clinical markers¹³. Many models are trained on small, unrepresentative datasets, leading to overfitting and limited generalizability especially when applied to populations with different ethnic, environmental, or socioeconomic backgrounds¹⁴. This is particularly relevant for South Asian populations, where genetic diversity, dietary patterns, and disease phenotypes differ significantly from Western cohorts that dominate public datasets¹⁵.

Another major constraint is the interpretability problem. Deep learning models often behave as “black boxes,” providing predictions without mechanistic explanations¹². Clinicians and regulatory bodies remain hesitant to adopt such opaque systems in decision-making. Furthermore, integrating AI workflows into real-world clinical settings requires standardized data pipelines, regulatory frameworks, skilled personnel, and robust QA/QC procedures elements that many healthcare systems lack¹⁰.

Ethical and privacy concerns also limit progress. Multi-omics datasets are sensitive and require stringent protections, while biases embedded in AI algorithms risk perpetuating inequities¹³. Without transparent model design, diverse training cohorts, and rigorous fairness evaluation, AI-derived biomarkers may inadvertently disadvantage certain patient groups¹⁴.

Despite these realities, the field is progressing rapidly. Hybrid approaches that combine wet-lab validation with network-based AI prediction, improved model interpretability, and the rise of explainable AI (XAI) are narrowing the translational gap¹¹. Increasingly, biomarker discovery pipelines integrate differential gene expression, co-expression networks, protein–protein interaction mapping, machine learning signatures, molecular docking, ADMET profiling, and in-vitro validation producing more robust, biologically grounded candidates¹².

What is needed now is a shift from hype to responsible implementation. This means prioritizing large, diverse datasets; standardizing workflows; enforcing multi-step validation; and fostering collaboration between computational scientists, clinicians, biochemists, and molecular biologists¹⁴. Pakistan’s rapidly growing biomedical research community, anchored by tertiary hospitals and molecular laboratories, is well-positioned to contribute meaningfully to this transition especially in fields such as metabolic syndrome, cardiovascular disease, infectious disease resistance, and oncology¹⁵.

In conclusion, AI-driven biomarker discovery embodies extraordinary potential but faces equally significant practical limitations¹. The future of translational medicine depends not on replacing traditional science with AI, but on integrating AI as a powerful partner one that requires rigorous oversight, multidisciplinary collaboration, and continuous validation². With responsible development, AI can transform biomarker discovery from a promising vision into a clinical reality that improves patient outcomes worldwide³.