Differential Diagnosis In Surgical Pathology

How We Actually Work Through Suspicious Cases Before Rendering a Final Call

Most pathology residents think differential diagnosis is something you write on a case presentation slide. It's not. It's a structured process of elimination that happens inside your head while you're looking at a glass slide at 2pm on a Tuesday and you're already running late for lunch. I've spent years in surgical pathology and the cases that make you stop are the ones where the architecture screams one thing but the cytology whispers another. A wrong call here isn't like missing a benign variant on a cytology prep. We're talking about whether someone gets a Whipple procedure or stays in surveillance. Whether a breast lesion gets lumpectomy or mastectomy. The stakes compound quickly because treatment decisions branch off the initial diagnosis like a river delta. The framework I use doesn't change much between years. It starts with asking what the lesion is trying to tell you before you reach for immunohistochemistry. You look at growth pattern first. Then cytology. Then stromal reaction. Most people reverse that order and it leads to expensive mistakes.

The Step-By-Step Process I Use Before Ordering Panels

I start with architectural context. Where is this lesion sitting relative to the surrounding tissue? Is it pushing into things or infiltrating? Pushing borders suggest benign or low-grade processes. Infiltrative patterns with individual cell invasion shift my thinking toward malignancy regardless of how bland the nuclei look. Next I grade the cytology. I'm looking for nuclear pleomorphism, nucleolar prominence, chromatin texture, and mitotic activity. A lesion can look deceptively bland at low power and reveal aggressive cytology at 40x oil. I always go back to high power at least three times during a difficult case because the answer is usually hiding in the periphery of the field. Then I examine the stroma. Desmoplasia tells you something different than myxoid change. Inflammatory infiltrate composition matters too. Lymphocytes versus plasma cells versus eosinophils each carry diagnostic weight depending on the context. I once spent twenty minutes staring at a gastric biopsy because the surface epithelium looked reactive but the lamina propria had a subtle monomorphic lymphoid infiltrate. That turned out to be early MALT lymphoma misdiagnosed initially as chronic gastritis. The key was the preservation of lymphoepithelial lesions at the edges where I should have looked harder the first time.

Practical War Story: When Two Entities Look Identical on H&E

Three years ago I received a resection specimen labeled as suspected inflammatory myofibroblastic tumor. The morphology was classic enough on H&E. Spindle cells, collagenous stroma, mixed inflammatory infiltrate with prominent plasma cells. ALK staining came back negative which threw the diagnosis into question because roughly forty percent of IMTs harbor ALK rearrangements. The rest don't show staining even though they're real tumors. At that point I had to differentiate between inflammatory myofibroblastic tumor, nodular fasciitis, IgG4-related disease, and a few rarer entities that mimic each other. I ended up running a panel: SMA, desmin, CD34, STAT6, IgG4, ALK IHC with both clone 5A4 and D5F3, and EBER in situ hybridization. The morphology pointed toward IMT but the negative ALK made me uneasy. The STAT6 ruled out solitary fibrous tumor. The IgG4 staining showed fewer than five positive plasma cells per high power field which effectively excluded IgG4-related disease. I went back to molecular testing and ordered ALK FISH which picked up a rearrangement that IHC missed. That case taught me that negative immunohistochemistry never equals absence when you're working with sarcomas and myofibroblastic proliferations. It just means the protein isn't being overexpressed to detectable levels.

Common Pitfalls That Waste Time and Risk Patient Care

The biggest mistake I see is ordering immunohistochemistry before you've exhausted the morphologic differential. Panels are expensive and turnaround takes time. More importantly, antibodies are tools not crutches. If you don't have a specific hypothesis about what you're looking for, you'll generate results you can't interpret reliably. Another trap is anchoring bias. Once you land on a diagnosis early in your assessment, you tend to see confirming features and ignore disconfirming ones. I catch myself doing this on high-volume days. The workaround is simple but hard to maintain: write down the top three differential diagnoses before you look at any ancillary tests. If your final diagnosis isn't one of those three, you have to explain why. A third pitfall is underestimating the value of clinical correlation. A radiologist's impression or a surgeon's note can change everything. I've had cases where the morphology was ambiguous but the clinical history provided the missing piece. Don't wait for the final report to read the clinic notes. Read them before you start scanning.

When to Escalate and When to Stop

There's a threshold where additional testing yields diminishing returns. If you've exhausted morphology, clinical correlation, and a focused immunohistochemical panel without resolution, the next step is molecular testing. Next-generation sequencing panels for soft tissue tumors and certain carcinomas have made this more practical than it was five years ago. Cost is still a factor but it's decreased significantly. Sometimes the right answer is to render a diagnosis of undifferentiated neoplasm and recommend further workup. That's not a failure. It's honesty. I've seen junior pathologists stretch to force a specific diagnosis when the evidence simply doesn't support it. That leads to treatment errors downstream.

Building Your Own Diagnostic Framework

Over the years I've developed mental algorithms for common scenarios. Carcinoma versus lymphoma in a poorly differentiated tumor. Melanoma versus carcinoma in a spindle cell lesion. Sarcoma versus reactive proliferation in a deep soft tissue mass. These aren't rigid protocols. They're starting points that get refined with experience. The best resource for developing this thinking is not a textbook. It's reviewing your own cases in follow-up. When a patient comes back with recurrent disease or metastatic spread that contradicts your initial diagnosis, that should feel uncomfortable. It should make you go back and figure out what you missed. That process of self-correction is where actual expertise develops. For those who want structured reference material, the WHO classifications remain the standard. Digital Atlas of Surgical Pathology and resources from the College of American Pathologists provide case-based learning that's closer to real practice than lecture slides. Case studies with follow-up data are particularly valuable because they show you what happens after the diagnosis is rendered.

Tools and Resources That Actually Help With Differential Diagnosis In Surgical Pathology

There isn't a single downloadable tool that solves diagnostic uncertainty. What helps is a well-organized personal archive of cases you've encountered. Mine consists of annotated image sets organized by anatomic site and diagnostic challenge. Each entry includes the differential I considered, the reasoning process, and the final outcome. I update it constantly. Open-source image analysis software like QuPath has become useful for quantifying immunohistochemical staining patterns. It won't make the diagnosis for you but it removes subjective estimation from antibody interpretation. When you're trying to distinguish between a benign and malignant process based on Ki-67 labeling index, having a precise number matters more than a visual estimate. For immunohistochemical panel selection, the literature on antibody validation and specificity is worth reading rather than relying on vendor recommendations. Many antibodies perform differently in surgical pathology fixed tissue than they do in research settings. Clone selection matters enormously. The same target stained with two different clones can produce completely different results and lead to opposite conclusions.

I also rely heavily on peer consultation through platforms like PathOut or informal networks with colleagues at other institutions. A fresh pair of eyes on a difficult case often catches something you've gone blind to through prolonged exposure. This isn't weakness. It's risk management. The core of differential diagnosis in surgical pathology is disciplined thinking under time pressure. The methods are straightforward. The difficulty lies in applying them consistently when you're handling twenty cases before noon and the eleventh one doesn't fit any of your mental templates. That's when the process matters most and that's when most errors creep in. Slow down at the critical moment. Look again. Write down your differentials. Order targeted tests instead of shotgun panels. And when you genuinely don't know, say so.