Casey Kim, MD

Still, a cool paper.

I think it shows:
- Further dev/integration of AI predictive models for personalized medicine in Oncology
- Modeling to help better ID which patients could benefit from earlier palliative care/geriatrics

As pt live longer w/ cancer, this will only 📈

CRP:Albumin ratio was also highlighted as a key predictive factor.

Easy to get or include with monitoring labs for patients getting treatment. The only issue is on its own, seems to have fairly mild predictive effect (at least as estimated in the Cox proportional hazards method)

What is the G8 score?

It’s a screening for frailty in geriatric cancer patients that involves BMI, weight loss, polypharmacy, dementia/depression, mobility.

The big con I see is that this score isn’t computed for most oncology visits & isn’t abstractable from just chart review

Using data on French pts > 70 yo with cancer referred to geriatrics, created predictive models for mortality:

1. Cox proportional hazards

Machine learning based:

2. Single Decision Tree
3. Random Survival Forest (best)

🔑 predictors: G8 score, tumor site/mets, & CRP/albumin

“How much time do I have left?”

As an oncology hospitalist, this is the most common ? I hear from my patients.

An interesting paper just came out in JCO using Machine Learning to predict mortality for geriatric oncology patients: ascopubs.org/doi/full/10....

My thoughts 👇

It's flu season, so that means it's also COPD exacerbation season in the hospital.

I always got a little confused as to what to do with all of their inhalers, but the gist is:

- Continue their long acting inhalers
- Duonebs will cover their rescue inhalers

Back to our pt, after the scan showed likely LC, he decided to stop tx and go home with hospice.

If time was short, he wanted to spend with family, not healthcare system

My take away point: Dx of LC can ground prognosis/time (months) for patients to help guide goals of care

The data is not great on prognosis (only case series b4 2000s), but roughly 2-7 months.

Even patients with resectable lung cancer (so not metastatic) had earlier mortality by about 2 years if they had positive post op cancer in lymphatics than those who did not

Most commonly associated with:

"BLS" 🚑

Breast cancer (most common)
Lung cancer (2nd most common)
“Stomach” Gastric cancers (3rd most common)

Diagnosis:

Usually t/ clinical presentation + imaging
- CT scan: interstitial thickening, nodular opacities, and Kerley B lines
- PET: high specificity (100%) and sensitivity (86%), but hard to get in inpatient

Transbronchial biopsy is technically gold standard, but more morbid

What is lymphangitic carcinomatosis (LC)?

LC occurs when cancer cells infiltrate the lymphatic vessels of the lungs, often leading to poor gas exchange → leading to dyspnea, cough, and fatigue

It’s often a 🚨 for poor prognosis - median survival is on the scale of months

I took care of a patient with lung cancer who went to the ICU for respiratory failure of unclear etiology.

CT chest done that suggested “lymphangitic carcinomatosis”

Though he got better, that dx changed what happened next for him

A 🧵 on lymphangitic carcinomatosis

Ever wonder if a certain symptom your patient with cancer is endorsing is related to their chemotherapy?

The short answer is usually "maybe", but here's a helpful chart of common toxicities by chemotherapy classes.

From hemeoncnotes.com, the solid cancer intro lecture!

7/ For more tips on biopsies and a general introduction to solid cancers, check out the Solid Cancers Introductory Lecture [Residents] on hemeoncnotes.com

Cancer of Unknown Primary Site: A New Era of Practice-Changing Approaches to Diagnosis, Staging, and Precision Therapy | JCO Oncology Advances The enigmatic syndrome of metastatic cancer of unknown primary (CUP) site has frustrated physicians and patients for decades. There has been debate whether CUP is a single biologically distinct cancer...

6/ Cancer of unknown primary

Path can show "carcinoma of unknown primary”

The cancer is so undifferentiated, there are no specific tissue markers. Repeat biopsy usually not helpful unless you don’t have enough tissue to run molecular/genetic studies.

ascopubs.org/doi/10.1200/...

Step 4: Interpret your biopsy result

Biopsies are a combination of:
- Histology: Tissue type
- IHC: Markers that can help determine origin & subtype
- Molecular: Genetic testing (can take longer to come back)

Ex: "Adenocarcinoma of lung [histology], positive for TTF-1... [IHC]

Step 3: What to biopsy?

Unfortunately, your pt may have multiple potential biopsy sites. Which one to target?

In general, you want your target to be: :
- Metastatic site over primary site
- Safe & easily accessible
- Core biopsies better than FNA [more tissue = more studies]

Step 2: Do you need a biopsy?

Usually the answer is yes, but there are some exceptions.

- Notably HCC can be diagnosed primarily through imaging (triple phase CT or MRI)
- Some cancers (RCC, testicular, CNS) may require total excision as opposed to biopsy

Step 1: Where is the cancer?

Get a CT Chest/Abd/Pelvis with contrast. [Don't bother w/ inpatient PET]

This is fast and will tell if the cancer is localized or metastatic, which will help guide next steps

🧠 scans not part of initial staging unless syx

"Tissue is the issue"

For new pts admitted for new malignancy w/u, almost universally, the next ? is “is the biopsy done?”

Seems obvious, but choosing when and where to biopsy has some nuance and directly impacts time to dx and tx

A 🧵 on biopsy tips and tricks: