OHDSI Home | Forums | Wiki | Github

Potential for Cancer/Oncology Studies in OHDSI--Need your thoughts!

Hi Ray, how about combining patient similarity research and comparative effectiveness research to stratify cancer patients by similarity and compare their treatment outcomes (e.g., post-treatment survival analysis or morbidity)? -Chunhua

1 Like

A student of my lab is currently investigating the trends of cancer interventions in the clinical trials space. Another potentially interesting question is to compare this research trend to the trend in real-world patient data and to see if there is any discrepancy, gap or lag between research interventions and patient care therapies.

Thanks @chunhua ! Any particular types of cancer and/or particular treatments in particular or just treatment outcomes in general among similar patients on different regimens across all different types of cancer?

@rchen
We have Korean representative database with cost information. I’m so interested in this topic, too!

‘Temporal trend of cancer incidence, overall survival and medical expenses’ would be one of the most important questions, which is feasible in current OHDSI. There have been many studies investigating trying to answer this question. But none of them used same analytic protocol and code across the countries.

It would be great
if we can compare the improvement of overall survival in each cancer across the countries.
or if we can compare the rise of medical expenses for cancer across the countries.

2 Likes
  • What Cancer Diagnoses:
    What were the patients’ cancer diagnoses? Described including histology. Ideally, also described including stage, grade and clinical/molecular tumor characteristics.

  • When Diagnosed:
    When were the patients first diagnosed with cancer? Ideally, via pathological confirmation.

  • How Treated:
    How were the patients treated? Described including type of treatment, constituent agents, begin date, end date, dosing and number of cycles/fractions. Ideally, also described in terms of standard-of-care or novel regimens.

  • When Recur/Progress:
    When did the patients’ cancer diagnoses recur or progress? Ideally, via pathological or imaging confirmation. Described including histology. Ideally, also described including stage, grade and clinical/molecular tumor characteristics.

  • What Vitality/Quality of Life:
    What were the performance statuses of patients over the course of cancer diagnoses? What were the patients reported outcomes over the course of cancer diagnoses?

  • When Deceased:
    When did the patients die? Described including cause of death.

These questions, while all very important, are generally impossible to answer without SEER level cancer registry data linked or loaded into the CDM.

Something that might be easier is to measure primary cancer time to treatment (resection, chemo-rads, immunotherapy) stage shift and then extent of followup. Variation in followup has been documented, but the “costs” in terms of overall survival and $ and evidence as to best practices are not well known. Its pretty much the Wild West in terms of post-treatment follow-up.

1 Like

Exactly. We are trying to implement a Sheriff in each town. :slight_smile:

Adding to Michael’s list:

  • Cost of treatment
    What is the cost of the treatment itself, and what is the overall cost of treating a patient including everything that is going on.

  • Side effects of treatments, and treatment of those
    Cancer treatment are usually pretty brutal, given the alternative, but finding out what the problems are and how to keep them in check is important.

  • Cancer screening
    Performance characteristics, burden and correlation with overall survival. There are plenty of debates about these screening tests like in the malignancies of breast and prostate. Some of them try to inform drastic vs not-so-drastic treatment options. Some of the screening tests have a high burden themselves, e.g. prostate biopsies.

  • Genetic characterization of the Conditions and effect on prognosis and treatment efficacy
    Traditionally, this has been the domain of histology, but increasingly it becomes clear that the underlying genetic drivers (and hence treatment choices) do not always jive with the traditional histology in prognosis and treatment choices.

1 Like

I will add my vote to cancer screening described nicely by Chris. In addition to the survival analysis listed but also add overdiagnosis; we can look at tumor size distributions.

@Christian_Reich
In this European OHDSI, we’ll prepare the poster visualizing relationship between genetic characterization according to the lung cancer stage based on CDM!

Hi Ray,

It would depend highly on the data availability. As Deppen indicated, cancer registry data (including SEER) contain limited information on treatment follow ups.

As for a population-level cancer survival trend study, from my experience, trends in cancer survival need to be considered together with trends in cancer incidence and mortality. See if the following paper is helpful:
When Do Changes in Cancer Survival Mean Progress? The Insight From Population Incidence and Mortality, JNCI Monographs, Volume 2014, Issue 49, 1 November 2014, Pages 187–197, https://doi.org/10.1093/jncimonographs/lgu014

1 Like

Thank you all for your thoughts and contributions! As part of the Oncology workgroup, we’re actually working (slowly) on incorporating and mapping our cancer registry into the CDM. Similarly, we hope side effects of treatments and treatment might able to be answered with our How Often/Incidence Rate project, the preliminary version of which was introduced during George’s talk at the Symposium.

Based on everyone’s understanding of the sites and databases currently in OHDSI, perhaps the question should be what would be the most impactful yet also lowest hanging fruit (in terms of data availability and what we accurately capture) study we can carry out?

Seems like there is interest in cancer screening test characteristics and survival (although we will need to be mindful of lead time biases). How good are we at identifying mortality in this population? There’s also interest in different aspects of treatment; how confidently can we identify what is treatment for cancer (often overlap with tx of other diseases)?

Also, some other topics of interest that I’ve verbally had conversations with others about recently:

  1. Palliative care/End of Life–where are patients at end of life? Hospital (and where in hospital) vs hospice vs home? What interventions are they receiving?
  2. Medication adherence
  3. Burden of comorbidities
  4. Length of stay; ICU stays
  5. Nutrition–tracking patient weights, potential associated w/ outcomes, tube feeds vs parenteral nutrition
  6. Complications such as infections or neutropenic fever
  7. Patient/family understanding of disease, informed consent, or of genetic counseling (likely can’t do with our data)
  8. Pain control

I like the idea of looking at cancer diagnosis and staging of diagnosis through the lens of whether that patient got screened for that cancer prior to diagnosis (thanks @Christian_Reich for the great idea!). So cut into two populations: those who got screened for the cancer he/she was diagnosed with and those who did not get screened for the cancer he/she got diagnosed with. Of those who were diagnosed with cancer, what percent got prior screening for cancer? I think, right off the bat, it would be interested to learn how many people who got cancer had prior screening. It would shed some light on the proactivity of the people in our respective databases. Are there noticeable differences in percentages among the respective sites? The next thing would be to get distributions of the stages for each population. Are the stages meaningfully lower for the screened population than they are for the unscreened pop? What about by cancer? This would have implications regarding prognosis.

@rchen
The study evaluating lung cancer incidence in different birth cohorts of US was published in NEJM
(https://www.nejm.org/doi/full/10.1056/NEJMoa1715907)

I think it would be interesting to evaluate incidence of cancer trends in different birth cohorts across the countries.

Yes @SCYou, I totally agree! Just saw that same study in one of my e-mail newsletters. Would be very interesting to try to execute this study internationally–especially with the high prevalence of smoking in South Korea (particularly among men) and Europe .

Perhaps we should create these cohorts locally then start a new thread to recruit collaborators?

@rchen
Good !

Korean National Health Insurance data has health exam and survey data such as BMI, exercise and smoking status and the cost.
I felt that female incidence of non small cell lung cancer in Korea became increased when I worked as a doctor. Air pollution is another big problem in Korea.

And we can analyze other cancers such as breast cancer or colon cancer. (In Asia most patients with breast cancer is young, while most are old in Western countries)

I can make a code for this too!

Hi everyone,
We’re planning to convert whole Korean cancer patients data into CDM from National Insurance data of HIRA (Korean national insurance data covers almost 99% population of Korea. This insurance covers 95% of cancer-related claim (If the patients should pay 100$ for the treatment, it covers 95$). Then, we can run @rchen 's treatment pattern in cancer patient on much bigger data.

We’ll perform descriptive analysis about incidence, overall survival and the whole cost within 1, 3 and 5 years after diagnosing of cancer in each age/gender group.

If other sites can participate, we compare the incidence, survival and the cost information year by year across the countries. The novel drug continues to be developed in oncology. Usually, these novel drugs are introduced in Korea 1~2 years after approval of American FDA. I want to answer to this question: “Does new drug or money cure more patients than health system?”. The health system can include early screening strategy of the country, insurance system, novel drug approval system, and other health-related situation such as obesity prevalence, air pollution, and etc.

Because I’m not the expert on the economic research, I need the help @Gowtham_Rao.
And I need to listen your thoughts, @rchen @Christian_Reich @mgurley @Patrick_Ryan @hripcsa

3 Likes

@SCYou this is fantastic. I think it would be a great opportunity for us all to collaborate! I sent you a message

1 Like

That’s very exciting @SCYou! Amazing to be able to cover essentially the entire Korean cancer population! Once it’s converted, it seems like a great opportunity to pursue the cancer by birth cohorts question you posed earlier based on the NEJM paper. Could even figure out a way to look at many different types of cancers across birth cohorts, maybe every type of cancer then run it across the network :slight_smile:

And your health system and environmental factors question is also interesting but may be quite complex. Because even within a specific country, especially a larger country like the US without nationalized health care, there can be huge variations in practice patterns, environment, demographics across regions or even on the hospital or provider level within a region. May need to further refine that question in order to get a meaningful answer, but certainly a good topic worth exploring

1 Like

@rchen thank you for your interest!
Yes, I have kept the NEJM paper in my mind to extend the result to other cancers, such as breast cancer (In Asia, most breast cancer patients are young, opposite to the Western patients).

I wrote that I wanted to answer ‘Does new drug or money cure more patients than health system’, but I don’t think we can answer this question with this simple protocols. But this would be the first step, for sure. Because this is the first study using HIRA’s whole cancer patients, I wanted to start with simple protocol first. We need to refine the protocol after that as you mentioned.

Still, the comparison of overall incidence, mortality and the cost of each cancer across the countries would bring huge impact.

1 Like
t