Cancer Epidemiology - Science topic

I agree with @James Leigh; three might be an error in the modeling.

The other possible explanation could be the fact that the attributable factor for lung cancer mortality is greater than the lung cancer diagnosis . . . .

Well, the cohort study is highly suggested in covid pendamic. however, the covid-19 is no more a cohort situation it is done for observation and per and post-test by training and other medication.

Today, August 6th, 2020 is the 75th anniversary of this tragic event, part of the even greater tragedy of the 2nd World War. I am surprised that no RG follower has offered a response to the question posed, given the significance of the event. On this day however, it is noteworthy that Hiroshima's mayor took the opportunity to warn the world about the rise of "self-centered nationalism" and appealed for more international cooperation to overcome the Covid-19 pandemic. Apparently, memorial events have been drastically scaled back because of the pandemic.

Hi

LGR5 is the best biomarker for detection of cancer stem cells in colorectal patients .

u can browse our my profile and see our paper about that

Regards

The SEER database.

Whenever you intend to study cancer, anyone, just try GLOBOCAN. It is no1 site most referenced and updated with all cancer info including types, incidence, mortality, year-by-year (even up to this moment), globally, intercontinental, nation-by-nation etc. You can download datasheets with statistics. You everything going on about all cancer types in the globe with free access. Check the link and the sample test of what to expect below;

http://gco.iarc.fr/today/fact-sheets-cancers

Hi Philipe,

I tend to agree with James, the first approach of selecting the controls with data up to a specific timepoint seems much easier.

Regards,

Hello, quality of food is decreasing due to know how technologies food additives, new pesticide residues, new chemicals and harm environments.

Hi Aparna,

Short courses on cancer epidemiology and biostatistics vary depending on what you have in mind (online / taught / duration) and how much you are willing to pay.

The international agency for research on cancer run summer schools on cancer epidemiology and the next one is 17th June - 5th July 2019 in Lyon, France with various modules: https://training.iarc.fr/summer-school-in-lyon/

There are also regular STATA Summer schools in London with various modules covering medical statistics and meta-analysis: https://www.stata.com/news/stata-summer-school-london-2018/

I attended one of these and found it very useful as well as learning to use STATA.

You can also explore various specific University modules offered in most places if you wanted a taught short course or even free online modules using MOOC - Massive Open Online Courses: https://www.mooc-list.com

I undertook an online course on medical statistics at the Stanford University and it was free and excellent. https://online.stanford.edu/courses/som-y0007-statistics-medicine Professor Kristin Sainani was awesome in delivering the lectures and you could always study at your pace and review video lectures.

Hope you find these suggestions useful.

Kind regards,

Oladejo Olaleye

The evidence for a shift or transition in the age risk of NCDs in LMICs need careful examination. A big international team solicitored by the WHO could be very helpful. Impressions are growing that NCDs risk is increasing with "extension"  not necessary shift towards younger age group. The main determinant is this contesxt is likely to be the  massive change in lifestyle behaviour. However, aretificial cause of such rise and shift or extension should not be ignored. All countries including LMICs ones are witnessing development in their health care systems with more chances to earlier diagnosis, better treatment and improved recording of events.  These last factors could explain part of the apparent risk and extension. True rise remains, however, a real contributor and  this needs extensive work to support, quantify and explain.

You're asking a question that the pharmaceutical industry, and biomedical science in general, has struggled with for decades.  Indeed, target discovery and validation is every bit as important as drug discovery when it comes to treating diseases, be they genetically driven or not.

Many people expected the Human Genome Project to essentially answer the questions you're asking - clearly it didn't.  While fabulously useful, the HGP resulted in very, very few simple answers or advances.  Similarly, to take just cancer: it doesn't have *a* genetic cause, it has a bunch of disparate causes.  Consider Weinberg and Hanahan's reviews, where they lay out an argument for cancer being caused by 12 key pathways - that's still 12 genetic pathways, that can be altered at multiple points to offer the same effect.  Or Celiac disease: is it caused by a problem with tissue transglutaminase?  For some people yes, that seems to play a big role.  For others, it's something else entirely.

Regarding drug treatment: curing chronic conditions is difficult to impossible, hence them being chronic conditions.  But just because a condition is genetic doesn't make it chronic.  Take cancer again as an example: somebody can have a mutation to their Ras gene that normally leads to cancer, but never develop a tumor.  So if somebody with that mutation DOES develop a tumor, we imagine we could cure it, and not have the disease come back.  We are, frankly, pretty bad at this right now - even the best cancer chemical therapies have low absolute success rates, but we get better each year.

Now, something like, say, celiac disease?  Or depression?  Or hormonal disorders?  It's unlikely those will be curable, at least without constant chemical treatment.  It all comes down to what the base state is - is the problem just in a specific place (ie: a tumor) or is it all over (ie: celiac disease and the whole intestine).  If you can't imagine just removing a few (or a lot of) cells and having the person both be cured AND still be alive, chances are drugs won't cure them either, since that's metaphorically all drugs do - remove the bad cells from play.

As for CRISPR/Gene editing in general: that's certainly a research track.  But it's an incredibly controversial pathway in general too - with a drug, you can wash it out of the system if something goes wrong.  But a gene edit?  It's very, very hard to undo what you did.  And while a protein (again, say Ras) might have a mutation that leads to cancer...  what ELSE is that mutation doing?  How much of the rest of the host is mutated to work with the mutant protein?  What good is it to cure the cancer if the person's skin stops regrowing as it sheds, for instance?  But, of course, as you note: a gene edit *could* fix the problem, then and there, simply and safely.  Then you've got the issue of what counts as the *correct* form of a protein, and who decides.  We've generally hesitated to release genetically modified organisms into the wild in case there are unplanned effects (and there are ALWAYS unplanned effects).  Now, we can sterilize corn, or salmon, or whatnot - but we can't ethically sterilize a person for daring to get gene alteration therapy.  So we could only alter them to 'normal', right?  But what's normal?  Who defines it?  And if we CAN alter your genes to 'normal', then what's to stop you from altering them to 'better'?  If we can equate a gene with height, or high IQ, or blue eyes, or whatever...  why shouldn't people be allowed to get cosmetic genetic surgery, if others are allowed healing genetic surgery?  It's an ethically difficult field in a way that more traditional drug treatment is not (and even that is changing - I saw a report at one point that there are active levels of Prozac in the Thames.  How ethical is it to treat people with drugs if they'll stay in the environment, essentially poisoning other people, for decades?)

But if you want to dig into genetic causes, cancer is probably one of the best places to start.  Not that it's intrinsically more genetically based than any other genetic disease, but there are a LOT of publicly available resources out there.  In particular, take a look at the Cancer Genome Atlas, and the Cancer Cell Line Encyclopedia.  You can get raw data, or use various online tools, to poke around and see if you can find something nobody else has ever noticed.

Just rename the caseId to Id and then sort both dataset by Id and follow these codes:

data merged_data;

   merge set1 (in=a) set2;

   by Id;

  if a ;

run;

I hope it works!

Great, thank you so much for this information Dr. Hughson, this is extremely helpful. 

Lieber Orlando,

da es ein Comorbidity Score ist, sollte die Grunderkrankung nicht hinzugerechnet werden.

Liebe Grüße und ein Frohes Neues Jahr

Christoph

The size of the study (no. of cases and controls) will depend on which causal effects you wish to detect, which degree of increased risk you wish to detect at which alpha level with which power. There are numerous reference to methods in RG answers including mine.

I assume you will be looking at things like aniline dyes, HIV, PAH, tobacco, alcohol, radiation therapy to other organs, general radiation and other chemicals..

you will also need some a priori estimates of likely effects for power studies., based on the literature.

This is a good question.  Proper survival analysis (Kaplan-Meier curves, log-rank test, and Cox proportional hazards models) does account for the discrepancy in follow-up time. 

However, I think you can make a reasonable argument that the additional follow-up time for patients in the older era adds no value to a comparison of the three eras (i.e. if you're looking at survival for cases diagnosed from 1970-1979, 1980-1989, and 1990-1999, the events occurring 25+ years after diagnosis in the 1970-79 and 1980-89 cohort are more-or-less meaningless in a comparison against the 1990-99 era).

I think that it does make intuitive sense, in the type of problem you're describing, to truncate your analysis at the longest follow-up time for the shortest era (so in my example above, truncate your analysis at 25 years after diagnosis, since patients from the "latest" era of 1990-1999 could have as much as 25 years of follow-up, and at least theoretically you can compute an estimate of 25-year survival in that group; consider all patients from the 1970-79 and 1980-89 era who survived >25 years to be censored and still alive at 25 years).  Then you should be able to produce a KM curve going out to 25 years, and create a Cox proportional-hazards model with the "era" as a covariate in the model.

I also think it would be worth reporting some pre-specified time points (1-year survival, 5-year survival, 10-year survival) which obviously also does not have this issue.

For what it's worth, I have been working with a transplant surgeon on a similar question comparing survival after heart transplant in older eras vs. more recently, and have been amused at his seeming inability to understand this.  He has asked me to calculate things like "10-year survival" in a cohort of patients transplanted from 2010-2014 to compare it to prior eras (clearly not understanding that we cannot estimate 10-year survival in patients who have not been followed for 10 years).  So although this may sound elementary, you need not feel bad for asking the question; this is definitely something that confuses researchers across different fields.

Dear,

On simple note, you need following variables : Survival time, Events 

Parameters required

- cut off date for follow up or censoring date, or study end point date. (A)

-date of remission/ or date of surgery(B)

-date of diagnosis(C)

For DFS, survival time:  A-B

For OS survival time: A-C

Designate your relapsed patients 1 and rest 0,( For DFS) Event  will be 1

Designate your died patients 1 and rest 0, (For OS) Event will be 1.

Be careful to exclude the death other than disease burden.

Subsiquntly use any statistical package to get Kaplan Meier curves followed by log rank test  to compare between the groups

Best Wishes..

Surender

Thanks dear Bhandari

Hi Wang,

The link that Jignasa provided seems to have been truncated and doesn't work. The book description is on the IARC website at: http://goo.gl/iSh2Es

You should be able to find a copy for around 90USD: http://goo.gl/nYcRWQ

I don't even understand the question.   Are you asking about genes that are co-regulated with a specific oncogene or tumor suppressor gene?   Are you looking for all gene expression changes in a cancer, compared to normal tissue?  What do you mean by "given a cancer gene"?   Lastly, are you familiar with Oncomine?  

dear  Victor.

we wait the final pathology !

thank you!

Babu: In SEER Survival Session you need to check the box "Case Listing" in the "Parameters" tab. Than you need to define all your analysis variables as required by your project (site, age, sex etc...) and hit the "Execute" button and you will get the table with single line for every individual. The most important columns from the table will be "Survival Months" and "Vital Status Recode". You can use highlight all-copy/paste commands to get the data into Excel, save it in whichever format is needed and process by a statistical SW of your choice for tests that I mentioned above (after coding for dead/censored individuals vs time between diagnosis and follow-up).

sir Asharaf tumours histopathology report came as Benign ovarian tumour.patient went home attending school.

here's a start:

Burkitt's lymphoma in Africa, a review of the epidemiology and etiology http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2269718/

Burkitt Lymphoma: A Review http://omicsonline.org/scientific-reports/srep337.php

Rainey et al Spatial distribution of Burkitt’s lymphoma in Kenya and

association with malaria risk. Tropical Medicine and International Health 2007: 12(8): 936–943 doi:10.1111/j.1365-3156.2007.01875.x

Alex

I Guess you might have seen this already http://www.stata.com/meeting/spain13/abstracts/materials/sp13_zlotnik.pdf In short there is a way in Stata to create nomograms based on regression results but this presentation at least does not point to a ready-made package to download for sure. (Nomograms can vary a lot in how it looks and how many variables etc are used.)

We once contemplated a study of the salience of public health issues (in our case stress related) that were loaded up and viewed on u-tube. This can be monitored if you have a good tech person. Another metric to think about Tee.

Agree that smoking duration is more important than pack years. In my view the deleterious effects of smoking on CVD pertain to artherothrombotic effects and are cumulative. But those effects diminish over time with stopping. So classify smokers as current (smoking during the period of follow up for CVD events) or past (quit smoking before the period of follow up for CVD events) or recent (quit smoking during the period of follow up for CVD events). Add another variable that gives the duration of smoking (eg 5 years). If you have a large enough sample size, you could also add a variable for age at onset of smoking.

As an example, I smoked for 6 years, I began smoking at age 18, I am now 63.

So class me as past smoker , 6yrs duration, age of smoking onset 18. These variables would exert very little effect on my current risk for CVD, which is what you wish (ie dose response). A current smoker, who smoked for 20 years, who started at 18 and is currently 38 should exhibit a much higher risk for CVD than a non smoker of similar age

for an insightful framing of the scientific questions, look at The Natural Step's "system conditions" for sustainability. The Natural Step seems to be somewhat out of fashion among Americans working in this area, but the framework is immensely valuable conceptually.

Gurdeep:

You are right to suspect a residual mammographic-independent component of increased DCIS incidence:

The effect of screening programs on incidence of DCIS (per 1,000 screening mammograms) was examined using data from the BCSC (Breast Cancer Surveillance Consortium) and the NBCCEDP (National Breast and Cervical Cancer Early Detection Program) [1,2,3], finding (1) that the incidence of screen-detected DCIS was greater than the incidence of nonscreen-detected DCIS; (2) the Incidence of DCIS in the United States increased over time regardless of the definition used. But most relevant was: (3) that the data revealed greater increases over time in incidence per 100,000 population than per 1,000 screened, that is, that the incidence of DCIS increased over time, even when the rate of mammography was constant, suggesting that although the clear preponderance of cases accounting for the continued increase in DCIS incidence was secondary to upswings in mammographic screening (as supported in eight population-based trials of mammography screening), not all of this increased incidence could be so accounted for. In addition, note that we have differential incidence dependent on morphology or histological subtype: the incidence of non-comedo DCIS, that is tumors without comedo necrosis (these tumors not being associated with subsequent DCIS or invasive cancer) has generally increased across all age groups, whereas rates of comedo DCIS (a type of DCIS which is associated with subsequent DCIS or invasive cancer), has held constant or decreased [4,5].

So data support that there is a residual and non-trivial incidence of increased DCIS not dependent on the changing dynamics of mammographic screening. We are only now learning that DCIS exhibits some substantively different clinical behavior as well as natural history depending on not only histology and morphology but also on the subtype of DCIS (endocrine-positive, HER2-positive, or basal-like), and it is clear that we need more, and more mature, data, as to the differential dynamics of DCIS subtypes to clarify both the dependencies of incidence patterns and any clinically relevant consequences for therapeutic intervention.

References

1. Ernster VL, Ballard-Barbash R, Barlow WE, et al. Detection of ductal carcinoma in situ in women undergoing screening mammography. J Natl Cancer Inst 2002 Oct 16; 94(20):1546-54.

2. Smith-Bindman R, Chu PW, Miglioretti DL, et al. Comparison of screening mammography in the United States and the United kingdom. JAMA 2003 Oct 22; 290(16):2129-37.

3. Smith-Bindman R, Ballard-Barbash R, Miglioretti DL, et al. Comparing the performance of mammography screening in the USA and the UK. J Med Screen 2005; 12(1):50-4.

4. Li C, Daling J, Malone K. Age-specific incidence rates of in situ breast carcinomas by histologic type, 1980 to 2001. Cancer Epidemiol Biomarkers Prev 2005;14(4):1008-1011.

5. Virnig BA, Wang SY, Shamilyan T, Kane RL, Tuttle TM. Ductal carcinoma in situ: risk factors and impact of screening. J Natl Cancer Inst Monogr 2010; 2010(41):113-6.

Thanks Freddy! That's exactly what I was looking for.

A specific method for calculating confidence interval of Mantel-Haenszel Odds Ratio was first described in Clayton D. & Hills M. (1993) Statistical Methods in Epidemiology. Oxford University Press, Oxford. It is also reproduced in Page 183 of Essential Medical Statistics by Betty Kirkwood and Jonathan E. Sterne. The following steps should be followed to calculate the Mantel-Haenszel CI :

i) Here 95% CI ranges from OR/EF to OR*EF, where OR is Mantel-Haenszel Odds ratio and EF is the exposure factor;

ii) EF = exp (1.96* SE) , where exp is exponential and SE is the Standard error of the Mantel-Haenszel Odds Ratio;

iii) SE = Sqrt [V/(Q*R)];

iv) Now just think about a, b, c, d as the values the four cells of a 2 X 2 table of each stratum (Remember that using this approach we usually describe the simple Odds Ratio as (a*d)/ (b*c) ). Thus ai, bi, ci ,di represent the a, b, c, d values of the i-th stratum. Similarly ni is the sum of ai, bi, ci and di in the i-th stratum;

v) Using the above notational style,

Q = Sigma (i.e. summation of) [(ai*bi) / ni]

R = Sigma [(ci*di) / ni]

V = Sigma [(ai +bi)* (ci+di)*(ai+ci)*(bi+di)]/ [ni*ni*(ni-1)].

You can easily generate the 95% Ci in Excel itself provided you carefully translate the above to Excel formula. If you are still not sure, you can directly consult the books mentioned in the beginning.

You may find this article interesting.

For your formula, higher limit and lower limit of Ci should be taken natural log first

e.g. CI of OR (2, 5), after taking natural log, it is (0.693, 1.609),

SE=(1.609-0.693)/3.92=0.2337

remark: 3.92 is 1.96*2

The question is how will you manage to add the studies, from which results are not available? From my perspective the best thing you can do is to discuss possbile limitation of publication bias. You are aware that the estimation done by metaanalysis could be different because of unpublished results. The issue has to be discussed and explained. Stats Direct shows a plot, from whoch possible publication bias can be assessed, although you know for shure that there is not available data, because you encountered such studies. On the other hand, if the result of the metaanalysis shows a heterogenity test p<= 0,1, that could mean possible heterogenity of the studies included to metaanalysis. Then a sensitivity analysis is a must taking into account discriminating factors. More on that topic here: http://www.ncbi.nlm.nih.gov/pmc/articles/PMC1767262/

It is true that using continuous variables generally helps you get the most information out of your data, but there are times when using a cutoff is necessary, for example to establish treatment guidelines. Blood pressure is an example of this: Anything over 140 systolic BP is considered high blood pressure because BP over this level is associated with adverse health outcomes. It may also be used to guide treatment with BP medications. The 140 figure used to be higher until epidemiologic studies showed that systolic BP as low as 140 was associated with bad outcomes too.

The cutpoint at where to categorize your data variable can be based on many things. But first, if a statistical method (such as linear regression) can be used that uses the continous data, by all means use it. Logistic regression will let you use either continous or binary variables. One way you can tweak the cutpoints when categorizing your data in logistic regression is to do a sensitivity and specificity, and positive predictive value / negative predictive value with the results, to see how your regression model performs at predicting outcome at various cutpoints.

Also for putting your results into practical terms, consider a NNT (number needed to treat) analysis.

If you can use more than one cutpoint, you may wish to stratify your data and do a stratified analysis. This will (hopefully) for example establish an increasing risk of the outcome with increasing exposure---the classic example being increasing mortality rates as cigarettes smoked per day increases. (you can do this using one cutpoint too, but using three or more makes a stronger case for a dose-response effect).

Some studies use the bottom ten percent of a value for the cutpoint, for example birthweight, if you can find normative data with which to classify your data. Or you can convert your continous data to Z-scores or a percentile rank and use, for example, the lower 25% or upper 10% as a cutoff. It depends what your data variable is, of course.

In short, you just have to play with the numbers. But as others have said, try to find some basis for your cutpoint in the literature, or something that makes biological sense (physiology frequently exhibits thresholds, for example muscle fatigue or the transition of cells into carcinoma). Or consider cutpoints that have practical value, such as for diagnosing or treating disease, or a population-level health impact of the risk factor, including economic health impacts.

Yes, there is a clear inverse correlation between vitiligo and melanoma. Those with vitiligo have a 3-fold decreased risk of melanoma (Teulings, Br J Derm 2013). This is not surprising, since vitiligo results from a cytotoxic response directed against melanocytes (van den Boorn, J Invest Derm 2009), which are the neoplastic cells in melanoma. Also, the appearance of vitiligo in a patient with late-stage melanoma is a good prognostic sign (Nordlund JAAD 1983), with some developing complete remission. Identical antigens are targeted in both vitiligo and melanoma, however the affinity of the T cell receptor (TCR) is greater on clones isolated from vitiligo patients compared to melanoma patients (Palermo Eur J Imm 2005). Some have proposed to use vitiligo TCR expression as a therapy for melanoma patients (Circosta Hum Gene Ther 2009). There is also support for this inverse correlation from genome-wide association studies (GWAS) on vitiligo and melanoma patients. The tyrosinase risk allele for vitiligo is protective against melanoma, and vice-versa (Jin NEJM 2010).

Thanks. Yes, I would be interested in a citable PDF version so that I can send it to some course leaders and students and start a debate on this in the medical school.

The best studies to answer to this question come from cancer registries. In western countries, the SEER is the most accurate. Tumors are located with the same rate in the right colon, left colon and rectum.

(right colon 28%, sigmoid 23%, rectum 25%, rectosigmoid 10%: Anatomic subsite of primary colon colorectal cancer.... Cancer 2013)

I think the main problem with capture-recapture methods is meeting the necessary assumption of the heterogeneity of mixing in the population, so that the recapture is independent of the first capture.

This is the most reliable database for cancer epidemiology.