Cigar Science with Rob VII: Cigar smoking and health, study #1

[bobbyg29-cl]

NOTE: Please read the discussion of Odd’s Ratios in Cigar Science VI before reading this post.

PREFACE: For the sake of objectivity, I thought it would be appropriate to tell you a little bit about my background so that you can take it into account when reading the analysis below. I have been a researcher in Psychology (in particular visual perception in driving, sports, and aviation) for about 14 years. I have about 50 publications and am a reviewer for several journals in my field. So I think I am pretty good at sifting through scientific mumbo-jumbo and critiquing research. In no way would consider myself an expert in the field of health research—I have not done any work in that area. I am just an interested reader

STUDY #1: The first study I want to review looks at the relative effects of cigarette, pipe, and cigar smoking on lung cancer. If you would like a copy of this paper please PM me your email address.

Boffetta, P., G. Pershagen, et al. (1999). "Cigar and pipe smoking and lung cancer risk: A multicenter study from Europe." Journal of the National Cancer Institute 91(: 697-701.

Basic design:
-5600 male lung cancer patients from Cancer centers in Germany, Italy, and Sweden analyzed between1988-1994.
-All given questionnaires used to assess smoking habits (type of tobacco, amount smoked, age when started, etc). In some cases family members were interviewed for deceased patients. To me this is a bit problematic b/c I know personally that my family could not give an accurate report of my smoking habits.
-Based on the questionnaire data subjects were divided into the basic categories shown in Table 1.
-7255 control subjects (that did not have lung cancer) were obtained from a random sample of the male population in these countries. They were given the same questionnaires and divided these into same categories.
-Taking into account exposure to workplace carcinogens did not effect the results described below.



Results:

Basic Probabilities:
-Table 1 shows the Odd Ratios for the different categories. Note that for this type of data there is actually a range of OR values (because individual people are different) as shown by the confidence interval (CI). This CI shows the upper and lower bounds of the OR. What is important to note is when the bottom bound is near 1.0 it means that for a lot of people there is no significant increase in risk.

I am not going to spend much time talking about the OR’s due to the problems with interpreting them described in my previous post. Instead let’s look at actual probabilities. Note these are just approximate calculations that I did from the data. Here are percentage chances that a person in a particular category will have lung cancer i.e., # cancer patients/ (patients + healthy controls):

Nonsmokers: 7%
Cigar smokers: 28%
Cigars+ cigarillo smokers: 36%
Pipe smokers: 32%
Cigarette smokers: 52%
Mixed smokers (cigarettes plus pipes and/or cigars): 47%

So by these calculations your risk of developing lung cancer is roughly 4 times greater if you are a cigar smoker, 5 times greater if you are a pipe smoker, and 7 times greater if you are a cigarette smoker or a mixed smoker.

Effect of amount smoked, age, etc:
Table 2 shows the OR’s for these different variables for combined cigar and cigarillo smokers. Note that this analysis could not been done for pure cigar smokers because there were not enough of them. Again let’s pull out some of the basic probabilities:



Average amount of tobacco consumed/day:
-For this study it was assumed that 1 cigar=4g, 1 cigarillo=2 g of tobacco. Here are again the percentages calculated as described above:

0.1-5g (up to 1 cigar) =18% * Note in Table 2 that the bottom end of the CI is near 1.0
5.1-12g (1-3 cigars) = 29%
12.1-15g (3-4 cigars) = 31%
>15.1 g (more than 4) = 55%

Age at start of tobacco use:
<19 years = 50%
20-26 years = 41%
>27 years= 17% * Note in Table 2 that the bottom end of the CI is near 1.0

Years smoking:
0.1-13 years = 16% * Note in Table 2 that the bottom end of the CI is near 1.0
13.1 – 26 years = 20%
26.1-39.0 years = 41%
>39.1 years = 54%
*Note these year smoking effects are slightly confounded by age; you have a greater chance of getting lung cancer as you get older.

Effects of inhaling
-I could not do the percentage calculations on these data because not enough detail was provided in the study, however out of the 74 cigar+cigarillo smokers that did provide information about inhalation there was a 5.2 OR for non-inhalers and a 28.1 OR for inhalers.

MY CONCLUSIONS:
1)The general result from this study is that smoking cigars leads to a elevated risk of lung cancer (whether you inhale or not) that depends on the amount smoked, number of years smoking and whether you combine cigars with other types of tobacco. I don’t think any of us is surprised by this.

2)If you smoke 1 cigar or less per day your risk of getting lung cancer is only slightly elevated as compared to a non-smoker (18% vs 7%) and depending on the individual there may be no significant increase in risk at all (bottom range of CI is near 1.0).

3)If you smoke more than 1 cigar/day you increase your percentage risk of lung cancer by about 12% compared to those that smoke 1 or less/day and by about 23% compared to a non-smoker.

4)Since the dosage effects in this study are based on the total weight of tobacco smoked it could be inferred that smoking smaller vitola cigars will lead to less of a risk of lung cancer although this was not tested directly.

5)One of the main things that effects your risk of getting lung cancer is the cumulative amount of tobacco smoked in your lifetime so you are more at risk if you started younger and smoke for a larger number of years. There are also major benefits to quitting cigars when you are older which I will discuss in future posts.

6)There are some important limitations of this study that need to be addressed including the relatively small number of pure cigar smokers in the study, the unusual way of breaking down the amount smoked, the fact that a lot of the cigar analyses combine both inhalers and non-inhalers, and the fact that smoking and general health behaviors may be different in Europe and the U.S.

Please let me know if you have any questions...

Cheers!

Rob

[JSC]

Very interesting and informative, thank you for taking the time to post this.

[bobbyg29-cl]

Originally Posted by JSC

Very interesting and informative, thank you for taking the time to post this.

Thanks Jake!

[g8trbone]

Thanks Rob... Your interpretations are what help the study make sense. It would take me a long time to figure out what those studies meant.

[countnikon]

Good post man. I usually have about 5 cigars a week so it is nice to know that I've only got a slightly elevated risk.

[Corona Chris]

Thanks for the post.

[Ceedee]

Rob, that's great stuff. Thanks for taking the time to post the info, I for one appreciate it!

CD

[kass]

Another great post Rob! Thanks.

[Cigarfiend1]

Good Post, now if I had read this article without you posting what OR's are I would have been lost at the first diagram. Thanks Rob for all your time.

[tobacmon]

Rob you are the Mad Cigar Scientist...This was a very interesting post and learned things about cigars that I would have never known if it were not for you. Good Job Bro!

[Augie754]

If you would like more info here are some links I've put together for myself. I've also threw together a excel spreadsheet of the risk for cigars and cigarettes from the cancer.gov website. This is one of the larger studies done on cigars and the health risks. If you would like a copy of the spreadsheet, click the link.

Unless you know how to read these, it my be boring. You should know what relative risk, confidence intervals, and confidence values are and what they mean in these studies.

http://www.augenstein1.com/files/TobaccoHealthRisks.xls

Info was gathered from this document
http://cancercontrol.cancer.gov/tcrb...phs/9/m9_4.PDF

If you haven't read health studies, read this page first so you know what the numbers mean! A 1.5 RR is a 50% increase, that sound high, but doesn't mean that much statistically, so read the following to understand why.
http://www.davehitt.com/facts/epid.html
Make sure you read about Relative Risk and CI (Confidence Intervals), this site also has a ton of good info on Second Hand Smoke.

More Links to tobacco health related sites that are not by the crazed anti-tobacco people.
A cigar smoking doctor's interperation of the above NCI's study
http://www.cigargroup.com/faq/health/

More Second hand smoke info
http://www.joejackson.com/smoking.php His Ebooklet was just published in the Winter 07/08 edition of Cigar Magazine. You can download it here. It has further links.
http://tobaccoanalysis.blogspot.com/

[foolwithnohair]

Originally Posted by augie754

http://www.davehitt.com/facts/epid.html
Make sure you read about Relative Risk and CI (Confidence Intervals), ...

Everyone has heard the phrase "lies, damned lies, and statistics." Well, I have a bit of an example in the above link. (I'm not referring to Augie's post.) Here's a passage quoted from the link noted above. Almost everything that is stated is true, but there a few things that are not quite right. Here is the section in the above link on Confidence Intervals:

[SIZE=+1]"Confidence Intervals [/SIZE]

Fact: The Confidence interval (CI) is used to determine the precision of the RR. It is expressed as a range of values that would be considered valid, for instance .90 – 1.43.

The narrower the CI, the more accurate the study. The CI can be narrowed in many ways, including using more accurate data and a larger sample size.

Fact: Confidence intervals are usually calculated to a 95% confidence level. This means the odds of the results occurring by chance are 5% or less.

This is one reason epidemiology is considered a crude science. (Imagine if your brakes failed 5% of the time.) The EPA, in their infamous 1993 SHS study, used a 90% CI, doubling their margin of error to achieve their desired results.

The RR could be any number within the CI. For instance, an RR of 1.15 with a CI of .95 – 1.43 could just as well be a finding of 1.25, an 25% increase, or .96, a 4% decrease, or 1.0, no correlation at all. Pay close attention to any study where the CI includes 1.0. (It does in virtually all ETS studies.) When the CI includes 1.0, the RR is not statistically significant."

Let's look at the last paragraph, and the last sentence of the quote (italicized by the original author). Each sentence in the paragraph is true - but doesn't the paragraph make you think that if the CI includes 1.0, then the study didn't find any relationship? What does "statistical significance" mean, anyhow?

My way of thinking about statistical significance is that significance statistically is about as relevant as asking my banker, "What are the interest rates on this mortgage?" and they answer "Yes, we think that there will be an interest rate" or "No, we do not think that will be an interest rate."

What do I mean? I mean that if the implied question is "how good are these results," then saying that the results are "statistically significant" is not really a useful answer.

To give some background, we have to determine what makes results good or bad. When a study or test is done, there are four possibilities:

a) the study shows an effect, and there really is an effect. For example:
- the pregnancy test was positive, and she is pregnant
- known user-id's can logon;

b) the study shows an effect, but there really isn't an effect
- the pregnancy test was positive, but she isn't pregnant
- known user-id's can not logon;

c) the study does not show an effect, but there was an effect
- the pregnancy test was negative, but she really is pregnant(!)
- unknown users can logon; and

d) the study does not show an effect, and there is no effect
- the pregnancy test was negative, and she really isn't pregnant.
- unknown users can not logon.

In other words, the study is "true" if a) or d), and the study is "false" if it is b) or c).

b) is called "Type 1 error," i.e., the study's results indicate a difference when there really is no difference (in statistician-speak, rejecting the null hypothesis when it is actually true). A false positive.

c) is called "Type II error" in statistics, a false negative, i.e., accepting the null hypothesis when the null hypothesis is false. In statistics, the lower the type II error, the greater the "power" of the test/study has statistically. Type I error has no effect on power.

Now, if one considers the pregnancy test example, type II error is the one that probably is of greatest concern. Sure, reducing type I error (b) is good, but (c) is usually the more important outcome to control.

Catch: It is impossible to measure type I or type II error except where predictions are being made (under some assumptions/conditions). It is possible to know for sure how good a pregnancy test was within nine months of the test :-). As soon as we know the truth, then we can measure the error, but generally there is no need for a study at that point in time.

With computers, it is somewhat possible to figure out a guess at what the type II error is. Before computers, it was practically impossible. It was, however, possible to figure out a way to guess at the type I error - if the guess was made before the study started.

In the early 1920's, a guy called Fisher realized that one could estimate - from the data collected only in the study - the probability of getting a greater yield from a different crop, e.g., corn, given that the yields were actually the same. This is called a "p-value." To simplify the manual calculations needed to calculate p-value, a type I error of 5% was assumed. (5% gives approximately 2 as the test statistic cutoff.) That means 5% of the time - e.g., every more-or-less twenty seasons, test plots could indicate one type of corn had a different yield from a different variety, but in reality both varieties of corn had the same yield. The other 95% of the time, or 19 times out of 20 seasons on average, the test plots would show a difference in yield, and the yields would really be different.

Now, we do have computers. It isn't 1925, and epidemiological studies can't be designed nor analysed (statistically) like crop yield experiments. We don't have to use look-up tables, and we don't need easy to compute-with numbers, yet that is the reason that 5% was chosen.

Today, every software spits out p-values, and p-values are somewhat related to type I error. So if a study or test comes up with a p-value of 5% or less, it is "statistically significant," but if the p-value is more, than it isn't. That's all it means, that the probability that a study shows an effect, given that there really isn't an effect. That's it, that's all - and we still know nothing about the type II error - being pregnant when the test said we weren't.

I hope that the above gives you the idea that considering "statistical significance at the 5% level" as the sole judge of the quality and/or usefulness of a study is not a good thing.

Links:

More details and one way to look at p-values in medical studies: http://www.bmj.com/cgi/content/full/322/7280/226

A longer and better explanation of “Statistical Significance” http://marketing.wharton.upenn.edu/i...gnificance.pdf

More on 5%: http://www.tufts.edu/~gdallal/p05.htm

Current theoretical foundations (textbook):
Royall, Richard M. 1997. Statistical evidence: A likelihood paradigm. New York: Chapman and Hall.

[6clicks]

Does anyone have any info on the relationship between non-inhaling cigar smokers and heart disease? Doesn't the nicotine still constrict the blood vessels or cause some other problems? I smoke about 3 a week. I'm not really worried (although I have stents in my heart) but I am curious.
George

[DocLogic77-cl]

Originally Posted by 6clicks

Does anyone have any info on the relationship between non-inhaling cigar smokers and heart disease? Doesn't the nicotine still constrict the blood vessels or cause some other problems? I smoke about 3 a week. I'm not really worried (although I have stents in my heart) but I am curious.
George

Yes, nicotine increases blood epinephrine levels which speeds up heart rate and constricts blood vessels. Having a history or coronary heart disease and stent placement...it is a very bad idea for you to smoke anything with nicotine. Vasoconstriction can reproduce ischemia symptoms (angina) and possibly disrupt plaques which can break lose and travel to smaller diameter arteries which can cause a MI (heart attack). Increased heart rate causes cardiac tissues to demand more oxygen which can be a serious issue if you also have vasoconstriction and an unstable plaque. I don't mean to be a downer but I think you need to know the risks. While moderation is a wonderful thing...I recommend to all my patients that anyone with heart disease refrain from nicotine and even caffeine. I hope that helps.

[6clicks]

Thanks doc. I spoke with my cardiologist last week and he approved my having a smoke a week. I took it upon myself to have 2 Rothschild sized smokes per week instead of one Churchill. I am on a 1 hour per day exercise program since January and have lost 35 lbs to date. Your advice is truly appreciated and will certainly keep me from pushing the envelope further.
George