Mirror Sites:
CSUDH - Habermas - UWP - Archives

California State University, Dominguez Hills
University of Wisconsin, Parkside
Created: January 17, 2005
Latest Update: January 17, 2005
jeannecurran@habermas.org
takata@uwp.edu

Understanding Research Abstracts and Their Statistics

The following abstract was taken from a lecture on health and nutrition, and is repeated here for your convenience. In this exercise we shall explore some of the terms and what they mean.

Reduction of C-reactive protein levels through use of a multivitamin. From C-reactive protein on Website of Anti-Aging Research.

Am J Med. 2003 Dec 15;115(9):702-7.
Church TS, Earnest CP, Wood KA, Kampert JB.

The Cooper Institute, Dallas, Texas 75230, USA. tchurch@cooperinst.org

PURPOSE: Elevated C-reactive protein levels are associated with the risk of cardiovascular disease and diabetes. We examined whether multivitamins reduce C-reactive protein levels. METHODS: We performed a post hoc subgroup analysis of a 6-month, randomized, double-blind, placebo-controlled trial. Patients (n = 87; mean age, 53 years) for whom frozen plasma samples were available; who did not have an inflammatory condition at baseline; and who were not hospitalized, taking antibiotics, smoking, or starting statin therapy during the study were included. C-reactive protein and plasma vitamin levels were measured at baseline and 6 months. RESULTS: At 6 months, C-reactive protein levels were significantly lower in the multivitamin group than in the placebo group (between-group difference = -0.91 mg/L; 95% confidence interval: -1.52 to -0.30; P= 0.005). The reduction in C-reactive protein levels was most evident in patients who had elevated levels (> or =1.0 mg/L) at baseline. Of the six vitamins measured (C, E, B(6), B(12), folate, and beta carotene), only vitamin B(6) (baseline: r = -0.31, P= 0.003; 6 months: r = -0.29, P= 0.006) and vitamin C (baseline: r = -0.25, P= 0.02) were inversely associated with C-reactive protein level. CONCLUSION: In a post hoc analysis of a randomized, double-blind, placebo-controlled study, multivitamin use was associated with lower C-reactive protein levels. Other similarly formulated multivitamins may yield comparable results."

Discussion Questions

1. What dos the first line tell you: Am J Med. 2003 Dec 15;115(9):702-7?

   That this study was published in the American Journal of Medicine on December 15, 2003, in Issue No. 9, Volume 115, at pp. 702-707.

2. What does this line mean: Church TS, Earnest CP, Wood KA, Kampert JB?

   The last names and initials of the four authors.

3. What's the Cooper Institute got to do with this?

   That's where the four authors are located, either working there, or affiliated there for research. And that's where you can reach them.

4. What's the purpose of the article, according to the abstract? Try to put it in your own words.

   High levels of a substance that can be tested for, C-reactive protein, have been associated with heart disease and diabetes. This study examined whether multivitamins could lower the levels of C-reactive protein. (Cite the study. You've just rephrased it. It's not YOUR idea.)

5. Methods and Concepts Drawn from the Study:

   1. post hoc - after the fact. Refers to the fact that analysis was done on patients for whom frozen plasma was already available for the base line, or starting point.

   2. subgroup analysis - Means that the patients were broken up into subgroups. Might have been, for example, subgroups of gender might have been used. You'd need to check the study to see what the subgroups were.

   3. randomized - The placement of the patient in the experimental or the control group was random. And/or patients were randomly selected from the available population of patients. Random means without plan or design, either by random numbers assigned to the whole population, or by choosing every sixth or every tenth patient or some method like that. Then you could place every other patient drawn from the population in either the experimental or control group by putting one in experimental, one in control, one in experimental, one in control. No way that either the experimenter or the subject patient would have any control over which group each subject went into.

   4. double-blind - means that the experimental and control groups were set up so that neither the experimenter or the subject patient knew which group the subject-patient was in.

   5. placebo-controlled - "(2) : an inert or innocuous substance used especially in controlled experiments testing the efficacy of another substance (as a drug)" - from Merriam-Webster Online Dictionary. Often we speak of the use of a "sugar pill," suggesting there's really nothing in it but sugar, but it looks like the pill the experimental group is taking. This means that no one could tell what group the subject-patient was in by looking at the pills he/she took.

   6. 6 month trial - The first test for C-reactive protein was taken from the frozen samples, and the next test was taken 6 months after the frozen samples were tested. The study was called a trial.

   7. n=87 - the sample size was 87. We aren't told in the abstract what the population was, but it seems to have been a population of patients in some facility to which the researchers had access to data.

   8. mean age=53 - the averaage age of the subject-patient included in the sample was 53. Since there is increasing concern for inflammatory diseases, heart attack, and diabetes as the population ages, the mean age seems appropriate.

   9. statistical controls - In order to be included in the sample the patient had to have frozen plasma records available, and "not have an inflammatory condition at baseline; and not be hospitalized, not taking antibiotics, not smoking, or not starting statin (statisns are used to treat high cholesterol, diabetes, etc.) therapy during the study. Each of these controls represents a variable tht might raise the C-reactive protein level, which indicates inflammation. That means that if at the end of the study, the patients had a higer C-reactive protein level it couldn't be attributed ot any of these controlled variables, since people with those problems were not included in the sample.

      Often there are statistical controls on gender. We study only boys, or only girls. Or race. We study only blacks and whites. This makes it easier to isolate variables that are having an effect, but then we can't generalize to those we didn't include in the study. This is one of the problems with medical research. Almost all the studies were done on white men. Good statistical control, but now we don't know if what we found is true for women, for non-whites.

      Other forms of control: experimental and control groups. Since we were looking at how vitamins affect the C-reactive protein level, we measured both the vitamin level and the C-reactive protein level. The experimental group was given the vitamin pill. The control group was not given the vitamin pill. So we would have a hypothesis that the experimental group, which had the vitamin pill, should have a lower level of C-reactive protein at the end of the study than the control group, which had a placebo instead of the vitamin pill. The control group gives a group to compare the results to. It represents the "what if we didn't do anything? or the null hypothesis."

   10. What does the abstract say that the study showed?

       Results: "At 6 months, C-reactive protein levels were significantly lower in the multivitamin group than in the placebo group . . . That means that we reject the null hypothesis, and our study supports out hypothesis that that the experimental group, which had the vitamin pill, should have a lower level of C-reactive protein at the end of the study than the control group, which had a placebo instead of the vitamin pill.

   11. Does this prove that vitamins lower c-reactive protein levels and consequently, inflammation, which can lead to heart problems and diabetes?

       Absolutely not. This is one small study. Could have been done with men alone. Lots of other variables that could affect inflammation in the body were controlled out of the study. And research like this doesn't prove anything. It shows that the inflammation levels registered "significantly lower." That means that we could not have gotten these results if the lower levels were due to random chance. So the vitamins seem to have had an effect. But it would take many studies, with solid controls, to understand the actual role of vitamins in inflammatory diseases.

   12. What does "significant" mean in this case?

       Significant has a very precise meaning in statistics. It means that a statistical test has shown that the results obtained could not have resulted from random chance alone, and that we can be 95% (alpha level) certain of that or 99% (alpha level) certain of that. That means that only 5 times out of 100 would we have gotten these results when vitamins didn't have any effect on the inflammation levels, and it was just random chance. That means we can be 95% sure there is some kind of relationship other than pure chance. But it doesn't begin to tell us what that relationship is, or how strong it is.

   13. What's an alpha level?

       The level of confidence we have in a statistical test that the results we got did not occur by chance alone. Or, put another way, our level of confidence that something different from zero happened. The alpha level is usually reported as 95% or 99%. But remember that it only tells us that we're 95% or 99% sure that vitamins had something to do with the inflammation levels, something different from nothing at all.

   14. between-group difference - this refers to the fact that a test, like the t-test, was used to measure whether there were any statistically significant differences between the levels of the vitamins shown in the experimental and control groups. We are looking here for differences between the group that got the vitamins and the group that got the placebo.

   15. Explain: "The reduction in C-reactive protein levels was most evident in patients who had elevated levels (> or =1.0 mg/L) at baseline."

       Taking the vitamins, instead of the sugar pills, had the greatest effect when the baseline inflammation level, as measured by tests for C-reactive protein, indicated that patients had an elevated level of inflammation. Vitamins didn't show so much effect when there was no indication of raised levels of inflammation at the outset.

       Two vitamins, B(6) and C, were inversely associated with the C-reactive protein level. What does that mean?

       When something is inversely related, as it goes up, the other variable goes down. In this case the vitamins would be the independent variable, the C-reactive protein level the dependent variable. As vitamin B(6) level was higher, the C-reactive protein level was lower. As vitamin C level was higher, the C-reactive protein level was lower. With all the other vitamins, as the vitamin level was higher, the C-reactive protein level was higher: A direct relationship.

   16. Note that even the abstract closes with a conclusion, and note that the conclusion says "multivitamin use was associated with lower C-reactive protein levels," not that multivitamins will lower C-reactive protein levels. To say that multivitamins will lower C-reactive protein levels is to imply causality. To imply causality in medicine in double-blind, experimental trials is costly, time consuming, and difficult. So imagine how hard it is to imply causality in human relationships.

---

---

This work is licensed under a Creative Commons License.
Individual copyrights by other authors may apply.