| Site Index |
|
|
         |
|
••Fall 1998 / Vol 2, No 4 |
|
|
|
|
|
|
|
Clinical Contributions The Gary Friedman Symposium | to
pdf >> On May 11, 1998, a well-attended affair was held at the University of California in Berkeley, entitled: Gary Friedman at the DOR: 30 Years (and Counting)A Celebration of a Career. The triggering event for the celebration was the end of Gary Friedman's seven-year term as Director of the Kaiser Permanente Medical Care Program Division of Research (DOR) in Oakland. The program included six Scientific Presentations, each of which emphasized Gary's role as an initiator of scientific research and as a mentor of developing researchers. They also serve as an admittedly incomplete catalog of the epidemiologic oeuvre of the DOR over the past few decades; for this reason and because of scientific interest in the presentations, it was decided that publication of manuscripts based upon these talks in The Permanente Journal would be of value. Five of the six presenters (including Gary) have submitted the manuscripts published here. Based upon somewhat informal talks, they should not be considered comprehensive, fully referenced reviews of the topics. Some anecdotal material relevant to the occasion was left in the manuscripts by several of the presenters. This deserved tribute to one of Kaiser Permanente's (KP's) most distinguished physicians (see biography) should interest and give pride to the Journal's readers. We hope it will also encourage others to submit original research articles to The Permanente Journal. The work of Gary and his colleagues is an excellent example of the important discoveries that can accrue when clinicians analyze their practices in a rigorous fashion and share their findings with others. --Arthur L. Klatsky, MD, and
Illegal, Immoral, or Bad for the Heart? Introduction First Alcohol-Coronary Heart Disease (CHD) Study Hundreds of items (history questions, health measurements, lab tests) were screened. Gary called the study a fishing expedition, but I think the hypothesis was that new predictors would be found. This hypothesis was amply fulfilled--Gary's very productive study1 resulted in published articles about lung function,2 psychological traits,3 coffee,4 medical history questions,5 and the first report of the leukocyte count as a heart attack predictor.6 As we surveyed the data, an inverse relationship of coronary risk to alcohol drinking habits was one of the more striking findings. Gary very generously suggested that I write the first report of prospective data suggesting that nondrinkers were apparently at higher CHD risk. Thus, an article appeared in the Annals of Internal Medicine in 1974, reporting a significantly lower myocardial infarction (MI) risk among alcohol drinkers, compared to abstainers.7 There were 464 patients with a first MI. All had a prior MHC exam. The computer selected a "risk control" group well matched to each case for demographics and seven established coronary risk factors; and an "ordinary control" group matched only for age, race, and sex. Thus it was possible to ascertain whether a predictor was associated with or independent of the established risk factors. The important smoking-drinking interaction was recognized. In retrospect, lack of control for smoking was probably a major reason why prior studies failed to recognize this inverse alcohol/MI relationship. Results stratified by smoking habit were graphically presented (Fig. 1). Since this was the first published epidemiologic report of this finding and no mechanism was apparent, interpretation was cautious. Nine possible (mostly spurious) explanations were discussed, only one of which was a protective effect of alcohol. Since then dozens of epidemiologic studies, including further Kaiser Permanente studies, have almost unanimously found that lighter drinkers are at lower coronary heart disease (CHD) risk than abstainers (both lifelong and former drinkers). Plausible mechanisms have been demonstrated, most notably that alcohol raises the coronary protective high-density lipoprotein cholesterol (HDL). Antithrombotic effects of alcohol may also be involved. Most scientists now accept this as a probable causal relationship. Benefit of lighter drinking for persons at coronary risk has found its way into U.S. and U.K. governmental advice to the public. Current debate is largely focused on possible additional benefits in specific beverages (wine, beer, liquor), and how best to advise the public. Due to extensive lay media reporting, much of the public is aware of this relationship. A 1995 publication commemorating the 25th anniversary of the National Institute of Alcohol and Alcohol Abuse8 cited the 1974 publication as one of 16 "seminal" articles in alcohol research. This got us started in alcohol-health research, and led to a 20-year series of grants from the Medical Advisory Board of the Brewers Association of America and later from the Alcoholic Beverage Medical Research Foundation. These resulted in many reports. Over the years major roles were fulfilled by Abraham Siegelaub, Mary Anne Armstrong, and Harald Kipp. Alcohol-Hypertension Association This was one of the first analyses of the alcohol-blood pressure relationship and remains one of the largest. By now dozens of cross-sectional and prospective epidemiologic studies, including additional Kaiser Permanente studies, have solidly established an empirical alcohol-hypertension link. These are supported by intervention studies showing blood pressure changes in several days to weeks with drinking or abstinence, with no convincing evidence that confounding is responsible. A mechanism has not been established although some data suggest a centrally mediated sympathetic nervous system action. The fact that explanations remain speculative is the only major deficiency in the case for causality. Public Affairs Aspects
As Gary and I were flying to Atlantic City for the first public presentation of the alcohol-CHD data, we had a drink with dinner. Gary said to me, "we may have to be careful about drinking in public after our presentation, because we'll be accused of wanting to justify our bad habits." Indeed we have been attacked over the years--not for our personal habits, but for possible bias related to our funding. The obvious harmful effects of alcohol make it difficult for many to accept the concept of possible benefit from lighter drinking. The attacks were bothersome, but Gary--as always--was a calming influence. The funding sources of the early studies became useful in rebuttal of suggestions of possible bias in our work. The apparent beneficial effect of lighter drinking in CHD was uncovered by work supported by the NIH, while the apparent harmful effect on blood pressure resulted from work indirectly supported by beverage industry funds. Of far greater importance in--I think--out success in maintenance of a reputation for honest reporting has been Gary's richly deserved reputation as in investigator of absolutely impeccable integrity. Alcohol-Mortality Relationship Alcohol Data In a 1978-1985 Cohort The long range objective was study of relationships of drinking behavior to a variety of health measurements and outcomes. Prospective alcohol related reports included analyses of pancreatic,13 breast,14 prostate,15 and large-bowel cancer;16 total hospitalization days,17 changes in drinking,18 liver cirrhosis,19 cerebrovascular disease,20 supraventricular arrhythmias,21 total mortality,22 unnatural deaths,23 and, of course CHD.24-27 New cross-sectional analyses of the alcohol-hypertension relationship28 and the alcohol-smoking association29 were also published. The first CHD study based upon the new data set was a prospective analysis
of 756 hospitalized persons.24 This multivariate analysis
showed that past drinkers and very infrequent drinkers (< once per
month) had CHD risk similar to that of lifelong abstainers. Lower risk
was The "Sick-Quitter" Hypothesis The Role of Beverage Choice Little is new under the sun. Actually, the first presentation pertinent to the "French Paradox" was in 1819 by Dr. Samuel Black,34 an Irish physician with a great interest in angina pectoris and of considerable perception with respect to epidemiologic aspects. His delightfully Francophile explanation of the disparity in CHD between Ireland and France, attributed the low French angina prevalence to "the French habits and modes of living, coinciding with the benignity of their climate and the peculiar character of their moral affections." The recent studies, plus the discovery of antioxidant phenolic compounds in red wine, have led many to believe that alcohol is protective against CHD, but that red wine is more protective than other alcoholic beverages. Prospective population studies, on the other hand, show no consensus, and suggest that each beverage type is protective.35 This controversy led us to study this aspect. We first examined CHD mortality by studying risk for persons who took almost all of their alcohol in the form of one beverage type.26 These preference groups represented only a fraction of the CHD deaths, and there were major user differences between the groups. All beverage types were protective, but wine and beer preferrers were at lower risk than liquor preferrers--this difference was statistically significant only for wine preferrers. There was no difference in risk between persons who took red wine and those who took other types of wine. We recently reported CHD hospitalization data with a much larger group of cases.27 As in all previous Kaiser Permanente studies, total alcohol intake was inversely related to CHD risk (Table 1). By assigning a proxy variable to use of wine, liquor or beer, we were able to utilize all beverage choice data. Again, each beverage type was protective, but beer use appeared most protective in men and wine (both red and white) in women (Tables 2 and 3). We concluded that the major benefit was from alcohol and that user differences and drinking patterns were most likely to be responsible for the beverage type disparities. I think that this view has become the prevailing current assessment of many workers in the field, but, as Yogi Berra's much quoted wisdom goes, "it isn't over until it's over." Conclusion References
The CARDIA Study and the Development of Clinical Research
at the Division of Research One of Gary Friedman's legacies to the Division of Research (DOR) was the development of a clinical research unit. Prior to 1984, the DOR (then known as Medical Methods Research) had been involved in few externally funded studies involving the clinical examination of research subjects, the most notable of which was a study of female twins, funded by NHLBI and headed by Gary. In December 1983, that changed with the funding of the Coronary Artery Risk Development in Young Adults (CARDIA) Study. Gary was the principal investigator and led the writing of a successful application for Kaiser Permanente (KP) to be one of the four field centers for this study. CARDIA was initially funded for a period of five years with a baseline and a two-year followup examination and with the primary aims: 1) to determine the distribution of coronary heart disease risk factors in a biracial cohort of men and women aged 18-30 years at entry; and 2) to identify habits and behaviors that are associated with both initial levels and later changes in these risk factors. Our initial plan was to conduct the clinical examinations for this study in the multiphasic health checkup area of the Oakland KP Medical Center, but we soon realized that this facility would not be adequate. The problemsolving process led us to develop research clinic space in the old DOR building on Piedmont Avenue. In February 1985, clinic equipment and supplies began to arrive at the DOR in preparation for the baseline exam startup scheduled to take place in the late spring--a low-temperature freezer, centrifuge, pulmonary function and exercise treadmill testing equipment, needles, syringes, food models, etc. It was both an exciting and chaotic time for us. At the baseline examination conducted for a year in 1985-86, 5115 participants were recruited into the 4 clinical centers (Birmingham, AL; Chicago, IL; Minneapolis, MN; and Oakland, CA) with 1426 recruited at Oakland, making it the largest center. The participants were 18-30 years of age, with relatively equal numbers of African Americans and whites, men and women, ages 18-24 and 25-30 years, and education <12 years and 12 years. The success of the study (high quality data and 91% cohort retention at the two-year follow-up) led to continued funding. Five examination cycles have been completed (baseline and 2-, 5-, 7-, and 10-year follow-up exams), and a 15-year follow-up examination is planned for the year 2000-01. Retention has remained very high for this relatively young and mobile cohort with nearly 80% of the cohort returning for the 10-year follow-up exam. The longitudinal study of cardiovascular risk factors in the CARDIA Study group has yielded very interesting findings. A sampling of these findings is shown in Figs. 1 through 4. Fig. 1 shows the mean cumulative weight change over time, one of the more dramatic findings from the CARDIA Study. The mean 10-year weight change ranges from a 15.7-pound increase in white women to a 25.4-pound increase in African American women. Fig. 2 suggests one of the reasons for this marked weight gain. Physical activity score, reflecting the frequency and intensity of participation in 13 different kinds of activity during the past year, decreased substantially and fairly consistently over 10 years with a mean decrease of about 20% in men and 25% in women. Fig. 3 demonstrates relatively small 10-year changes in systolic blood pressure (SBP). All race/gender groups show a decrease in SBP between the baseline exam and the two-year follow-up, a finding common to most observational epidemiologic studies which is probably related to acclimatization of the participant to the research environment. Mean SBP at the 10-year follow-up exceeds the baseline SBP only for African American women, is nearly equal for African American men, and has not yet returned to baseline for white study participants. Fig. 4 demonstrates divergent trends in mean low-density lipoprotein (LDL), the "bad" cholesterol, in men and women. LDL cholesterol has trended upward in men and downward in women since the baseline examination. The observation that LDL cholesterol is actually decreasing in women (and not increasing dramatically in men) in spite of large weight increases might seem surprising to some. One of the strengths of the CARDIA Study is that the longitudinal measurement of many factors that influence cardiovascular risk, and the aging of the younger part of the study cohort has allowed us to understand to a great degree how this perplexing situation has occurred. CARDIA is the first study to document directly the impact that secular trends are having on individual changes in LDL cholesterol level with age. Two concepts of group change need to be understood:
In the CARDIA Study, Bild et al. examined seven-year secular trends by comparing older participants at the baseline CARDIA examination (ages 25-30 year, N=2788) with younger CARDIA participants who aged into the 25-30 year group at the seven-year follow-up exam (ie, those who were 18-23 years old at the baseline exam in 1985-86 became 25-30 years old at the seven-year follow-up exam in 1992-93, N=1395).1 Comparison of these two groups (those who were 25-30 years-old at baseline and those who were 25-30 years old at the seven-year follow-up) provides information on the secular influences that have caused changes in the characteristics of 25-30-year-olds at these two points in time. Table 1 shows the secular percentage changes in LDL-cholesterol, body weight, and Keys score. The Keys score predicts cholesterol change based on changes in dietary fat and cholesterol and was measured from dietary data collected at both points in time. We see that a secular decrease in LDL cholesterol occurred during this time period, ranging from 5.2% to 8.9%, in the setting of a secular increase in body weight (ranging from 5.3% to 6.9%) and a secular decrease in Keys score (ranging from 9.5% to 15.2%). We conclude that the secular decrease in LDL cholesterol offsets the expected increase in LDL cholesterol with age and increasing weight and that this decrease was probably due to secular dietary changes that lowered lipids. As the CARDIA Study prepares for a 15-year follow-up exam, the participants are still too young to experience many incident events of atherosclerotic cardiovascular disease. However, it has only recently become clear that the presence of subclinical cardiovascular disease can be quantified noninvasively in a substantial proportion of the population in the age spectrum represented by the CARDIA Study population. For example, coronary artery calcium can be quantified quickly and noninvasively by electron beam computed tomography (EBCT) and has been found to correspond to the quantity of coronary plaques and the probability of significant disease as assessed by angiography. The Oakland and Chicago centers participated in a CARDIA EBCT substudy that was conducted during 10-year follow-up exams. In the 443 CARDIA Study participants, coronary artery calcium was detected in 40% of African American men, 46% of white men, 37% of African American women, and 17% of white women.2 The high prevalence of subclinical CVD in this population of young adults is not surprising, reflecting the fact that atherosclerosis is a lifelong process that is highly prevalent in this country. Plans for the 15-year follow-up exam include the assessment of atherosclerosis in the carotid artery of all study participants using ultrasound imaging. Gary remained the principal investigator of the CARDIA Study until he was awarded an Outstanding Investigator Grant from the National Cancer Institute in 1991. During his tenure as the Principal Investigator, he wrote two papers based on the CARDIA data. The first was a descriptive paper about the design of the CARDIA Study with a report of the recruitment experience for the baseline examination and some of the sociodemographic and risk factor characteristics of the study cohort.3 The second paper has a story to go along with it because it represents
part of a longer story about the pioneering aspect of Gary's work, of
his being the "discoverer" of epidemiologic findings that
other researchers later replicated and verified. In this case, it is
the story of the leukocyte count (white blood cell count) and coronary
artery disease. In 1974, The New England Journal of Medicine
published a paper by Gary, Art Klatsky, and Abe Siegelaub showing that
leukocyte count predicted the development of first acute myocardial
infarction.4 In an accompanying editorial, Dr. Henry Blackburn,
a noted cardiovascular epidemiologist, expressed the importance of search
for new indicators of coronary risk and said about Gary's paper, "The
Kaiser group has gone fishing and caught an interesting specimen, the
leukocyte count."5 Gary's second CARDIA paper, published
in 1990, was an analysis of association of the leukocyte with sociodemographic
and cardiovascular risk factors, and represented his continued interest
in the relationship of the leukocyte to cardiovascular risk.6
Why would the leukocyte count be associated with coronary heart disease
(after accounting for The seed sown by Gary in the start up of the CARDIA study has grown into the implementation of many clinical research studies at the DOR, and the development of a large clinical research facility. These studies account for about 20 to 30 percent of the annual DOR research budget. The growth of clinical research activity has been particularly rapid during Gary's tenure as director, and is one the many legacies of his research career at the DOR. References
"Sensitive" and "Specific" Epidemiologic
Studies: The Division of Research of the Kaiser Permanente Medical Care
Program As critics of epidemiologic research are all too eager to point out,1,2 this type of research can lack both "specificity"it can erroneously observe an association--and "sensitivity"it can fail to identify an association that truly is present. There are a number of strategies available to epidemiologists to increase the specificity and sensitivity of their investigations, the use of which generally makes the results they obtain more valuable. Nearly all of these have been employed commonly by Gary Friedman and his colleagues at the Division of Research (DOR), which is one of the reasons their work is valued so highly. Minimization of Measurement Error Exposure Subclassification Disease Subclassification In their study of gastric cancer in relation to prior infection with Helicobacter pylori,8 DOR investigators acknowledged this possibility by subdividing cases according to anatomic location within the stomach. No association with H.pylori infection was observed for tumors of the gastroesophogeal junction, in contrast to a 3.6-fold increased risk of other gastric cancers. Had they not disaggregated the case group, an overall relative risk would have been obtained that would not have reflected the size of the association with either cancer subsite. Conclusion References
Screening for Colorectal Cancer: Research Contributions
of The Permanente Medical Group Screening for colorectal cancer has been a topic of exceptional interest to clinician researchers in The Permanente Medical Group since the early 1950s. Rigid sigmoidoscopy was employed in at least two medical centers at that time1 and became a part (albeit an optional part) of the multiphasic health checkup from its earliest days. The work of a number of investigators has added to and, in several cases, led development of public policy on screening for this common malignancy. Gary Friedman had a hand in many of these studies, as we shall see. The Multiphasic Evaluation Study After 16 years of follow-up, mortality from potentially preventable causes was 30 percent lower in the study group, and approximately half of this reduction was due to fewer deaths from colorectal cancer (Table 1).3 Because colorectal cancer was not the focal point of the Multiphasic Evaluation Study, Friedman et al dedicated a relatively modest amount of space in their report to discussing details of the finding. They suggested that a specific trial of this question was needed to determine whether sigmoidoscopy indeed reduced mortality. Despite these cautions, the Multiphasic Evaluation Study was embraced in the early 1980s by advocates of "evidence-based guidelines" as randomized trial evidence that sigmoidoscopy can lower mortality from colorectal cancer.4,5 The American Cancer Society, among others, used the study to endorse periodic (every three years) sigmoidoscopy.5 Concerned that the full story may not have been told, Dr. Friedman and I began a further set of analyses of the Multiphasic Evaluation Study in 1985.6 We reasoned that if sigmoidoscopy were responsible for the lowered mortality from colorectal cancer, the mortality reduction should principally be for cancers of the rectum or distal colon (ie, cancers within reach of the sigmoidoscope). We further suggested that sigmoidoscopy could lower mortality by either or both of two mechanisms. It could lower incidence of cancer (by detection and removal of adenomatous polyps), or it could improve the stage distribution of cancers. Finally, because the sigmoidoscopic examination was an optional part of the multiphasic, dependent upon scheduling by the patient and occurring at a separate later examination, we wanted to assure ourselves that there really was an excess of screening sigmoidoscopy in the study group. After reviewing records of all incident cases of colorectal cancer and detailed chart abstraction forms from the entire cohort, we observed the following: 1) the mortality reduction in the study group was somewhat greater for cancers within reach of the sigmoidoscope than for cancers above reach of the sigmoidoscope (Table 2). Two-thirds of the reduction in mortality from distal colorectal cancers was due to lowered incidence; the remaining third appeared to be due to earlier detection with an improved stage distribution and lower case fatality rate. However, there was no difference between groups in removal of adenomas during the follow-up period. Thus, the incidence reduction can not comfortably be attributed to sigmoidoscopy. In further analyses, only a small fraction of the cancers were detected by screening in either group and the study group's stage shift improvement was as great for cancers detected after onset of symptoms as for screen-detected cancers. Thus, it does not appear that screening sigmoidoscopy accounted for much of the stage shift either. When we tallied the total number of sigmoidoscopies in each group, the study group had only a slight excess of sigmoidoscopies and this was contributed by a very small group of individuals (N=212 excess persons) who heeded the telephone urgings meticulously and had, on average, 3.3 sigmoidoscopic examinations in the ten-year period. This small excess of screening would be expected to prevent less than one fatal cancer. In summarizing, we could say only that the Multiphasic Evaluation Study had not been designed or sized to test the efficacy of sigmoidoscopy and its results should not be used as evidence either for or against sigmoidoscopic screening. The US Preventive Services Task Force A Case-control Study of Screening Sigmoidoscopy Funded by the National Cancer Institute, the study was undertaken in 1989. The case definition was fatal colorectal cancer arising within reach of the rigid sigmoidoscope. Controls were drawn from the health plan membership, were matched on age and sex to cases, and had to outlive the case and not die of colorectal cancer. A total of 261 eligible cases were identified. These were matched to 868 controls. A large deficit of exposure to screening sigmoidoscopy was observed among cases during the ten years preceding diagnosis of the fatal cancer. This yielded an odds ratio of 0.30, suggesting a 70% reduction in risk of fatal colorectal cancer associated with screening. After adjustment for a number of possible confounders, including personal and family history of colorectal cancer or polyps, exposure to other screening tests, and history of periodic health checkups, the odds ratio increased slightly to 0.41, still suggesting nearly a 60% risk reduction. To determine whether this association might have been due to additional unmeasured self-selection factors, we studied an additional group of 268 patients who died of colorectal cancers arising above reach of the sigmoidoscope and a similar number of matched controls. For these cancers, sigmoidoscopy should offer no benefit. Any observed benefit would most likely be due to confounding by self-selection factors that made persons undergoing screening sigmoidoscopy less likely to develop or die from colorectal cancer. However, after adjusting for the same set of measured confounders, there was no residual benefit in this group (Table 3). The striking difference in benefit between cancers within reach and those above the reach of the sigmoidoscope strengthened the inference that sigmoidoscopy was truly beneficial. Another important finding of the study was the observation that a person's risk for developing fatal colorectal cancer remained extremely low for at least ten years following a negative sigmoidoscopy. This observation fit with what is known of the biology of colorectal cancer (ie, that cancer develops slowly from adenomatous polyps over a period of many years). However, it clashed with recommendations of the time that sigmoidoscopy should be performed every three years. A longer interval between screenings would make a sigmoidoscopy screening less costly as well as more acceptable to patients. The findings from this study were published in The New England Journal of Medicine in 199210 and reviewed by the US Preventive Services Task Force in 1994. On the basis of this study, the Task Force recommendation for sigmoidoscopy was upgraded from "insufficient evidence" (Grade C) to "fair evidence" (Grade B), meaning that the Task Force now supported screening with sigmoidoscopy. On the strength of this change, coverage for screening by sigmoidoscopy has now been added by HCFA for Medicare recipients and by many private insurers. The CoCaP Program The CoCaP program (Colon Cancer Prevention), 11 begun in January 1994, was the first sigmoidoscopy screening program in the country designed to serve an entire population. During its first four years, CoCaP screened approximately 60,000 members annually. A recent survey of Kaiser Permanente members found that more than 50 percent of age-eligible members now report having had a sigmoidoscopy within the past 5-10 years. This figure is much higher than levels reported nationally. Data from the CoCaP screening examinations, along with pathology reports and follow-up colonoscopies, have been collected at DOR and entered into a large database for continued study of questions related to screening with sigmoidoscopy. Using this database, we have recently shown convincingly that neither the presence nor the size of tubular adenomas found at screening sigmoidoscopy predicts a greater likelihood of finding important adenomas in the proximal colon at follow-up colonoscopy. Rather, it is the presence of villous features in adenomas of any size that increases this risk. We are also monitoring cancer incidence rates in the entire Kaiser Permanente membership. Given the degree of effectiveness shown in the case-control study, the CoCaP screening effort should eventually show up as a decline in the incidence of advanced colorectal cancers from the distal half of the colon and rectum. Just how long it will take to demonstrate this effect is uncertain. With complete cancer incidence data through 1996 (after three years of CoCaP), we do not yet see a significant decline in age-, sex-adjusted incidence of advanced stage cancers. Studies of Alternative Screening Strategies (Fecal
Occult Blood Testing) Summary References
Reflections Summary One of the main reasons that KP in Northern California is one of the best settings in the world to conduct epidemiologic and health services research is our access to comprehensive, often long-term, medical records on millions of people. Although our newer computerized records are very valuable, our collection of manual charts going back over 50 years is a national treasure and must be preserved despite the storage and retrieval costs entailed. We must also guard against external threats to our access to our records for legitimate research resulting from overzealous protection of privacy. Other precious resources for research, such as the Kaiser-Orentreich frozen serum collection and tissue specimens in our pathology departments, also require care and preservation. While recognizing the value of all of our records we must be cautious about errors in data retrieved from either computer storage or medical charts. Investigator-initiated, outside-funded research published in peer-reviewed journals must always be a primary activity of DOR. The keys to our success, both past and future, lie in our resources, objectivity, quality work completed and published, and helpfulness to others in KP. Thanks and Acknowledgments My Assistant Directors, Drs. Robert Hiatt and Joe Selby, have contributed much to the leadership of DOR, in part by covering for my inadequacies and making me look better than I was. I have been greatly stimulated by collaborating investigators in DOR and KP, of whom three of today's speakers, Arthur Klatsky, Joe Selby and Stephen Sidney are prime examples. Young investigators have also been stimulating; mentoring them and watching them grow to be independent has been one of the most satisfying aspects of my career. I have received much help from DOR biostatisticians, originally Abe Siegelaub and Hans Ury and more recently the five excellent colleagues that we now have. (It is not well known that biostatisticians are ordained clerics, empowered to pronounce one's analyses and statistical tests as kosher.) I have received indispensable computer programming by Donna Wells and Harald Kipp for many years. Similarly, I have relied on our excellent medical record analysts--most recently, Merril Jackson, May Kuwatani and Bill Frank--and superb secretarial assistance from Agnes Lewis, Stephanie Tang, Pat Crump, Lyn Wender, Susan Mignano and Mary Jeaniene Luck. DOR has had excellent administration, led by P.H. Kidd, Carolyn Quan, and Donna O'Connor. Though Donna's staff has grown along with the department, it is still lean, hardworking, and productive. Fortunately, there has been relatively little bureaucracy in DOR. I am grateful to our outside collaborators who have brought good ideas and expertise to DOR; two of the speakers today, Julie Parsonnet and Noel Weiss are excellent examples, as is Leonard Syme here at the University of California. Finally, DOR and I owe a great deal to the overall KP organization. Although research is one of its important goals, it is not its main mission. KP has a wealth of talent that can participate in our research efforts and generally employees throughout KP have been willing to go the extra mile to help us. Changes in Epidemiologic Methods and My Use of them
During My Career In 1971, Art Klatsky, Abe Siegelaub, and I began a study to search
for as-yet-undiscovered risk factors for myocardial infarction and sudden
cardiac death. (This effort led to some important discoveries, especially
the negative correlation between alcohol use and risk and the positive
association of the leukocyte count and risk.)3,4 Because
I neither understood nor trusted multivariate analysis methods such
as multiple logistic analysis (having had almost no exposure to them
in my biostatistics training at Harvard) I developed an elaborate scheme
to account for the known standard coronary risk factors. We selected
our subjects from the large number who had extensive computer-stored
multiphasic health checkups (MHCs) and matched cases and controls on
these risk factors, made possible by the huge number of MHC examinees.
Continuing my conservatism, multivariate analysis was barely mentioned
in the first edition (1974) of my textbook, Primer of Epidemiology.
Then in 1978, led by Loring Dales, we published a case-control study
of diet and colon cancer in blacks.5 Loring did introduce
us to the use of multivariate analysis, in this case the method of Alvan
Feinstein called prognostic stratification. Finally, I was converted,
and we used logistic regression in a study of cigarette smoking and
mortality6 published in 1979 and the Cox proportional hazards
model in a study of the effects of quitting smoking on mortality in
1981.7 In 1987, the third edition of The Importance of Preserving our Medical Records,
A National Treasure This collection began over 50 years ago and, obviously, many old records are no longer needed for the care of patients. The costs of storage of these records and retrieval for research are considerable and periodically questioned. Every few years I have been called upon to defend these old records from destruction and have zealously done so because I view them as a national treasure. The question has been raised again because of the merger of KP in Northern and Southern California into one division. The storage issue is especially urgent in Northern California because our long-term storage facility in Livermore will run out of space by the end of this year. KP in Southern California has more space available partly due, unfortunately, to their policy of destroying records of subscribers after they have left the Health Plan for seven years (except records of children and cancer patients). Why I regard these records as a national treasure is well illustrated by one of the most influential studies ever done in DOR, the well-controlled study by Joe Selby and colleagues that showed that screening sigmoidoscopy prevents death from colorectal cancer.8 Because the ascertainment of cancer cases dated from 1971 and the review of the sigmoidoscopy experience of cases and controls went back ten years before the cases' diagnoses, we needed to include records of care in the early 1960s. To my knowledge, no one ever predicted that those records would prove to be so valuable. We have looked at even earlier records in assessing the effects on children of their parents' exposures or characteristics or the effects of exposures or characteristics early in life on health and disease many years later. Lisa Herrinton is now abstracting records going back to the 1940s for her study of adolescent height and weight in relation to risk of breast cancer in our female subscribers. There is no guarantee that our old records will prove so valuable in the future, but I think that the likelihood is great. I know of no comparable medical record collection in a defined population in which long-term medical follow-up, or follow-back, is so feasible. The Mayo Clinic has a wonderful system, which goes back further--to the early 1900s,8 but the population of Rochester, Minnesota in which the Mayo group can do population-based studies is much smaller than ours and lacks the tremendous ethnic and socioeconomic diversity of our subscribers. I have been meeting with persons in charge of medical record storage in KP's California Division, and they understand how important our records are, especially in view of KP's increasing promotion of research as a way to demonstrate its leadership in providing health care. We have been exploring ways to lower the cost of storing the little-used, older records and to support their preservation and retrieval through research-funding mechanisms. External Threats to Research Using our Medical Records Fortunately, recent proposals for legislation in California and at the federal level have been more reasonable, protecting confidentiality but allowing legitimate medical and public health research to go forward under the watchful eyes of Institutional Review Boards. Other Priceless Resources Some Cautions About our Valuable Computer-stored
Records I urge my colleagues to make good use of our computer-stored records. At the same time, recognize their limitations and maintain a healthy skepticism about what you find in them. Usually, a more full story can be found in the chart. In a postmarketing study of the antibiotic, clindamycin,11 we found two cases of diarrhea during three months of follow-up in the computer records. In the manual charts, we found ten. There were probably several reasons for the discrepancy, one of the most important being that the drug was prescribed frequently by otolaryngologists and there was no provision on their clinic diagnosis form for them to easily record the occurrence of diarrhea in their patients. In a study of the possible carcinogenic effects of lindane,12 applied topically for scabies or pediculosis, four of the skin cancer diagnoses turned out to be Kaposi's sarcoma in AIDS patients, which was not apparent in the computer records. Lifestyle factors could readily connect these infestations with AIDS. Another apparent lindane user who developed breast cancer had used her KP identification card to obtain the prescription for her husband. These errors were sufficient to reduce the apparent excess risk of cancer among lindane users to statistical insignificance. So be careful. If you cannot review all the manual records of study subjects, at least review the critical ones or a random sample of all of the records for validation. Chart Review is Subject to Error, Too. Be Especially Careful of Rare Findings or Events,
Whatever the Data Source My experience in searching for patients with rare diseases or findings in various data bases has confirmed this concern about accuracy. When you find only a few such patients, check their records carefully. You will often discover that about half are people with other conditions that have been miscoded. Investigator-initiated Research in our Setting Such a change would be most unfortunate,15 as is well illustrated by Joe Selby's study of sigmoidoscopy.16 Previously, Joe had helped us on the US Preventive Services Task Force (USPTF) to review the literature on whether screening by sigmoidoscopy prevents death from colorectal cancer. He and I concluded17 that there were no good, well-controlled studies that demonstrated the efficacy of this procedure, even though it was being recommended by prestigious authoritative bodies such as the American Cancer Society. The Task Force agreed, and its recommendation concerning the clinical application of screening sigmoidoscopy was neutral.18 Clearly, this procedure could detect cancer early and lead to the removal of premalignant polyps; but since it was unpleasant and costly, more evidence concerning its efficacy was needed. Joe recognized that KP was an excellent setting for a case-control study of this question. Our records could reveal who died of colorectal cancer--the case subjects, and screening sigmoidoscopy had been offered at some of our medical centers and recorded in our records. He designed a case-control study in which the case subjects were compared with control subjects, the main focus being on their experience of sigmoidoscopy in the ten years before the cases' diagnoses. Motivated by his personal interest and recognition of the need for such a study, Joe applied for a National Cancer Institute (NCI) grant. Peer review led to the application initially being turned down because Joe and I had not considered a methodologic problem known as healthy-screenee bias. We consulted with Noel Weiss, a leading expert in case-control studies of screening tests. As a result, an improved study design was resubmitted, and the study was approved and funded by NCI. The study was completed and showed elegantly that screening sigmoidoscopies do prevent death from colorectal cancers within their reach. After additional peer review, the study was published in the prestigious New England Journal of Medicine16 and received considerable attention both nationally and within KP. Confirmed in a smaller investigation elsewhere,19 our findings led the USPTF to change its recommendation regarding screening sigmoidoscopy from neutral to favorable.20 The KP gastroenterologists were also impressed, and a systematic screening program for our subscribers, known as Colorectal Cancer Prevention or CoCaP, was implemented. This study was investigator-initiated, outside-funded, and published. What would have happened if DOR had been totally devoted to directed in-house research (assuming that someone of Joe Selby's caliber would have been willing to work under such an arrangement)? First, the grant application was submitted in 1987, and the study was conducted from 1988 to 1991 at a cost of $457,000. I sincerely doubt that management at that time would have viewed this question of sufficient concern to devote almost a half million dollars to it. Second, with total internal support, we would probably not have benefited from external peer review of the study design, and the methodologic flaw would not have been detected. If anyone in KP had been aware of healthy-screenee bias, it should have been Joe and I, and we were not. Finally, if the study had not been published in a respected peer-reviewed journal, the findings would have been less impressive to our physicians and managers. If our leadership had instituted the CoCaP program on the basis of a purely internal unpublished study, it would probably have been met with skepticism and resistance, given the extra work, costs, and difficulties involved. This example clearly shows several advantages of peer-reviewed published research and why it should always be a big part of our mission. This does not mean that DOR should become an ivory tower lacking organizational concerns. We must respect and respond to the interests and needs of KP. Not only would we not exist without its support, but we in DOR believe in KP, its principles and its social mission. We want to contribute to KP's success in providing high quality care to its subscribers. Just as our work benefits the organization, the interaction with KP leaders and clinicians stimulates us to identify and investigate interesting questions. The Keys to DOR's Success, Past and Future The keys to DOR's success, both past and future, are listed.
References
Gary D. Friedman, MD, MS: A Biography Gary Friedman is an internationally known, outstanding physician-epidemiologist who has been with the Kaiser Permanente Medical Care Program (KPMCP) since 1968. Born in Cleveland, Ohio on March 8, 1934, he attended Antioch College and the University of Chicago (BS 1956) and received his MD at the University of Chicago in 1959. His internship and residency were served at the Harvard Medical Services, Boston City Hospital and at University Hospitals in Cleveland, Ohio. This was followed by obtaining an MS degree (biostatistics) from the Harvard School of Public Health in 1965 and American Board of Internal Medicine certification. He worked with the Framingham Heart Study (1962-66) and as Chief of the Epidemiology Unit, Heart Disease Control Program, U.S.P.H.S. in San Francisco before joining the KPMCP. At the KPMCP Division of Research (DOR), he served as Senior Epidemiologist (1968-76), Assistant Director (1976-91), Director (1991-98), and is now Senior Investigator. Gary is substantially responsible for developing one of the first and largest epidemiological research programs in an HMO. His prodigious output of work has been in diverse areas of epidemiology, including cancer, cardiovascular disease, gallbladder disease, adverse drug reactions, twin research, effects of smoking and alcohol, and evaluation of screening tests. His textbook, Primer of Epidemiology, now in its Fourth Edition, is widely used by medical schools and health professionals (>80,000 copies sold) and has been translated into Spanish, Italian, and Chinese. He has authored more than 250 scientific papers or book chapters. He is one of the few epidemiologists to win a coveted seven-year, $4-million Outstanding Investigator Grant from the National Cancer Institute (1989-96; renewed from 1994-2001). Among Gary's epidemiologic "firsts" are the following:
His academic affiliations are: Gary is a devoted husband (married Ruth Schleien on 06/22/58), father (Emily, Justin, and Rick), and grandfather (Sophie and Nathaniel). His diverse interests include music and hiking. He has been a regular runner since age 27 and now runs 15-20 miles per week (bicycles regularly another day). Gifted with perfect pitch, he has played the piano since childhood and took up the oboe at age 54. He has played the oboe in numerous chamber music groups (was on Board of Directors, Chamber Musicians of Northern California) and is a performing member of the San Francisco Civic Symphony, the University of California San Francisco Orchestra, and the Bohemian Club Band. Selected Publications By Gary D. Friedman, MD, MS
(of over 260) Papers:
|
|
|
|
|
|
|
|
|
|
|
 |
|
