GRACE principles: recognizing high-quality observational studies of comparative effectiveness

Nonrandomized comparative effectiveness studies contribute to clinical and biologic understanding of treatments by themselves, via subsequent confirmation in a more targeted randomized clinical trial, or through advances in basic science. Although methodological challenges and a lack of accepted principles to assess the quality of nonrandomized studies of comparative effectiveness have limited the practical use of these investigations, even imperfect studies can contribute useful information if they are thoughtfully designed, well conducted, carefully analyzed, and reported in a manner that addresses concerns from skeptical readers and reviewers. The GRACE (Good Research for Comparative Effectiveness) principles have been developed to help healthcare providers, researchers, journal readers, and editors evaluate the quality inherent in observational research studies of comparative effectiveness. The GRACE principles were developed by experienced academic and private sector researchers and were vetted over several years through presentation, critique, and consensus building among outcomes researchers, pharmacoepidemiologists, and other medical scientists and via formal review by the International Society of Pharmacoepidemiology. In contrast to other documents that guide systematic review and reporting, the GRACE principles are high-level concepts about good practice for nonrandomized comparative effectiveness research. The GRACE principles comprise a series of questions to guide evaluation. No scoring system is provided or encouraged, as interpretation of these observational studies requires weighing of all available evidence, tempered by judgment regarding the applicability of the studies to routine care.

Incidence of pelvic inflammatory disease among women treated for gonorrhea or chlamydia

BACKGROUND

No estimate exists of the incidence of pelvic inflammatory disease (PID) after the occurrence of a recent bout of sexually transmitted disease (STD).

METHODS

We used a computerized data file of prescriptions and medical encounters from the Fallon Community Health Plan to estimate the incidence rate of pelvic inflammatory disease (PID) among women with a recently treated episode of gonorrhea or chlamydia (STD). First we identified women with presumed gonorrhea or chlamydia on the basis of a combination of diagnostic codes for cervicitis, vulvovaginitis or Bartholin's abscess, and a computer record of a prescription for doxycycline. We then followed these women to estimate the incidence rate of PID after their treatment for gonorrhea or chlamydia. We estimated the number of cases of pelvic inflammatory disease in this cohort by selecting all women with an International Classification of Diseases (ICD) code of 614.9 entered for either an outpatient or inpatient diagnosis.

RESULTS

We estimated the overall risk of PID to be about 9% during the 1-year period following treatment for gonorrhea or chlamydia, with a steep rise in risk coming within the first 45 days.

CONCLUSION

The risk of PID in the year after an episode of treated STD is high, but the highest period of risk is in the first few weeks. The shape of the risk curve indicates that some PID cases may result from treatment-resistant infections, or possibly from untreated reinfections.

Randomized field trial of vaginal douching, pelvic inflammatory disease and pregnancy

BACKGROUND

Several case-control studies have reported that women who use vaginal douche products are at increased risk for pelvic inflammatory disease. Women who douche regularly may do so for reasons related to their risk of acquiring a sexually transmitted infection, introducing confounding that is difficult to control in non-experimental studies.

METHODS

We conducted a multicenter randomized field trial with a 1-year follow-up period. The study comprised 1827 women age 18-34, with no current indication of pelvic inflammatory disease, who were regular users of a douche product and who had been treated recently for a sexually transmitted bacterial infection or bacterial vaginosis. Women were randomly assigned to use either a newly designed and marketed douche product or a soft cloth towelette, and were resupplied with product at each bimonthly follow-up visit. We measured the occurrence of pelvic inflammatory disease using a combination of clinical and laboratory indicators. We also recorded pregnancy occurrence among participants.

RESULTS

The risk of PID among women assigned to use the douche product, relative to that among women assigned to use the wipe product, was 1.05 (95% confidence interval = 0.57-1.9). Using an alternative, less sensitive definition of PID gave a risk ratio of 1.26 (0.62-2.6). The probability of becoming pregnant was 15% lower among women assigned to use a douche product, and 33% lower among women who douched more frequently (ratio = 0.67; 0.42-1.08).

CONCLUSIONS

There was little or no indication of a greater risk of PID among women assigned to use the douche product. Douching may be related to a lower probability that a woman becomes pregnant.

Methods and objectives of a large US multicenter case-control study of post-transplant lymphoproliferative disorder in renal transplant patients

A large multicenter case-control study is in progress in the United States, the primary goal of which is to provide information about the effects of specific immunosuppressants and other risk factors on posttransplant lymphoproliferative disorder (PTLD) in renal transplant patients. It will also provide incidence data and case characterization on PTLDs arising in a large contemporary population. Medical record data are being collected on up to 120 PTLD cases and up to four controls per case transplanted at 20 large US centers. Participants all received transplants on or after July 1, 1995 and PTLD cases will be identified through December 31, 2001. All cases undergo central clinical and pathologic review. Abstracted information includes detailed data (dosages, duration) on all immunosuppressants (induction, maintenance, anti-rejection) as well as antiviral treatment. Other data include demographics, transplant history, HLA matching and viral status (e.g., Epstein-Barr virus, cytomegalovirus). Information associated with the PTLD diagnosis and initial therapy for PTLD is also collected. To date, 86 potential cases have been reported. Twenty (24%) are pediatric patients (< or =18 years). Median time between transplant and PTLD is 268 days; 53 (62%) were diagnosed within the first year. Cumulative incidence through 1998 is 0.7% for adults and 4.5% for children. The most common single site for PTLD is the allograft. Common treatments included either a reduction or discontinuation of immunosuppression (90%) and antiviral treatment (66%). Overall, the allograft appears to be an important site of PTLD recurrence. Also, the incidence of renal PTLD since the introduction of new immunosuppressive therapies is similar to that reported earlier.

Accessing third-party data for research: trust me? Trust me not?

Epidemiologists are often asked to evaluate product safety or provide information about the history of disease and use of health services. Most of this research is conducted through sponsorship from a governmental or other non-profit agency, or with direct support from a for-profit company with an economic interest in the outcome. Vast amounts of information that are routinely collected for administrative and billing purposes may also be used for research. These data can reveal information about the etiology of disease, utilization patterns of prescription drugs, and trends in disease occurrence. Using third-party data requires a trusting partnership between researchers and data custodians. Although such data offer public health benefits, their use can also lead to embarrassment and legal action. Five guiding principles will help outside contractors facilitate access to third-party data and avoid pitfalls. (1) Understand the sponsor's objectives by understanding the purpose of the research. (2) Identify and approach data resources that have appropriate information. (3) Consider special issues relating to accessing confidential information. (4) Establish terms of the research engagement with the sponsor. (5) Establish ground rules with the data provider.

A combination study design to examine mycophenolate mofetil (MMF) and PTLD in renal transplant patients

Renal transplant recipients who are chronically immunosuppressed by drugs are at a higher risk of developing malignancies. Commonly observed malignancies are several forms of posttransplant lymphoproliferative disorders (PTLD), skin, lip and gynaecological cancers. The risk is associated with many risk factors including the extent of immunosuppression. Mycophenolate mofetil (MMF) is an immunosuppressant, indicated for the prophylaxis of organ rejection in patients receiving allogenic renal and heart transplants. During the European approval of MMF for renal transplantation, the question was raised as to whether the use of MMF was associated with an increased risk of PTLD in comparison with alternate immunosuppressive regimens. In response, F. Hoffman-La Roche Ltd set up a prospective observational cohort study with a companion case-control study. This paper describes the objectives and the methods of these studies along with the rationale of the methodology.

The risk of serious cardiac arrhythmias among cisapride users in the United Kingdom and Canada

PURPOSE

Serious, although rare, ventricular arrhythmias and deaths have been reported in patients taking cisapride monohydrate. Without quantification of the risk involved, it is impossible to develop rational therapeutic guidelines.

SUBJECTS AND METHODS

Arrhythmic events (sudden deaths and other events compatible with serious ventricular arrhythmias) were sought among 36,743 patients prescribed cisapride in the United Kingdom and Saskatchewan, Canada. Prescriptions and cases were identified from computerized medical claims data and physicians' office records. We compared rates of events between periods of recent cisapride use and nonrecent use, using cohort analysis. Potential confounding factors, including concomitant treatment with agents that inhibit CYP3A4 metabolism or that prolong the QT interval, were assessed in a nested case-control study.

RESULTS

In the cohort analysis, the incidence of the arrhythmic events was 1.6 times greater (95% confidence interval [CI]: 0.9 to 2.9) in periods of recent use. With adjustment for clinical history, use of CYP3A4 inhibitors, and use of drugs that prolong the QT interval, the odds ratio for cisapride and cardiac outcomes was 1.0 (95% CI: 0.3 to 3.7). There was no identifiable increase in risk when cisapride was dispensed at about the same time as QT-prolonging drugs or CYP3A4 inhibitors. QT-prolonging agents were associated with a 2.5-fold increase in the risk of arrhythmic events (95% CI: 1.1 to 5.8).

CONCLUSIONS

Serious rhythm disorders were not associated with cisapride use, although the upper confidence bounds do not rule out an increase in risk.

Lymphatic and haematopoietic cancer mortality in a population attending school adjacent to styrene-butadiene facilities, 1963-1993

STUDY OBJECTIVE

To evaluate the risk of mortality from lymphatic and haematopoietic cancers and other causes among students.

DESIGN

The study used school records, yearbooks, and Texas Department of Health records for the school years 1963-64 to 1992-93 to construct a cohort of 15,403 students. Three mortality databases were searched to identify deaths, and mortality rates in the cohort were compared with mortality rates from the United States and Texas. Computed standardised mortality ratios and 95% confidence intervals were used.

SETTING

Eastern Texas high school adjacent to facilities that have been producing synthetic styrene-butadiene since 1943.

MAIN RESULTS

338 deaths were identified. The all causes standardised mortality ratio was 0.84 (95% confidence intervals 0.74, 0.95) for men and 0.89 (0.73, 1.09) for women. The standardised mortality ratio for all lymphatic and haematopoietic cancers was 1.64 (95% confidence intervals 0.85, 2.87) for men and 0.47 (0.06, 1.70) for women. The slight male excess in lymphatic and haematopoietic cancers was stronger among men who attended school for two years or less.

CONCLUSIONS

The overall mortality from lymphatic and haematopoietic cancer among the students was little different from that of the United States as a whole. A moderate excess for men, predominantly among the shorter-term students, was offset by a deficit among women. These variations are compatible with random fluctuations; the overall pattern is not indicative of an effect of environmental exposure sustained while attending the high school.

Resource utilization and cost of care for rheumatoid arthritis and osteoarthritis in a managed care setting: the importance of drug and surgery costs

OBJECTIVE

To describe the frequency and costs of medical services for patients with osteoarthritis (OA) or rheumatoid arthritis (RA) in a managed care setting.

METHODS

Individual utilization records of medical and pharmacy services for OA and RA patients were obtained from a group-model health maintenance organization (HMO). Estimates were made for costs of drugs and medical services for arthritis from July 1, 1993 to June 30, 1994 using Medicare reimbursement schedules and average wholesale drug prices. Calculated rates for each population were expressed as counts of events or as dollars per person-year.

RESULTS

The average individual cost rate of arthritis-related care for 365 RA patients was $2,162 per year, and the total cost of RA care to the HMO was $703,053. Prescription medications accounted for 62% ($436,440) of the total cost of RA care, while ambulatory care accounted for 21% ($150,938), and hospital visits accounted for 16% ($115,674). With regard to 10,101 OA patients, the average individual cost rate was $543 per year, and total cost to the HMO was $4,728,425. Hospital care accounted for 46% ($2,170,890) of the total cost of OA care, medications accounted for 32% ($1,509,637), and ambulatory care accounted for 22% ($1,047,898).

CONCLUSION

RA care, in the setting of this study, was characterized by intensive treatment, especially frequent use of medications that were delivered to most patients. Although the cost of RA care per patient was high, cost to the managed care provider was relatively low, owing to the rarity of RA. OA care tended to be infrequent, and the largest component of cost was hospital care for a small proportion of patients (5%). Owing to the greater prevalence of OA, care of OA was nearly 7 times more costly to the managed care provider than was care of RA.

5-Aminosalicylates, sulfasalazine, steroid use, and complications in patients with ulcerative colitis

OBJECTIVES

The choice between sulfasalazine and 5-aminosalicylate (5-ASA) drugs in the management of patients with ulcerative colitis often depends on idiosyncrasies of drug tolerance and control of the disease in individual patients. We sought to evaluate whether there were population differences in the effect of 5-ASA and sulfasalazine on the occurrence of clinically recognized adverse events. We also attempted to determine whether there were differences in the use of concomitant steroids and in the rates of hospitalization.

METHODS

We reviewed a large computerized database drawn from general practices in the United Kingdom. There we found records of 2894 patients in whom general practitioners had diagnosed ulcerative colitis, and who were receiving ongoing medical therapy specific to ulcerative colitis. The period of data availability ran from the beginning of 1990 to the latter part of 1993. The average duration of observation was 2.1 yr per patient. Patient histories were categorized into distinct periods according to the dose of 5-ASAs and sulfasalazine, steroids, and immunosuppressants, and were further separated according to the activity of ulcerative colitis. Within these categories, we examined the initiation and discontinuation of steroids, incidence of new hospitalizations for ulcerative colitis, and clinical mention of adverse events.

RESULTS

New clinical mentions of hepatic, pancreatic, renal, and hematological events other than anemia were similar among the 5-ASAs and were very infrequent overall. Hospitalizations for ulcerative colitis occurred with similar frequency (about 15 hospitalizations per 100 patients per year) among users of those drugs. Patients receiving sulfasalazine had lower rates of initiation of prednisolone than did patients receiving 5-ASA, but sulfasalazine was used proportionately less often in patients who had been recently hospitalized, and it may be that sulfasalazine patients were somewhat less sick, overall, than were 5-ASA-using patients. The choice of drug did not affect discontinuation rates for prednisolone among established users.

CONCLUSIONS

In the United Kingdom, during the period of this study, serious adverse reactions to drugs were not an important aspect of the management of patients with ulcerative colitis. Renal and pancreatic complications of sulfasalazine and 5-ASA therapy were extremely rare. Sulfasalazine and 5-ASA drugs have similar steroid-sparing properties. Disease-specific hospitalizations are approximately 100 times more common in ulcerative colitis patients than are serious adverse drug effects. Considerations of drug efficacy should therefore dominate the choice between therapeutic agents.

Characterisation of asthma management in the Fallon Community Health Plan from 1988 to 1991

In order to characterise asthma management in a managed care setting, we identified 10,301 patients who were diagnosed with asthma between 1 January 1988 and 31 December 1991 at a group model health maintenance organisation in central Massachusetts, US. We obtained for these patients automated utilisation files containing data on medications, hospitalisations, emergency room visits, office visits, and estimated costs of these services. The medication dispensed to the greatest proportion of patients was beta 2 agonists either by inhalation (56%) or orally (21%). Theophylline was dispensed to 23% of the patients. Maintenance therapy was inhaled anti-inflammatory medication was uncommon, as inhaled corticosteroids (17%) and sodium cromoglycate (cromolyn sodium) [8%] were dispensed to fewer patients than other asthma medications. Among patients who had been hospitalised in the previous year, 36% were presently receiving inhaled corticosteroids, and among patients who used at least one beta 2 agonist metered-dose inhaler per month, 49% were presently receiving inhaled corticosteroids. Economic analyses showed that only 8% of the patients had either a hospital admission or an emergency room visit, but hospital costs among these patients accounted for 25% of the total costs of asthma care. In addition, the top 10% most expensive patients accounted for 42% of the total cost of asthma care. We conclude that a substantial proportion of patients at increased risk of a severe attack, by virtue of having a recent hospitalisation, do not receive maintenance anti-inflammatory therapy, and that hospitalisations among a relatively small proportion of asthma patients contribute significantly to the cost of asthma care.

A cost density analysis of benign prostatic hyperplasia

We assessed the frequency and cost of care for benign prostatic hyperplasia (BPH) among approximately 165,000 subscribers to Fallon Community Health Plan (FCHP), a group model health maintenance organization located in central Massachusetts. We computed rates of episodes of medical services for BPH using automated utilization files, and we estimated costs using Medicare reimbursement schedules and medication average wholesale prices. We identified 3919 men who visited a physician for BPH from January 1, 1991, until December 31, 1994, during which time they contributed 8336 person-years to the analysis. This population comprises approximately 12% of men at least 40 years old at FCHP. From 1991 to 1994, 696 (18%) men received terazosin, 219 (6%) men underwent a prostatectomy, and 41 (1%) men received finasteride. Men averaged 1.66 office visits per year to a physician for BPH. Most office visits (61%) were to a primary care physician, with 39% of the visits to a urologist. Among patients who received terazosin, the frequency of office visits increased slightly after receiving terazosin, from 2.14 to 2.62 visits per year. Among surgery patients, the frequency of visits declined after prostatectomy, from 6.31 visits per year to 1.67 visits. The individual annual cost rate for BPH care ranged from $25.00 to $25,352.00, with an average of $364.00 per person and a median cost of $126.00. The major components of the overall costs were hospital admissions (35%), terazosin dispensings (29%), and physician office visits (19%), with outpatient hospital care and ambulatory procedures accounting for the remaining 17%. Among men receiving terazosin, the average cost was $1190.00 per person-year, and among patients undergoing prostatectomy, the cost was $2630.00 per person-year. The prostatectomy rate declined by nearly 80% during the study period, while the dispensing rate for terazosin doubled, resulting in an overall decline in the total cost of care for BPH from 1991 to 1994.

Utility of telephone company records for epidemiologic studies of cellular telephones

We conducted a survey of over 5,000 telephone users who were customers of one large cellular telephone company covering four major geographical areas. Our primary goal was to assess the utility of ascertaining information on telephone use and type from telephone company records. We compared information from 3,949 respondents with corresponding data from company billing records. We found that 48% of the account holders were sole users, and 69% were the primary user, meaning that they accounted for at least 75% of the use. Respondent reports of amount of telephone use were highly correlated with data on the billing record (r = 0.74). Respondent reports of telephone type were similarly correlated with data from the manufacturer (r = 0.92). We also inquired about telephone holding patterns, since these have implications for exposure. Most users reported favoring one side of the head when using the telephone, but the side of the head used was not strongly associated with handedness.

Overall mortality of cellular telephone customers

Unlike mobile cellular telephones, in which the antenna is not part of the handset, a portable cellular telephone exposes the user's head to radio frequency energy transmitted from the antenna. This exposure has prompted concerns about potential biological effects, including brain cancer. As a first step in a record-based mortality surveillance of cellular telephone customers, we report on overall mortality of a cohort of more than 250,000 portable and mobile telephone customers during 1994. We found age-specific rates to be similar for users of the two types of telephones. For customers with accounts at least 3 years old, the ratio of mortality rates in 1994 for portable telephone users, compared with mobile telephone users, was 0.86 (90% confidence interval = 0.47-1.53).

Patient factors associated with strut fracture in Björk-Shiley 60 degrees convexo-concave heart valves

BACKGROUND

Previously established predictors of outlet strut fracture in Björk-Shiley convexo-concave (CC) valves include larger valve size, larger opening angle (70 degrees versus 60 degrees), younger age at implant, and date of manufacture. We sought to identify patient characteristics that might be predictive of strut fracture and to refine the estimates associated with previously identified predictors.

METHODS AND RESULTS

We conducted a case-control study of CC60 degrees valves implanted in the United States and Canada and manufactured between January 1, 1979, and March 31, 1984. Cases included all valves with verified outlet strut fractures reported to the manufacturer from January 1979 through January 1992. Up to 10 controls were selected for each case. Control valves were matched according to implanting surgeon and were required to have been functioning at least as long as their matched case valves. Case and control medical records were reviewed for information on patient medical history before the valve implant. There were 96 case and 634 control valves for which clinical data were available. Patient age and valve size and implant position were confirmed as important determinants of fracture. There was a strong inverse gradient of risk with age. The risk of fracture was 42% lower for each 10-year increment of patient age at time of implant. Large mitral valves were at greatest risk of strut fracture, with the largest mitral valves (33 mm) estimated to be 33 times more likely to fracture than the smallest (21 to 25 mm) aortic valves. Date of manufacture was also associated with risk; valves welded from mid-1981 through March 1984 were more likely to fracture than those manufactured in 1979 and 1980. Body surface area < 1.5 m2 was associated with 1/16 the risk of body surface area > or = 2.0 m2. No other patient factor was strongly associated with the risk of strut fracture.

CONCLUSIONS

Few patient features identifiable in the implant record are predictive of strut fracture. Our analysis supports previous work in identifying valve size, patient age, and date of manufacture as predictors of fracture and adds body surface area. A number of these associations suggest that conditions associated with higher cardiac output may also place patients at increased risk.

Manufacturing characteristics associated with strut fracture in Björk-Shiley 60 degrees Convexo-Concave heart valves

Björk-Shiley Convexo-Concave (CC) valves sometimes experience fracture of the outlet strut. Previously implicated valve characteristics that predict strut fracture include larger valve size, larger opening angle (70 degrees vs 60 degrees), remilling, weld date, and implant in the mitral position. While the associations between risk, size, and opening angle suggest that part of the elevated incidence of strut fracture might be due to the design of the Björk-Shiley valves, only a small fraction of implanted valves have experienced strut fracture. In consequence, previously unexamined variations in the manufacturing process have been suggested as possible factors affecting the failure risk of individual valves; materials, manufacturing steps, quality control, and specific workers have all been put forward as potential explanations for valve-to-valve variation in risk. We conducted a case-control study of CC60 degrees valves implanted in the USA and Canada and manufactured between January 1, 1979 and March 31, 1984. Cases included all verified strut fractures reported to the manufacturer from 1979 through January, 1992. up to 10 controls were selected for each case. Controls were matched to cases on implanting surgeon and were required to have been implanted and functioning at least as long as their respective case valves. We reviewed case and control manufacturing records. There were 150 cases and 1095 surgeon-matched controls. Large mitral valves were at greatest risk of strut fracture; 33mm mitral valves were estimated to be 23 times more likely to fracture than 21-25mm aortic valves. Valves welded in 1979 and 1980 were less likely to fracture than those welded in any other time period; however, no specific manufacturing procedures or personnel were uniquely associated with this time period. Valves with more flexible outlet struts, as determined by the hook deflection and load deflection tests during manufacture, appear to have been at higher risk than valves with more rigid outlet struts. There were three welders who had worked on a sufficient number of valves to allow separate estimation of the risk in the valves they welded. One welder's work was associated with about one-third the risk of valves worked on by the other two. Examination of receiver operating characteristic curves revealed, however, that welder identity added little to the discriminating information already available in the form of valve size and implant position. It is concluded that welder identity and strut flexibility appear to contribute to the risk of outlet strut fracture in Björk-Shiley CC60 degrees valves. Neither of these factors, however, is sufficient to account for much of the previously unexplained variation in risk. No other characteristic measurable in existing manufacturing records appears to predict risk of strut fracture in any useful way.

Peptic ulcer and gastrointestinal hemorrhage associated with nonsteroidal anti-inflammatory drug use in patients younger than 65 years. A large health maintenance organization cohort study

BACKGROUND

Nonsteroidal anti-inflammatory drugs (NSAIDs) are associated with an elevated risk of peptic ulcer and upper gastrointestinal hemorrhage, but published reports have lacked information on rates of outpatient disease, have concentrated on the elderly, and have not provided comparisons of rates for specific types of NSAIDs.

METHODS

We compared incidence rates of peptic ulcer and upper gastrointestinal hemorrhage in 68 028 people younger than 65 years who used diclofenac sodium, naproxen, piroxicam, or sulindac, and who were members of a network of health maintenance organizations. We reviewed automated insurance claims data and medical records to ascertain cases and included conditions treated on an outpatient basis.

RESULTS

Medical claims data were adequate for crude identification of potential cases, but review of medical records led to rejection of 63% of these, representing either no abnormality or diseases other than peptic ulcer or upper gastrointestinal hemorrhage. Of the total 112 cases, 64 (57%) were treated as outpatients. The crude incidence rate per 1000 person-years in users of any current, recent, or past NSAID was 2.2 and in distant-past users of NSAIDs was 0.75. For diclofenac, naproxen, piroxicam, and sulindac, we found a consistent pattern of decreasing NSAID effects from current to recent to past exposure. The risk of peptic ulcer or upper gastrointestinal hemorrhage was 1.6 cases per 1000 people using NSAIDs.

CONCLUSIONS

Combining use of automated claims records with review of medical records promotes efficiency while maintaining specificity of case ascertainment. This study, with 57% of cases treated as outpatients, had results consistent with other published reports that were based on hospitalized patients. Within the limits of statistical error, the incidence rates of peptic ulcer and upper gastrointestinal hemorrhage appeared to be similar for the various NSAIDs studied.

Mortality among synthetic fiber workers exposed to glycerol polyglycidyl ether

We studied mortality among 8,878 employees who worked at any time from 1965 to 1988 at a synthetic fibers plant in North Carolina that used a finishing agent containing glycerol polyglycidyl ether. Some glycidyl ethers are mutagenic and tumorigenic in laboratory animals. The main route of exposure to workers was inhalation of the spray mist, although there was also skin contact. We identified 553 deaths in the cohort and the standardized mortality ratio (SMR) from all causes of death combined was 0.80. For most causes of death, mortality rates in the cohort were similar to mortality rates in the U.S. population. Among categories with at least five observed deaths, the largest effect estimate was for cancer of the central nervous system (SMR = 1.77), and the SMR for lung cancer was 0.94. The cancer categories of central nervous system (brain) and "other" lymphopoietic cancers (lymphoma and myeloma) showed weak associations with duration of employment. In case-control analyses in which we utilized work history data to compute effect estimates by duration of exposure, we found no increased risk of lung cancer or brain cancer among employees with more than 5 years of exposure. Effect estimates for lymphoma and myeloma tended to increase with duration of exposure, although there were only seven deaths in this category and the effect estimates were very imprecise. To date, this study has identified no clear carcinogenic effect of glycerol polyglycidyl ether, but plausible induction periods have not yet elapsed. The cohort should continue to be monitored to obtain more precise estimates after moderate or long induction times.

Kidney cancer and hydrocarbon exposures among petroleum refinery workers

To evaluate the hypothesis of increased kidney cancer risk after exposure to hydrocarbons, especially those present in gasoline, we conducted a case-control study in a cohort of approximately 100,000 male refinery workers from five petroleum companies. A review of 18,323 death certificates identified 102 kidney cancer cases, to each of whom four controls were matched by refinery location and decade of birth. Work histories, containing an average of 15.7 job assignments per subject, were found for 98% of the cases and 94% of the controls. To each job, industrial hygienists assigned semiquantitative ratings for the intensity and frequency of exposures to three hydrocarbon categories: nonaromatic liquid gasoline distillates, aromatic hydrocarbons, and the more volatile hydrocarbons. Ratings of "present" or "absent" were assigned for seven additional exposures: higher boiling hydrocarbons, polynuclear aromatic hydrocarbons, asbestos, chlorinated solvents, ionizing radiation, and lead. Each exposure had either no association or a weak association with kidney cancer. For the hydrocarbon category of principal a priori interest, the nonaromatic liquid gasoline distillates, the estimated relative risk (RR) for any exposure above refinery background was 1.0 (95% confidence interval [CI] 0.5-1.9). Analyses of cumulative exposures and of exposures in varying time periods before kidney cancer occurrence also produced null or near-null results. In an analysis of the longest job held by each subject (average duration 9.2 years or 40% of the refinery work history), three groups appeared to be at increased risk: laborers (RR = 1.9, 95% CI 1.0-3.9); workers in receipt, storage, and movements (RR = 2.5, 95% CI 0.9-6.6); and unit cleaners (RR = 2.3, 95% CI 0.5-9.9).

Mortality update of cellulose fiber production workers

Mortality ascertainment was extended through 1990 for a cohort of 1271 workers involved in the production of cellulose triacetate fiber at a plant in Rock Hill, South Carolina. Each subject was employed for at least three months between 1 January 1954 and 1 January 1977 in jobs that entailed exposure to the highest concentrations of methylene chloride. Median exposures in 1977 ranged from 140 to 745 ppm (8-h time-weighted average). The observed numbers of deaths from specific causes were compared with the expected numbers of deaths computed from rates in York County, South Carolina. For most causes of death, there was little if any association with employment. Among causes of particular interest, no new deaths were observed from cancer of the liver and biliary tract, although the excess from the earlier study persisted (4 observed, 1.34 expected). No excess mortality was observed for cancer of the pancreas (2 observed, 2.42 expected) or for ischemic heart disease (43 observed, 47.8 expected).

Determinants of serious liver disease among patients receiving low-dose methotrexate for rheumatoid arthritis

OBJECTIVE

To assess the risk of serious liver disease in patients with rheumatoid arthritis (RA) taking methotrexate (MTX).

METHODS

We surveyed members of the American College of Rheumatology to determine previous use of MTX in the treatment of rheumatoid arthritis and to identify cases of cirrhosis and liver failure. Cases were confirmed by review of pathology specimens, findings from diagnostic testing, and clinical presentations. A case-control study was then conducted to ascertain prognostic factors. Case and control medical records were reviewed for information on MTX therapy as well as other possible determinants of serious liver disease.

RESULTS

Twenty-four cases of cirrhosis and liver failure were identified, giving a 5-year cumulative incidence of approximately 1/1,000 treated patients. Six of the 24 patients had died: 4 died of the initial liver disease, 1 of hepatic complications of another illness, and 1 of unrelated causes. Two patients continue to have active liver disease. Late age at first use of MTX and duration of therapy with MTX were independent predictors of serious liver disease.

CONCLUSION

Serious liver disease is an uncommon, age- and dose-related complication of low-dose MTX therapy for RA.

Depressive symptoms and headaches in relation to proximity of residence to an alternating-current transmission line right-of-way

Electric power transmission lines have become objects of public controversy. Hypotheses have linked neurobehavioral effects to the electric and magnetic fields that these lines produce. The authors conducted a telephone interview survey in November 1987 to assess the prevalence of depressive symptoms and headache in relation to proximity of residence to an alternating-current transmission line in the United States. Proximity to the line, defined as residing on a property abutting the right-of-way or being able to see the towers from one's house or yard, was positively associated with a measure of depressive symptoms. The association was not explained by demographic variables associated with depression or by attitudes about power lines or other environmental issues. The estimated prevalence odds ratio was 2.8 (95% confidence interval (CI) 1.6-5.1). The estimate did not change appreciably when the definitions of depressive symptoms or of proximity to the line were altered. Nonmigraine headaches had a weaker association with proximity to the line (odds ratio = 1.5, 95% CI 0.76-2.8), and self-reported migraine headaches exhibited no association (odds ratio = 0.99, 95% CI 0.29-3.4). Additional studies of psychological and behavioral measures should be conducted in relation to electric and magnetic fields, with a strong emphasis on improved exposure assessment.

Leukemia among nuclear workers with protracted exposure to low-dose ionizing radiation

We conducted an analysis of leukemia mortality and protracted exposure to low-dose ionizing radiation. We used data from seven published epidemiologic studies of nuclear workers that reported individual monitoring information for radiation exposure. We found 83 leukemia deaths among white males from a combined total of more than 1.4 million person-years. We calculated an overall relative risk for leukemia of 1.5, after adjustment for age and calendar time, for workers with cumulative occupational doses of 10 mSv (1 rem) or greater, compared with those with cumulative doses of less than 10 mSv. An adjusted relative risk of 1.8 was observed when individuals with 10-50 mSv (1-5 rem) were compared with those who had cumulative occupational doses of less than 10 mSv. The adjusted relative risk for those with doses greater than 50 mSv relative to those with doses less than 10 mSv was 1.2, although the data were especially sparse in this dose range. These combined data indicate a small elevated risk of leukemia for doses of ionizing radiation under 50 mSv.

Mortality of cellulose fiber production workers

Mortality was studied among 1271 employees of a cellulose fiber production plant in Rock Hill, South Carolina, in the United States. Each subject was employed for at least three months between 1 January 1954 and 1 January 1977 in jobs that entailed exposure to the highest concentrations of methylene chloride. In the cohort 122 deaths were identified through 1 September 1986, and mortality rates for the cohort were compared with mortality rates for York County, South Carolina. Deficit mortality was observed for cancers of the respiratory system, breast, and pancreas and from ischemic heart disease. Excess mortality was observed for cancers of the buccal cavity and pharynx and the liver and biliary tract, and for melanoma as well. The largest relative excess was for liver and biliary tract cancers. There were only four deaths in this category; however, three of the four deaths were cancer of the biliary tract (3 observed, 0.15 expected, standardized mortality ratio 20).

Bromocriptine and puerperal seizures

Case reports have prompted concern that the use of bromocriptine mesylate to prevent lactation in the puerperium increases the risk of postpartum seizure. We conducted a record-based case-control study of postpartum seizures in three data bases to evaluate this relation. We identified 43 women who had a postpartum seizure, and we matched 319 controls individually by hospital of delivery, quinquenium of age, and time of delivery. Overall, women taking bromocriptine had a 22% lower risk for seizures, that is, the relative risk estimate was 0.78, with a 90% confidence interval of 0.29 to 1.87. A reduction in seizure risk is consistent with reports of antiseizure activity for bromocriptine in various species, including humans. We found a small positive association between bromocriptine use and seizures occurring more than 72 hours after delivery, with a relative risk estimate of 1.6 after controlling for seizure history. This association was offset by a strong negative association between bromocriptine use and early-occurring seizures. The pattern of an initial reduced risk followed by an increase to normal or above-normal levels of risk could result from an antiseizure activity of bromocriptine, with a rebound in risk when bromocriptine is withdrawn.

Toxic shock syndrome, contraceptive methods, and vaginitis

To elucidate further the etiology of toxic shock syndrome, we assessed the effects of certain contraceptive methods and recent history of vaginal infection on the incidence of toxic shock syndrome, with confounding effects of other risk factors controlled. We found a strong but imprecise positive association between toxic shock syndrome and tubal ligation (rate ratio = 7.9, 90% confidence interval 1.4 to 42.7). We also observed a negative association with oral contraceptives (rate ratio = 0.49, 90% confidence interval 0.22 to 1.1) and a positive association with a recent history of vaginitis (rate ratio = 2.1, 90% confidence interval 1.2 to 3.9).

Projections of asbestos-related disease 1980-2009

Approximately 19,000 cases of mesothelioma and 55,000 cases of lung cancer will arise in U.S. men with histories of nontrivial occupational exposure to asbestos. There are approximately 65,000 U.S. men now alive with clinically diagnosable asbestosis. These estimates are based, in the case of the cancer, on estimates of the effective number of asbestos-exposed workers required to produce the current national incidence of mesothelioma. The asbestosis estimates are based on a number of rough measures relating the prevalence of asbestosis to the incidence of mesothelioma, the incidence of compensable asbestosis in other countries, the prevalence of and mortality from pneumoconioses generally, and the number of workers heavily exposed to asbestos.

Identifying the health risks from very low-dose sparsely ionizing radiation

The health risks from low-dose sparsely ionizing (low-LET) radiation have been the subject of continued debate. At present, quantitative estimates of risk are extremely uncertain due to the controversy surrounding both the dosimetry for A-bomb survivor data and the choice of mathematical models for extrapolating risk from high to low doses. Nevertheless, much can be learned about the nature of the health risks by reviewing the epidemiologic literature. We present a summary of diseases which have been associated with low-LET radiation (less than 1000 rad) in at least two independent studies, according to the mean cumulative organ dose at which the disease was observed. At organ doses of less than or equal to 50 rad, the only diseases that have been reported consistently are thyroid cancer, salivary gland tumors, and leukemia. The first two diseases were observed in association with x-ray epilation of the scalp for tinea capitis, a therapy which is no longer employed. On the other hand, leukemia has been observed repeatedly to occur at cumulative doses of greater than or equal to 30 rad low-LET radiation.

An independent analysis of the National Cancer Institute study on non-nutritive sweeteners and bladder cancer

We evaluated the possible relation between use of non-nutritive sweeteners and bladder cancer using data obtained from the National Cancer Institute Bladder Cancer Study under the Freedom of Information Act. In the general study group, there was no evidence for an association between non-nutritive sweeteners and bladder cancer. Control for a variety of factors through multivariate techniques diminished the plausibility of earlier interpretations of these data, which had raised the possibility that certain subgroups of users or non-nutritive sweeteners might be at an increased risk for bladder cancer. We found that the putative effects of non-nutritive sweeteners were not consistent among subgroups with similar baseline risk, did not display consistent dose-response trends, and were subject to considerable sampling error. We concluded that the data provided little evidence that non-nutritive sweeteners increase risk for bladder cancer among subgroups of users, and that definitive evidence on this question is beyond the reach of conventional research.

Choosing populations to study the health effects of low-dose ionizing radiation

In January 1978, the United States Congress requested information about the utility of additional epidemiologic studies for quantifying the health effects of low-dose ionizing radiation. In our judgment, no single population can be recommended for study on purely scientific grounds, since the largest group offers only a small chance to obtain a definitive result. On the other hand, if social pressures and regulatory agencies mandate that such studies be attempted, we would recommend prospective cohort studies of occupational populations. We propose that a national worker registry be developed using ionizing radiation as the prototype for studying other occupational exposures. The problems related to studying low-level radiation are not unique, but apply equally to investigations dealing with a great variety of toxic agents. A national plan for collecting information on workers' exposure and health could provide a cost-efficient means to answer public health questions posed by the Congress, scientists and the public.

Fibrinolytic response and oral contraceptive associated thromboembolism

A case-control study of fibrinolytic activity was conducted comparing 12 women with a recent history of thromboembolism while taking oral contraceptives and 28 matched female controls without a history of thromboembolism. All subjects had stopped using oral contraceptives at least 12 months prior to study. A new assay, recently developed in this laboratory, was used to evaluate the mean fibrinolytic response to venous occlusion in both cases and controls. The fibrinolytic response of all subjects was stratified into quartiles as previously described. While the controls segregated as expected, all 12 cases occurred in subjects whose fibrinolytic response fell in the first two quartiles with 8 of the 12 subjects having first quartile responses. The mean fibrinolytic response for all controls was 12.3 units while cases showed a mean response of only 3.9 units. The above data supports and extends our recent suggestion that low fibrinolytic response may signal a natural predisposition to venous thromboembolism which could be triggered by use of synthetic estrogens. However, these data do not support the use of this assay as a screening test for oral contraceptive related risk to thromboembolism since at least 50% of the control population also segregate in the first two quarters.

Blood coagulation and idiopathic thromboembolism among fertile women

A case-control study of fibrinolytic activity and antithrombin-III (AT-III) was conducted comparing 16 women within the childbearing ages who had a recent history of idiopathic thromboembolism to 29 age- and race-matched female controls without a history of thromboembolism. All subjects had stopped using oral contraceptives at least 12 months prior to study. No difference in AT-III was evident between cases and controls. The mean fibrinolytic response to venous occlusion in cases was significantly lower; however, baseline fibrinolysis was essentially identical. These data demonstrated a bimodal distribution of fibrinolytic response among controls. By dichotomizing fibrinolytic response between the two modes in the controls, we estimated that the rate of low fibrinolytic response was 6 times as frequent in cases as controls. Thus, low fibrinolytic response may signal a natural predisposition to thromboembolism which could be triggered by use of synthetic estrogens.