Diet, reproductive factors and lung cancer risk among Chinese women in Singapore: evidence for a protective effect of soy in nonsmokers

The factors associated with risk of lung cancer among nonsmokers have not been fully elucidated, but dietary factors have consistently been shown to play a role. Chinese women are unique in having a high incidence of lung cancer despite a low smoking prevalence. This population is also known to have a high intake of soy, a dietary source of phytoestrogens. We conducted a hospital-based case-control study among Singapore Chinese women, comprising 303 cases and 765 age-matched controls, of whom 176 cases and 663 controls were lifetime nonsmokers. Data on demographic background, reproductive factors and dietary intake of fruit, vegetables and soy foods were obtained by in-person interview. We observed an inverse association between intake of total, cruciferous and non-cruciferous vegetables and risk of lung cancer among smokers. Although smokers in the highest tertile of fruit intake also had a lower risk, this was not statistically significant. Higher intake of soy foods significantly reduced risk of lung cancer among lifetime nonsmokers, but not among smokers. When soy isoflavonoid intake in mg/week was computed based on frequency and portion size of intake of eight common local soy foods, the adjusted OR among nonsmokers for the highest tertile compared to the lowest was 0.56, 95% CI 0.37-0.85 (p for trend <0.01). Fruit intake was also significantly associated with reduced lung cancer risk among nonsmokers, but the effect was not significant after adjustment for soy intake. On the other hand, soy intake remained an independent predictor of risk after controlling for fruit intake. Reproductive effects were also primarily confined to lifetime nonsmokers, among whom having 3 or more livebirths (adjusted OR 0.65, 0.44-0.96) and a menstrual cycle length of more than 30 days (OR 0.46, 0.25-0.84) accorded a significantly reduced risk of lung cancer. Place of birth was significantly associated with risk among nonsmokers (OR 2.6, 1.7-3.9 for China-born vs. local born) but not among smokers. When analysis was restricted to nonsmokers with adenocarcinomas, the dietary effects were consistent or enhanced. On stepwise regression, soy intake and cycle length emerged as the independent dietary and reproductive predictors of lung cancer risk in nonsmokers. These findings are consistent with other evidence suggesting an involvement of estrogen-related pathways in lung cancer among non-smoking women.

Risk factors and incident coronary heart disease in Chinese, Malay and Asian Indian males: the Singapore Cardiovascular Cohort Study

OBJECTIVE

This prospective study in Singapore investigated the relationships of established coronary risk factors with incident coronary heart disease (CHD) for Chinese, Malay, and Asian Indian males.

SUBJECTS

A cohort (consisting of 2879 males without diagnosed CHD) derived from three previous cross-sectional surveys.

METHODS

Individual baseline data were linked to registry databases to obtain the first event of CHD. Hazard ratios (HR) or relative risks for risk factors were calculated using Cox's proportional hazards model with adjustment for age and ethnic group and adjustment for age, ethnic group and all other risk factors (overall adjusted).

RESULTS

There were 24,986 person-years of follow-up. The overall adjusted HR with 95% CI are presented here. Asian Indians were at greatest risk of CHD, compared to Chinese (3.0; 2.0-4.8) and Malays (3.4; 1.9-3.3). Individuals with hypertension (2.4; 1.6-2.7) or diabetes (1.7; 1.1-2.7) showed a higher risk of CHD. High low density lipoprotein cholesterol (LDL-C) (1.5; 1.0-2.1), high fasting triglyceride (1.5; 0.9-2.6) and low high density lipoprotein cholesterol (HDL-C) (1.3; 0.9-2.0) showed a lesser but still increased risk. Alcohol intake was protective with non-drinkers having an increased risk of CHD (1.8; 1.0-3.3). Obesity (body mass index > or =30) showed an increased risk (1.8; 0.6-5.4). An increased risk of CHD was found in cigarette smokers of > or =20 pack years (1.5; 0.9-2.5) but not with lesser amounts.

CONCLUSIONS

The increased susceptibility of Asian Indian males to CHD has been confirmed in a longitudinal study. All of the examined established risk factors for CHD were found to play important but varying roles in the ethnic groups in Singapore.

Population-based cancer survival in Singapore, 1968 to 1992: an overview

The Singapore Cancer Registry has provided comprehensive population-based incidence data since 1968. This paper describes the population-based survival analysis of the registry data. All invasive primary cancers diagnosed from January 1, 1968 to December 31, 1992 were passively followed up until December 31, 1997. Only 5.8% were lost to follow-up. Cumulative and observed survival rates were calculated using Hakulinen's method. Overall 5-year relative survival rates have increased dramatically over the 25-year period in both genders. Significant increases are seen with nasopharynx, stomach and colo-rectum cancers, non-Hodgkin's lymphoma, leukemias and cancers of the testis, cervix, ovaries and breast. When compared with the Surveillance, Epidemiology and End Results (SEER) rates in the United States, the 5-year relative survival rates in Singapore are generally lower. However, the rate of change between the two countries is fairly similar. On the average, the rates are 10 to 15 years behind the SEER rates and 5 to 10 years behind Finland, Switzerland and Japan, but they are close to the UK rates. The age-standardized 5-year survival rate for Singapore is higher for most sites compared with other developing countries like Qidong (China), Madras (India), Bombay (India) and Chiang Mai (Thailand). The 25-year trend in cancer survival in Singapore showed two extreme groups: those showing no change and those showing significant improvements. Reducing the incidence of cancers belonging to the first group remains the only viable mode of cancer control. For cancers in the second group, improvement in survival is due to a combination of successful early detection measures and effective treatment services in Singapore.

Academic achievement, close up work parameters, and myopia in Singapore military conscripts

AIM

To determine the relation of refractive error to environmental factors, including close up work, in Singapore military conscripts.

METHODS

A cross sectional study was conducted on 429 Singapore military conscripts. Non-cycloplegic refraction and A-scan biometry were performed in both eyes. A detailed questionnaire was administered by in-person interview to obtain information about current and past near work activity, extra tuition lessons, educational experiences, and family demographics.

RESULTS

Myopia associated with the conscript having been educated in the (gifted, special, or express) educational streams (adjusted odds ratio (OR) = 3.8, 95% confidence interval CI 2.0-7.3), and having completed pre-university education (OR=4.1, 95% CI 1.9-8.8). The reported close up work activity at age 7 years did correlate with age of onset of myopia (p<0.001). In parallel, supplemental tuition lessons in primary school has (OR=2.6, 95% CI 1.4-4.9) associated with conscript myopia. Parental myopia was positively associated with myopia (p<0.001), but this relation disappeared when adjusted for environmental factors. Current (p=0.83) and past close up work activity at age 7 years (p=0.13) did not correlate with myopia.

CONCLUSION

Educational level and educational stream positively related to myopia. A relation was observed with reported close up work activity in early childhood and with tuition classes during elementary school, but not with current close up work activity. These results underscore the strong influence of environment in myopia pathogenesis but a role for close up work activity remains indeterminate.

Does education explain ethnic differences in myopia prevalence? A population-based study of young adult males in Singapore

PURPOSE

To study interethnic variation in myopia prevalence and severity in young adult males in Singapore and to determine whether these variations are related to differences in education level.

METHODS

A population-based survey of refractive errors in a cohort of 15,095 military conscripts between July 1996 and June 1997 using noncycloplegic autorefraction and a standard questionnaire. Prevalence rates of myopia (<-0.5 D) and severe myopia (<-6.0 D) were determined for Chinese, Malay, and Indian men; prevalence rate ratios were compared after adjusting for education level.

RESULTS

Singapore has one of the highest prevalences of myopia (79.3%) and severe myopia (13.1%), with Chinese having higher rates (82.2%, 95% confidence interval 81.5, 82.9) compared with Indians (68.7%, 95% confidence interval 65.1, 67.1) and Malays (65.0%, 95% confidence interval 62.9, 67.1). Education was strongly associated with prevalence and severity of myopia. However, significant interethnic variation persisted after adjusting for education.

CONCLUSIONS

There is a high prevalence of myopia in Singapore. Although prevalence and severity of myopia were strongly associated with education, interethnic variation observed was not fully explained by differences in education level.

Myopia and night lighting in children in Singapore

AIM

To examine the role of night time lighting and myopia in children in Singapore

METHODS

A cross sectional study was conducted on 1001 children in two Singapore schools. Cycloplegic refraction and A-scan biometry measurements were made in both eyes. A detailed questionnaire was completed by the parents to obtain information on night time lighting, near work activity, educational and demographic factors.

RESULTS

There was no difference in myopia prevalence rates in children exposed to night time light (33.1%) compared with children who slept in the dark (31.4%) before age 2. In addition, vitreous chamber depth was not related to night light (p = 0.58) before age 2. These results remained even after controlling for near work.

CONCLUSION

Myopia is not associated with night light in Asian populations.

Renal dysfunction in workers exposed to inorganic lead

INTRODUCTION

We studied the relationship between renal dysfunction and occupational lead exposure in a local factory.

MATERIALS AND METHODS

A cross-sectional study with a cohort component was conducted with 55 male workers of a factory producing PVC stabilisers as subjects. Workers from this factory have been followed up with 6-monthly blood lead measurements since 1982. Two indices of overall lead exposure, i.e. the number of times a worker's blood lead exceeded 40 micrograms/dL (PbB40) and cumulative blood lead index (PbBint), were obtained from this data. Recent blood lead concentration (PbB) was measured. 4-hour creatinine clearance and various other urinary and serum markers of renal dysfunction were used as effect indices.

RESULTS

There was no relationship between PbB and any of the renal markers. However, creatinine clearance decreased significantly (P < 0.001) with increasing PbB40 and PbBint after adjustment for age and smoking habits. Urinary albumin (Ualb), urinary alpha-1 microglobulin (U alpha 1m), urinary beta-2 microglobulin (U beta 2m) and urinary retinol-binding protein (URBP) increased significantly with both increasing PbB40 and PbBint. Total urinary activity of N-acetyl-beta-D-glucosaminidase (NAG-T) and its heat-stable isoenzyme (NAG-B) increased significantly with increasing PbB40. A significant difference in renal parameters occurred when PbB40 was 1 or more.

CONCLUSIONS

We have found a positive association between overall lead exposure and renal dysfunction. The renal parameters were significantly higher among those who had at least one episode of blood lead above 40 micrograms/dL. Our findings also strengthen the case for the use of Ualb, U alpha 1m, U beta 2m, URBP, NAG-T and NAG-B as early markers of lead nephropathy.

Urinary protein excretion in Type 2 diabetes with complications

This study examined the association between urinary markers of early diabetic nephropathy and non-renal diabetic complications in 946 patients with type 2 diabetes mellitus. The association with hypertension was also studied. Data on macrovascular complications (ischaemic heart disease, stroke, peripheral vascular disease) and microvascular complications (retinopathy, peripheral neuropathy) were obtained from case records and clinical examination. Urine samples collected were analysed for albumin, beta(2)-microglobulin, retinol-binding protein (RBP), and N-acetyl-beta-D-glucosaminidase (NAG). Results showed that urinary albumin, RBP and beta(2)-microglobulin levels were higher in patients with macro- and/or microvascular complications, compared to those without. NAG levels were higher only in patients with both types of complications. A higher proportion of patients with complications had abnormally raised urinary protein and enzyme levels, compared to those without. Patients with associated hypertension had higher urinary levels of albumin and beta(2)-microglobulin, regardless of whether complications were present or not. RBP excretion was, however, markedly higher only in patients with microvascular complications, whereas hypertension did not influence NAG excretion. Urine albumin and RBP excretion were predictive of microvascular, as well as both macrovascular and microvascular complications, whereas NAG excretion was predictive of macro- and microvascular complications. These findings could mean that increased urinary protein and enzyme excretion were associated with more severe disease in these patients.

Urine protein excretion among Chinese patients with type 2 diabetes mellitus

BACKGROUND

Urinary excreton of low molecular weight proteins such as beta2-microglobulin and retinol binding protein (RBP), and enzymes such as N-acetyl-beta-D-glucosaminidase (NAG), may be useful as indicators of renal tubular dysfunction in diabetes mellitus.

OBJECTIVE

To describe the profile of urinary protein and enzyme excretion in 240 Chinese patients with type 2 diabetes mellitus in Singapore.

MATERIALS AND METHODOLOGY

Cross-sectional study of consecutive patients presenting for follow-up at a Government primary care clinic. Information was obtained from interview, physical examination and laboratory analysis. Data analysis included descriptive statistics on urinary protein and enzyme excretion, comparison of unadjusted and adjusted means of these among patient subgroups, as well as correlation with control of diabetes and other clinical parameters.

RESULTS

Albuminuria correlated with urine beta2-microglobulin (r=0.34, p<0.01) and RBP (r=0.46, p<0.01). Hypertensive patients had significantly higher mean urine albumin (geometric mean 15.13 mg/gCr) and beta2-microglobulin (363.18 microg/gCr) levels compared to patients without hypertension (7.07 mg/gCr; 219.20 microg/gCr; p<0.05). Patients with complications of diabetes also had higher albumin (15.55 vs 6.20 mg/gCr), beta2-microglobulin (344.47 vs 288.83 microg/gCr) and RBP excretion (152.02 vs 94.54 mg/gCr). Two-hour postprandial sugar correlated with beta2-microglobulin (r=0.33, p<0.01), RBP (r=0.35, p<0.01) and NAG (r=0.28, p<0.01). Urinary protein excretion did not correlate with HbA1c, fasting blood sugar, age of patient or duration since diagnosis.

CONCLUSION

These results among 240 Chinese patients in Singapore were consistent with reports from other study populations.

Myopia: gene-environment interaction

INTRODUCTION

Myopia has reached epidemic proportions in Japan, Hong Kong, Taiwan and Singapore. This review summarises the evidence for environmental and genetic factors as well as gene-environment interaction for myopia for both epidemiologic studies as well as animal models.

METHODS

A literature review was conducted after a Medline search on articles on the genetic or environmental aetiology of myopia in animal or epidemiologic studies. Articles on the methodology of gene-environment studies were also reviewed. All articles reviewed were articles published in peer-reviewed journals.

RESULTS

Cross-sectional studies have found a positive association between myopia and near work activity such as reading and writing. Likewise, laboratory research has shown that environmental factors such as visual deprivation may lead to the development of myopia in animals. While linkage studies in humans are currently being conducted to identify possible markers for myopia in the human genome, several neurotransmitters, modulators and growth factors that influence refractive development have already been identified in animal models that may help identify candidate genes. Epidemiologic studies have also evaluated the combined effects of hereditary factors, environmental factors and gene-environment interaction on myopia development.

CONCLUSIONS

Both genes and environmental factors may be related to myopia. There are no conclusive studies at present, however, that identify the nature and extent of possible gene-environment interaction. Further linkage analysis, affected sib-pair studies, and family-based association studies may better identify the nature of gene-environment interaction.

Incidence of ischaemic heart disease and stroke in Chinese, Malays and Indians in Singapore: Singapore Cardiovascular Cohort Study

INTRODUCTION

This is the first prospective cohort study in Singapore to describe the incidence of ischaemic heart disease (IHD) and stroke among Chinese, Malays and Asian Indians.

MATERIALS AND METHODS

The Singapore Cardiovascular Cohort Study is a longitudinal follow-up study on a general population cohort of 5920 persons drawn from 3 previous cross-sectional surveys. Morbidity and mortality from IHD and stroke were ascertained by record linkage using a unique identification number with the death registry, Singapore Myocardial Infarct Registry and in-patient discharge databases.

RESULTS

There were 193 first IHD events and 97 first strokes during 52,806 person-years of observation. The overall incidence of IHD was 3.8/1000 person-years and that of stroke was 1.8/1000 person-years. In both males and females, Indians had the highest IHD incidence, followed by Malays and then Chinese. For males after adjusting for age, Indians were 2.78 times (95% CI 1.86, 4.17; P < 0.0001) and 2.28 times (95% CI 1.34, 3.88; P = 0.002) more likely to get IHD than Chinese and Malays respectively. For females after adjusting for age, Indians were 1.97 times (95% CI 1.07, 3.63; P = 0.03) and 1.37 times (95% CI 0.67, 2.80; P = 0.39) more likely to get IHD than Chinese and Malays respectively. For stroke, male Chinese and Indians had higher incidence than Malays (though not statistically significant). However, in females, Malays had the highest incidence of stroke, being 2.57 times (95% CI 1.31, 5.05; P = 0.008) more likely to get stroke than Chinese after adjustment for age.

CONCLUSIONS

This prospective study of both mortality and morbidity has confirmed the higher risk of IHD in Indians. It has also found that Malay females have a higher incidence of stroke, which deserves further study because of its potential public health importance.

Incidence and survival characteristics of retinoblastoma in Singapore from 1968-1995

PURPOSE

To describe the incidence and survival of 69 Singapore residents with retinoblastoma in all Singapore hospitals from 1968-1995.

METHODS

Data of all Singapore residents diagnosed with retinoblastoma from 1968-1995 were collected by the Singapore Cancer Registry based on notifications from physicians, pathology records, hospital discharge records, and death certificates. The medical records of 46 patients were traced, and information on laterality of tumor, spread of tumor, mode of treatment, and family history of retinoblastoma was obtained. Time trends and survival characteristics of the cohort were described.

RESULTS

The incidence rate of retinoblastoma was 2.4 per 1 million for children <9 years and 11.1 per 1 million for children <5 years. The incidence of retinoblastoma has been almost uniform over time from 1968-1995, except for an apparent increase in 1988-1992. The 3-year survival rate for retinoblastoma was 83%. Survival rates were higher in children <2 years because children who present at a younger age may have tumors diagnosed at earlier stages of the disease. There was no difference in survival rates for sex, race, laterality, family history of retinoblastoma, treatment, or year of diagnosis.

CONCLUSION

Retinoblastoma is the most common eye cancer in children that may cause blindness or death. The incidence rates of retinoblastoma in Singapore have remained fairly stable over 28 years, and the survival rate is higher in younger children. This study will be helpful in monitoring future disease patterns in Asian populations.

Markers of diabetic nephropathy

It is well established that the detection of microalbuminuria in a patient with diabetes mellitus indicates the presence of glomerular involvement in early renal damage. Recent studies have demonstrated that there is also a tubular component to renal complications of diabetes, as shown by the detection of renal tubular proteins and enzymes in the urine. In fact, tubular involvement may precede glomerular involvement, as several of these tubular proteins and enzymes are detectable even before the appearance of microalbuminuria. This review looks at the studies reported so far on serum and urinary markers of diabetic nephropathy, both glomerular and tubular, and their roles in the early detection of renal damage. The advantages and disadvantages of some of these markers are also discussed. The markers reviewed include (1) glomerular--transferrin, fibronectin, and other components of glomerular extracellular matrix, and (2) tubular--low molecular weight proteins (beta 2 microglobulin, retinol binding protein, alpha 1 microglobulin, urine protein 1), other proteins such as Tamm-Horsfall protein, beta 2 glycoprotein-1, urinary enzymes (N-acetyl-beta-D-glucosaminidase, cholinesterase, gamma glutamyltranspeptidase, alanine aminopeptidase), and tubular brush-border antigen.

Helicobacter pylori infection and gastric emptying of indigestible solids in patients with dysmotility-like dyspepsia

BACKGROUND

The role of Helicobacter pylori and gastric motility in dysmotility-like dyspepsia is unclear. The aim of this study was to determine whether delayed gastric emptying of indigestible solids and H. pylori infection are associated with dysmotility-like dyspepsia.

METHODS

Thirty-two healthy volunteers and 72 patients fulfilling the criteria of dysmotility-like dyspepsia received a gastric emptying test using radiopaque markers, and the H. pylori status was determined by histology.

RESULTS

Twenty-seven percent of volunteers were H. pylori-positive, compared with 32% in the dyspeptic groups (P = NS). Gastric emptying was significantly slower in dyspeptic patients than controls and in H. pylori-positive patients than H. pylori-negative patients. Subjects with gastroparesis have a higher chance of developing dysmotility-like dyspepsia (odds ratio (OR), 2.5) than subjects with normal gastric emptying. Subjects with H. pylori and gastroparesis have an increased likelihood of developing dysmotility-like dyspepsia (OR, 4.3) than if either factor were present alone.

CONCLUSION

Our data suggest that gastroparesis alone and gastroparesis and H. pylori infection are associated with dysmotility-like dyspepsia.

Three-year follow-up of serial nerve conduction among lead-exposed workers

OBJECTIVES

The purpose of this study was to investigate the serial nerve conduction parameters of a group of lead-exposed workers and determine their correlation with the serial blood lead results over a three-year period. A "nonresponse" level (defined as no significant changes in the nerve conduction in response to changes in the blood lead level of each exposed worker observed over the period of study) was also determined for blood lead in respect to the peripheral nerves.

METHODS

Seventy-two male workers from a lead battery manufacturing factory were followed at six-month intervals for three years. At each follow-up, the blood lead level was determined and nerve conduction tests (ulnar and median nerves) were conducted. A group of 82 unexposed subjects served as referents.

RESULTS

Significant differences were observed for some of the mean values of the median nerve conduction parameters between the exposed and reference groups. The 28 exposed workers who completed the follow-up were divided into the following two blood lead categories: < 40 micrograms. 100 ml-1 (< 1.93 micromol.l-1 and > or = 40 micrograms. 100 ml-1 (> or = 1.93 micromol.l-1. In the latter, the median motor conduction velocity, median distal latency, median amplitude, ulnar motor conduction velocity, and ulnar amplitude were significantly correlated (adjusted for age and within-subject variation) with the blood lead levels, but not so in the former.

CONCLUSIONS

According to the three-year serial results, the "nonresponse" level for blood lead with respect to the peripheral nerves would be < 40 micrograms.100 ml-1 (< 1.93 micromol. l-1).

Biomarkers: a molecular approach to cancer epidemiology

The rapidly evolving field of molecular epidemiology represents a paradigm shift in medical research, illustrating the convergence of basic, clinical and epidemiologic research, as well as the application of laboratory methods to population studies. As a complement and improvement to traditional epidemiologic approaches, the use of biomarkers offers more accurate exposure assessment at an individual level, better understanding of the biological mechanism of carcinogenesis by tracking a continuum of events between exposure and disease, as well as the ability to discern host or genetic factors that may account for variations in cancer susceptibility among individuals or ethnic populations. Further, biomarkers have also proved useful in cancer diagnosis and prognosis. This article attempts to review the latest progress in molecular cancer epidemiology and to summarise the current state of research in Singapore.

Molecular epidemiology: issues in study design and statistical analysis

Traditional analytical epidemiology is directed at identifying the association between risk factors and occurrence of disease by using crude exposure data derived from questionnaires or clinical measures, and taking clinical disease as the end point. With the rapid development in molecular biology and laboratory methods, it is now possible to use biomarkers which are capable of identifying molecular events for epidemiologic research. This improved sensitivity enables us to develop a mechanistic understanding of disease causation: a step closer to the unravelling of the "black box" of traditional epidemiology. Biomarkers may be classified as internal indicators of exposure (biomarkers of exposure), indicators of preclinical adverse effect (biomarkers of effect) or indicators of an intrinsic or acquired susceptibility to disease (biomarkers of susceptibility). Biomarkers provide a better definition of exposure and disease status and consequently they could help to reduce misclassification bias in both exposure and disease, reduce the follow-up time in prospective studies, as well as identify possible interactions between risk factors on disease occurrence. However, a biomarker needs to be validated and its distribution in large populations described before it can be used profitably for aetiologic research. Also, the use of biomarkers in epidemiologic research raises other interesting epidemiological and statistical issues like confounding, effect modification and the analysis of repeated measurements. Molecular epidemiology is a multidisciplinary endeavour which comprises molecular biology, epidemiology and biostatistics. Clearly then, to carry out research in this field profitably, the molecular biologist, epidemiologist and biostatistician must acquire not only expertise in their respective fields, but also an integrated understanding of all three fields. The molecular biologist is not merely a laboratory bench worker; the epidemiologist, a field data-collector and the biostatistician, a number cruncher. They must work together to pry open the "black box" to gain a greater insight into how risk factors operate to initiate disease onset and ultimately to make use of this knowledge base to implement preventive measures.

Regression analysis of biomedical research data based on a repeated measure or cluster sample

Research in biology and medicine often entails estimating the effect of an exposure variable X on a response variable Y from a cluster sample, that is, where X and Y may be measured repeatedly from the same subject (cluster), or X and Y may be measured from two or more subjects based on some related grouping such as a litter or household (cluster). Typically, regression analysis of the data is performed ignoring the subject or grouping (cluster) identity. This analytical approach has two drawbacks. First, the statistical inference of the regression coefficient of Y on X (beta) ignored cluster identity will likely be biased. More serious though, is that beta ignored cluster identity will likely be quite discrepant from the average within-cluster beta. It is the latter that is of relevance to the research question. Indeed it is not clear what beta ignored cluster identity really conveys. We describe a multiple regression model for the analysis of data from a cluster sample. The model treats the cluster as a nominal confounding variable to be adjusted. The idea is to represent the cluster by a set of dummy variables to be included as explanatory variables in the regression model. This model gives the average within-cluster beta with valid statistical inference. Numeric examples were used to illustrate the application of the dummy variable multiple regression model. This statistical method can be implemented by any software package that includes multiple regression, and virtually all commercial packages include this procedure.