Serum ferritin as a predictor of host response to hepatitis B virus infection.

With hemodialysis patients, a high serum ferritin before there was serological evidence of hepatitis B virus infection increased the likelihood that the infection would be persistent. This finding suggested that hepatitis B virus is likely to infect and actively replicate in liver cells with the propensity for increased ferritin synthesis. The virus itself could stimulate the synthesis of ferritin in a cyclic positive feedback mechanism that increases intracellular ferritin concentration and, eventually, intracellular iron. Transformed liver cells have low iron content, do not replicate hepatitis B virus, and require iron for growth. Infected, nonmalignant liver cells could supply iron to the transformed cells and nourish their expansion.

Body weight, dietary practices, and tumor susceptibility in the rat.

The mature body weight of outbred CD-Crl:COBS CD(SD)BR male rats permitted to select their own diets throughout life was shown to correlate linearly with the frequency of spontaneous tumors. In an analysis of the dietary practices of these animals, a multifactorial model was derived--for which prepuberty and early post-puberty data only were used--that accounted for a large proportion of the variance in mature weight. The level of food or calorie intake proved less informative than the following: the interactive effects of food intake relative to body weight, the proportion of protein and carbohydrate constituents in the diet, the intake of each of these components, and, particularly, the efficiency with which the diet consumed was used for growth during early postweaning life. In a prospective study of other rats maintained on a similar feeding regimen, the model was found to be an accurate predictor of mature weight. Several of the explanatory factors necessary for estimation of body weight also contributed to tumor susceptibility. Nonetheless, both dietary and body weight-related factors in a specific time-related sequence were required to explain variation in spontaneous tumor susceptibility.

Dietary practices of early life and age at death of rats with tumors.

The association between the age at death of rats that develop spontaneous tumors and their early life dietary practices and growth responses was investigated. The data were obtained from outbred male Charles River CD Crl:COBS CD (SD)BR rats maintained for life on a self-selection dietary regimen that permitted them to regulate their protein intake apart from energy intake. A series of multivariable models was derived, after the rats were grouped by tumor type, that accurately accounted for the age of death of the individuals in each class. The models that explain the variance in death age of rats that develop tumors also differed from the combination of factors for rats that do not develop tumors but succumb to other age-related diseases. Although dietary protein and measures of growth are ubiquitous features in these models, no one set of idealized conditions for longevity can be proposed.

Contribution of body weight and growth to risk of anterior pituitary gland tumors of the rat.

The dietary and growth histories of noninbred male Charles River COBS rats permitted freedom of dietary choice throughout life were analyzed to determine the conditions associated with the spontaneous occurrence of anterior pituitary gland tumors (PT). The animals were allowed to regulate their protein intake apart from energy intake. Multivariable models were derived distinguishing rats that develop PT from those remaining free of tumors throughout life. Conventional fixed diet studies suggested that the quantity and quality of the diet consumed would be the best explanatory variables. However, body weight-related variables were found to be more important. The interacting, time-related conditions conductive to a high probability of PT occurrence included an unusually high growth rate associated with a high level of conversion of the food consumed into body mass during early life and a high intake of protein relative to body weight during early adult life. The accuracy in identifying rats with PT was even greater if mature body weight was included as an explanatory variable.

Validation of a breast cancer risk assessment model in women with a positive family history.

BACKGROUND: Gail et al. developed a statistical model for estimating the risk of developing breast cancer in white women screened annually with mammography. This model is used for counseling and for admission to clinical trials. PURPOSE: We evaluated the model prospectively in a cohort of women with a family history of breast cancer. METHODS: We followed women who participated in the American Cancer Society 1987 Texas Breast Screening Project. The model was evaluated by comparing the observed (O) and expected (E) numbers of breast cancers using composite background rates from both the Breast Cancer Detection and Demonstration Project and the Surveillance, Epidemiology, and End Results program of the National Cancer Institute. Data were partitioned by adherence to American Cancer Society screening guidelines. RESULTS: The Gail et al. model predicted the risk well among women who adhered to the American Cancer Society guidelines (O/E = 1.12; 95% confidence interval = 0.75-1.61) but overpredicted risk for women who did not adhere to the guidelines. There was an indication that the model overpredicted risk for women younger than 60 years old and underpredicted risk in women aged 60 years and older. CONCLUSIONS: Overall, the Gail et al. model accurately predicts risk in women with a family history of breast cancer and who adhere to American Cancer Society screening guidelines. Thus, the model should be used as it was intended, for women who receive annual mammograms.

Forecasting the development of primary hepatocellular carcinoma by the use of risk factors: studies in West Africa.

An association between hepatitis B virus (HBV) and primary hepatocellular carcinoma (PHC) has been found in several studies in Africa, Asia, and elsewhere. In this paper we considered the interrelations between several events related to HBV infection, which include the presence of: 1) hepatitis B surface antigen (HBsAg), 2) antibody to hepatitis B core antigen (anti-HBc), 3) antibody to the surface antigen (anti-HBs), 4) chronic liver disease, 5) elevated alpha-fetoprotein, and 6) PHC. With the use of preliminary epidemiologic data, risk factors related to these events were calculated. We suggested that the interactions between these events and HBV infection in parents be used to estimate the risk of PHC for an individual in this environment.

Lack of influence of T-cell marker and importance of mediastinal mass on the prognosis of acute lymphocytic leukemias of childhood.

One hundred consecutive new cases of acute lymphocytic leukemia (ALL) were studied in patients that were 4 months to 16 years of age when admitted to The Children's Hospital of Philadelphia. Various prognostic factors were examined and related to the duration of the first remission. These factors included lymphoblast surface markers, age at diagnosis, sex, race, initial white blood counts (WBC), presence of mediastinal mass, degree of hepatosplenomegaly, and lymph node size. Classification by lymphoblast surface markers showed 22 T-cell, 71 null cell, and 3 B-cell leukemias; 3 cases were unclassifiable and 1 had both T- and B-cell markers. Statistical analysis indicated that stratification by presence or absence of mediastinal mass was necessary. Most patients with mediastinal masses, 5 of thymus and 4 of the non-thymus type, fared poorly with a median duration of continuous complete remission of less than 12 months as compared with greater than 48 months for those without such masses. The most satisfactory model to estimate remission duration in children without mediastinal masses was dependent upon initial WBC, sex, race, and surface markers. Low WBC, white race, female sex, and T-cell markers in patients without mediastinal masses were associated with a favorable prognosis. The findings suggest that patients with mediastinal masses need special therapy and that T-cell ALL without a mediastinal masses need special therapy and that T-cell ALL without a mediastinal mass does not carry a poorer prognosis than does null cell ALL.

Identification of an NAD(P)H:quinone oxidoreductase polymorphism and its association with lung cancer and smoking.

The enzyme NAD(P)H:quinone oxidoreductase (NQO1) catalyses bioreduction and bioactivation reactions. A mutation in the NQO1 gene had previously been demonstrated in a cancer cell line with reduced NQO1 activity. In this study, several regions of the NQO1 locus were examined for constitutional variation at the DNA level. The previously described mutation in exon 6 was detected by the single-strand conformation polymorphism technique. This was confirmed by sequencing to result from a C-->T substitution. Genotype analysis in the Centre d'Etude Polymorphisme Humain (CEPH) reference panel revealed two alleles with frequencies of 0.87 and 0.13 and demonstrated Mendelian transmission. Genotype distributions were consistent with Hardy-Weinberg equilibrium. Linkage analysis mapped the gene locus to chromosome 16q. NQO1 was felt to be a candidate gene for the susceptibility to lung cancer, given its potential role in protection against carcinogenic compounds. The frequency of NQO1 variants was examined in 150 lung cancer cases and in two reference populations. The allele distribution in CEPH parent controls was significantly different from cases (chi 2 = 5.52, p = 0.019), but no difference was noted between cases and a healthy local reference population. When the local reference distribution was stratified on smoking status, a significant difference was observed (chi 2 = 3.88, p = 0.048).(ABSTRACT TRUNCATED AT 250 WORDS)

Dietary habits and the prediction of life span of rats: a prospective test.

A multivariable statistical model for predicting the duration of life of rats permitted to select their own diets was evaluated prospectively. With information relating to pre-adult dietary habits and growth responses, the predicted length of life for an individual closely matched the observed life span; the average absolute error was 11%. This suggests that, even under normal feeding conditions, the diet/growth history is an important factor bearing on death rate. In general, an early adult death age is associated with a high food intake prior to adulthood particularly when coupled with a high efficiency of food utilization during the post-puberty period, a rapid growth rate and early attainment of mature weight. Deviations from this pattern serve to increase the duration of life of the individual.

Identification of women at increased risk for breast cancer in a population-based screening program.

A multivariate model to assess breast cancer risk was developed by Gail et al. (M. H. Gail, L. A. Brinton, D. B. Byar, D. K. Corle, S. B. Green, C. Schairer, and J. J. Mulvihill, J. Natl. Cancer Inst., 81: 1879-1886, 1989) based on data analysis of the Breast Cancer Detection and Demonstration Project. We evaluated the model's usefulness for assigning women to risk groups for counseling and follow-up by applying it to the 1987 Texas Breast Screening Project data. We identified 3165 women with one or more first-degree relatives affected with breast cancer. The mean risk score for the group was 3.3 (range, 2.7-11.8), indicating a greater than 3-fold elevated risk. The mean risk score for the remaining 27,439 women without affected first-degree relatives was 1.5 (range, 1.24-3.2). Risk perception was found to be a motivator for participation. Women with a risk score greater than 5 perceived themselves to be at high risk for breast cancer. The perception of risk was related to the type of affected first-degree relatives: 80.0% of the women with three affected first-degree relatives and 71.5% of women whose mother and sister were both affected with breast cancer perceived themselves to be at high risk. The Gail model is potentially useful in the clinical setting because women at high risk for breast cancer can be entered into etiological studies, enrolled in primary prevention trials, or referred to programs seeking to improve compliance with screening mammography. The Gail model needs validation, but it is useful for estimating the risk of breast cancer in large populations.

Some problems of inference in cohort studies.

This paper uses a detailed example to illustrate how to detect individual observations that disproportionately influence the results of hypothesis testing with relative risk regression models.

Immune responses to hepatitis B virus and tuberculosis infections in Southeast Asian refugees.

Both Mycobacterium tuberculosis and hepatitis B virus infections are common in the Philadelphia Southeast Asian refugee population. Among 224 hepatitis B carriers identified between January 1, 1982 and March 31, 1984, there was a statistically significant association between a negative tuberculin skin test (purified protein derivative (PPD)) and viral replication (hepatitis B e antigen positivity (HBeAg)). This finding suggests that bacille Calmette Guérin (BCG) vaccination might reduce the prevalence of infectious carriers, thereby ultimately reducing the incidence of hepatitis B infection.

Assessing the adequacy of the logistic regression model for matched case-control studies.

We apply results for the detection of outliers in logistic regression to the analysis of matched pairs data and illustrate the techniques with data from a matched pairs study of the relationship between iron-binding proteins and mortality in the Solomon Islands. Although the study includes 90 matched pairs, the conclusions change substantially with deletion of one or two pairs. Thus, even with reasonably large data sets, single observations may have a great impact on the results of an analysis. The identification of influential observations should constitute part of the analysis of matched case-control data.

Somatic allele loss in genetic linkage analysis of cancer.

The ability to detect or reject genetic linkage in studies of human cancer is often diminished because multiple affected relatives in a pedigree are unavailable for analysis. The observation of somatic allele loss in tumors can provide knowledge about gametic phase. Therefore, consideration of tumor genotype data could be used to obtain knowledge about gametic phase ordinarily gained from a larger sample of individuals in cancer families. The objective of the present study is to describe a method for improving the power to detect or reject genetic linkage by using knowledge about somatic genetic changes in tumor tissue. A modification to the lod score method of linkage analysis is proposed in which knowledge of gametic phase in the linkage likelihood is inferred from observations of loss of constitutional heterozygosity (LoH) in tumor tissue. This methodology was evaluated using a double backcross nuclear family with a pair of offspring. The expected lod score improved substantially when tumor genotype data were included in the analysis. For example, when the haplotype remaining in tumor tissue was identical to the inherited haplotype in constitutional tissue 99% of the time, linkage analyses without tumor genotype data would require a 2-5 times larger sample of offspring pairs to conclude linkage with an expected lod score value of 3 or greater, compared to analyses incorporating tumor genotype data. These results suggest that consideration of tumor genotype data using the proposed method can substantially improve the power of linkage analyses in cancer families.

Using loss of heterozygosity data in affected pedigree member linkage tests.

Linkage analysis can be used to test the hypothesis that a marker locus of known location segregates independently from a presumed disease gene. One way to test this hypothesis is to measure the similarity of marker alleles among pairs of relatives affected with the disease. When the disease under consideration is cancer, it is possible to take advantage of the marker alleles in tumors to revise the similarity measure obtained from the observations made in constitutional tissue. Only cancers that arise through the model of recessive oncogenesis are amenable to this revised analysis. This model postulates that cancer is caused by somatic genetic changes which result in the loss of one or both copies of a normal allele at a tumor suppressor locus. If an individual's inherited genotype is heterozygous at the marker locus, the model of recessive oncogenesis suggests that we may observe loss of constitutional heterozygosity at the marker locus in the tumor. In this report, we how how to incorporate this loss of heterozygosity data into affected pedigree member linkage tests. The revised procedure is illustrated using data obtained from relatives with breast cancer. Substantial improvement in the power to reject the different chromosome hypothesis is obtained when loss of heterozygosity is observed in multiple relatives with the same marker alleles retained in the tumors.

Study design for a hepatitis B vaccine trial.

A short-time trial of small sample size for an evaluation of the hepatitis B vaccine is proposed and designed. The vaccine is based on the premise that antibody to the surface antigen of the hepatitis B virus is protective against viral infection. This premise is verified by using the presence of the surface antigen as the marker of infection and comparing infection rates in renal dialysis patients who had naturally acquired antibody to patients without antibody. Patients with antibody have an extremely low risk of infection. The probability of remaining uninfected decreases at an exponential rate for patients without antibody, implying a constant risk of infection at any point in time. The study design described makes use of this time independence and the observed infection rates to formulate a clinical trial which can be accomplished with a relatively small number of patients. This design might be useful if, in preliminary studies, it is shown that the vaccine produces antibody in the patients and that protection against hepatitis B virus would be beneficial to the patients.

Dietary practices of early life and spontaneous tumors of the rat.

A complete life history was obtained of the dietary practices, growth responses, and diseases of outbred male Charles River COBS rats permitted to select their own diets. Simple correlation analyses proved inadequate for detecting an association between any single factor at any age and tumor development. A multifactorial model showed that animals that developed tumors could be distinguished from those that did not, solely on the basis of age- or weight-specific information prior to maturity. The combination of variables that maximize the probability of a neoplasm is: 1. a high absolute protein intake shortly after weaning, 2. a high degree of efficiency in converting consumed food into body mass at the time the individual enters puberty, 3. a high level of protein intake relative to body weight during the early adult period and concomitantly, 4. a high level of food intake, and 5. a rapid growth rate during early postnatal life so that comparatively less time is required to attain a specified body weight than a subsequent weight increment. The opposite set of conditions reduces the probability of neoplasm occurrence, which is even further minimized if the proportion of protein in the diet in early postweaning life and the absolute intake of protein during the early adult period are relatively high. Differences in temporal-specific dietary practices and growth responses accounted for differences in the tissue origin of the tumors.

Changes in serum iron levels due to infection with hepatitis B virus.

We found in two previous studies (Down syndrome patients and end-stage kidney patients receiving renal dialysis) that total serum iron is higher on average in carriers of the hepatitis B virus than in those who are not. The elevation of the serum iron is independent of elevations of serum L-alanine:2-oxoglutarate aminotransferase, EC 2.6.1.2) (SGPT), an indicator of liver cell damage. We have followed for 10 yr a large number of patients with end-stage renal disease receiving renal dialysis. In this paper we describe studies of serum iron and SGPT levels in patients (i) 1 mo before infection, (ii) after infection but within the month of infection, and (iii) 6-12 mo after infection. Comparisons of serum iron levels were made between those infected who retained the virus (carriers) and those who rejected the infection (transients). There were no differences between these groups before infection. Serum iron remained high in the carrier group and dropped in the transients. However, not all of the carriers retained high levels, although this was the case in general. Individual changes in the pre- and postconversion period were then considered. All carriers who had a preconversion decline in iron had an increase after infection, whereas this occurred in only some of the transients. Those carriers who had a decline after infection had raised levels before infection, and the decline was generally less than the increase. Consideration of the SGPT and the iron levels together led to the same conclusion as the previous studies, that elevation of iron may be independent of rise in SGPT. Several hypotheses were derived from these findings. Individuals who are carriers in general have higher iron levels and, therefore, are more likely to become infected with bacteria; this may contribute to increased morbidity and mortality. From experimental evidence, iron is required for the growth of tumor cells. Carriers with elevated iron levels may be more likely to develop detectable cancer of the liver than those who do not.