Trends of testicular cancer within the United States by histological subtype amongst whites and blacks, 1973-2003

PDF

Download

Open document

Flip pages

Summarize

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content TRENDS OF TESTICULAR CANCER WITHIN THE UNITED STATES BY HISTOLOGICAL SUBTYPE AMONGST WHITES AND BLACKS, 1973-2003 by Janelle Rodriguez ______________________________________________________________ A Thesis Presented to the FACULTY OF THE GRADUATE SCHOOL UNIVERSITY OF SOUTHERN CALIFORNIA In Partial Fulfillment of the Requirements for the Degree MASTER OF SCIENCE (BIOSTATISTICS) May 2007 Copyright 2007 Janelle Rodriguez ii TABLE OF CONTENTS LIST OF TABLES iii LIST OF FIGURES iv ABSTRACT v Chapter 1: INTRODUCTION 1 Chapter 2: METHODS 5 2-1: Sample Population 5 2-2: Age-Period-Cohort Model 8 2-3: Statistical Analysis 11 Chapter 3: RESULTS 13 3-1: Descriptive Analysis 13 3-2: Age-Drift, Age-Period, Age-Cohort Analysis 19 3-3: Age-Period-Cohort Analysis 24 3-3-1: Seminomas 24 3-3-2: Nonseminomas 28 3-3-3: US Black Males 32 Chapter 4: CONCLUSIONS 34 Chapter 5: FUTURE WORK 37 REFERENCES 38 Appendix A: Model Derivations 42 iii LIST OF TABLES Table 1: Histological subtypes of testicular cancer from SEER9 by race for ages 0-85 years 6 Table 2: Percent change in incidence rates of testicular germ cell tumors: SEER9, 1973-1978 to 1999-2003 17 Table 3: Age- specific incidence rates of US white seminomas (per 100,000) from SEER9 for period 1973-2003. The number of cases for which rates are based upon are in parenthesis. 20 Table 4: Age-Specific incidence of white seminomas rearranged in a longitudinal series by central year of birth 21 Table 5: Age effects in US white males among seminomas adjusted for period (in period drift model) and for cohort (in cohort drift model). Both models have a deviance of 118.6 with 59degrees of freedom and an estimated drift of 0.12 for both period and birth cohort. 23 Table 6: Age, period, and cohort effects in US white males 1973-2003 amongst seminomas. Additive effects, deviances and degrees of freedom (d.f.) for the age-period and age-cohort models estimated using Poisson regression. 24 Table 7: Predicted number of US white seminomas for 1973-2003. Pearson residual for each predicted value appears in parenthesis. 27 Table 8: Age- specific incidence rates of US white nonseminomas (per 100,000) from SEER9 for period 1973-2003. Number of cases for which rates are based upon are in parenthesis. 20 Table 9: Age-Specific incidence of white nonseminomas rearranged in a longitudinal series by central year of birth 21 Table 10: Predicted number of US white nonseminomas for 1973-2003. Pearson residual for each predicted value appears in parenthesis. 31 Table 11: Age- specific incidence rates of US black TGCTs (per 100,000) from SEER9 for period 1973-2003. The number of cases for which rates are based upon are in parenthesis. 33 iv LIST OF FIGURES Figure 1: Age-Specific Incidence of Testicular Germ Cell Tumors by Histological Subtype for Cases Diagnosed in 1973-2003 Amongst all Cases from SEER9 13 Figure 2: Age-Specific Incidence of Seminomas by Year of Diagnosis: SEER9 1973- 1978 to 1999-2003 15 Figure 3: Age-Specific Incidence of Nonseminomas by Year of Diagnosis: SEER9 1973-1978 to 1999-2003 16 Figure 4: Calendar period trends of the incidence of testicular germ cell tumors by histological subtypes amongst white and black SEER9 men from 1973-1978 through 1999-2003 18 Figure 5: Suggested order of analysis for temporal data of white seminomas 23 Figure 6: Maximum likelihood estimates of 10-year birth cohort effects from an age- period-cohort model fit to seminoma incidence data amongst white men aged 15-74 years using SEER9, 1973-2003 25 Figure 7: Maximum likelihood estimates of 5-year calendar period effects from an age-period-cohort model fit to seminoma incidence data amongst white men aged 15-74 years using SEER9, 1973-2003 26 Figure 8: Suggested order of analysis for temporal data of white nonseminomas 28 Figure 9: Maximum likelihood estimates of 10-year birth cohort effects from an age- period-cohort model fit to nonseminoma incidence data amongst white men aged 15-64 years using SEER9, 1973-2003 29 Figure 10: Maximum likelihood estimates of 5-year calendar period effects from an age-period-cohort model fit to nonseminomas incidence data amongst white men aged 15-64 using SEER9, 1973-2003 30 v ABSTRACT In order to address the question of whether or not the incidence of testicular germ cell tumors (TGCTs) has begun to abate, the incidence by histological subtype (seminoma and nonseminoma) among racial groups from 1973-2003 was examined on a national level. Incidence rates of TGCTs were obtained from the Surveillance, Epidemiology, and End Results Program and subjected to age-period-cohort analysis. Since 1973 to 2003, the incidence of all TGCTs has risen 50% (95% CI: 40.2 59.8). The overall incidence of TGCTs among white men has been consistently increasing in the US since 1973 with black males showing a recent increase in the last 10 years. Analysis of the incidence of TGCTs amongst white men revealed that birth cohort was the dominating risk factor for both seminoma and nonseminoma subtypes. 1 Chapter 1: INTRODUCTION Testicular cancer is a rare malignancy responsible for 1% of all male tumors in the United States. 23 Ninety-five percent of all malignant testicular tumors develop from germinal cells (or reproductive cells) while 5% are other histological subtypes that originate from interstitial cells such as Sertoli and Leydig. 11,18 Metastatic testis tumors are uncommon, however they do exist as lymphomas and are usually seen amongst older patients. 39 The classification of germinal tumors falls into two main categories, seminomas and nonseminomas and will be the main focus of this paper. Unlike most other tumors, the incidence of testicular germ cell tumors (TGCTs) peak in young adulthood. Specifically, these tumors peak between the ages of 15 and 34 years and are the most common cancer of males seen across the United States and European populations for this age group. 5,14 Smaller peaks of TGCTs are seen in children younger than 5 years and in men older than 60 years. From the 1970s to the 1990s, the overall survival rate from testicular cancer rose from 10% to 90%. This increase in survival is due, in part, to effective diagnostic and chemotherapy techniques, improvements in serum tumor markers, and enhancements in surgical procedures. 39 In addition, since TGCTs are germ cell in origin they are sensitive to chemotherapy agents and radiation therapy favoring the higher survival probability. Recent studies have suggested that although an increase in survival is evident, survivors experience undesirable sequlae, including elevated rates of infertility 21,38 , sexual dysfunction, and other cancers 22 , as well as occurrence of second primary testicular cancers in numbers far exceeding those predicted by population incidence rates. 34 2 Seminomas, a subtype of germinal tumors, are further classified histologically as classic, anaplastic and spermatocytic. Classic seminoma accounts for roughly 80% of all seminoma cases and usually occurs in the late third decade. Anaplastic seminoma accounts for 10% of seminoma cases. However, in spite of its lower overall incidence, anaplastic seminoma is more aggressive than classic seminoma, and approximately 30% of seminoma deaths are due to this histologic type. Spermatocytic seminoma accounts for 10% of all seminomas and are believed to have few features in common with other TGCTs. For example, incidence of this histologic type usually peaks over the age of 60. Major histologic classifications of nonseminomas include embryonal carcinoma, choriocarcinoma, teratocarinomas and yolk sac tumors. Compared to seminomas, nonseminomas occur earlier in life, generally around age 20, and are highly malignant. Embryonal carcinoma accounts for roughly 40% of all nonseminomas while teratocarcinomas account for 30%. Yolk sac tumors occur more often in infants and children than in adults. The incidence of testicular cancer has been increasing at an uninterrupted rate in the last 50 years. 12,17 From 1973 to 1999, in the United States alone, incidence nearly doubled (going from 3.3 to 5.4 per 100,000). 12 Analysis by histologic subtype reveals that the rates of both seminoma and nonseminoma have been increasing, although seminomas have demonstrated a higher incidence. The increase in incidence has been consistent in areas that are considered to have good cancer registration and during a time interval wherein investigations of cases were complete and the accuracy of diagnostics high. 33,44 Increases of this magnitude and duration are unlikely to be due to improvements in clinical registration or screening of the disease. 3 In 1998, 7,200 cases were predicted to occur in the United States alone. 23 Significant variation in incidence rates is apparent when testicular cancer is observed on the level of racial ethnic groups. Within the United States, incidence is highest amongst white males (incidence is 5 times greater in white than in black men 26 ). In fact, an American white male has a 0.2 probability of being diagnosed with a TGCT in his lifetime. 39 These findings are consistent with studies worldwide, specifically studies in the United Kingdom and northern Europe. 5 Approximately, 1,400 new cases are diagnosed within the United Kingdom annually. 15 Incidence amongst white males differs depending on geographical region. In 1995, for example, rates in Denmark were reported to be 11.1 per 100,000 while in Finland, rates were much lower at 2.8 per 100,000. 5 Rates among black and white males follow the same age pattern, but rates amongst blacks are substantially lower in magnitude, both in the United States and in Africa. 12 Several risk factors, both environmental and genetic, have been suggested to be associated with testicular cancer, including urogenital conditions such as cryptorchidism, other congenital abnormalities, and family history of TGCTs. Several studies have reported an elevated risk of TGCTs to be 2 to 17 fold greater amongst boys born with cryptorchidism compared to those born with scrotal testes. 8 Cryptorchidism occurs when one or both testes fail to descend into the scrotum before birth. Approximately, 10% of TGCT cases have been diagnosed with undescended testes. Men with a family history of TGCTs are 8 to 10 times more likely to develop a malignancy if a male sibling has been previously diagnosed with testicular cancer. Furthermore, a man whose father has been diagnosed with a TGCT is 4 times more likely to develop a malignancy compared to a man with no family history of testicular cancer. 4 Age, period, cohort analysis has demonstrated that year of birth is a far more important risk factor than year of diagnosis. These results indicate that critical exposures may occur early in life and potentially in utero. 12 Recent studies have confirmed these findings by examining the difference of incidence of TGCTs across northern European countries and determining that later birth years were significantly associated with an increased risk independent of the level of incidence within the country. 29 Furthermore, studies have also demonstrated that a significant increase in risk seen in successive years of birth have predominately been seen following World War II. Certain studies in the United States and the United Kingdom suggest that specific racial groups are at a four or five times greater risk of TGCTs. 16 To that extent, conflicting opinions exist as to whether or not the increase in incidence of testicular germ cell tumors has abated in recent years. To address this question and to investigate in depth the effects of age, period, and birth cohort, we examine the incidence of TGCT by histological subtype among racial groups from 1973-2003 on a national level. In Chapter 2 of this paper, the sample population and statistical methods are described. A derivation of the age-period-cohort (APC) model and a detailed explanation of how the effects of the models components are measured are provided. In Chapter 3, the trends of TGCT incidence rates by histological subtype and race are provided based on the results of the age-period-cohort model. In Chapter 4, we interpret these findings and provide a discussion of how the effects of environmental and/or genetic exposures in the last 30 years have influenced the incidence of TGCTs. Finally, in Chapter 5 we discuss future work that will concentrate on the analysis of testicular germ cell tumors by racial/ethnic group in Los Angeles County from 1973 to 2003. 5 Chapter 2: METHODS 2.1 Sample Population National incidence rates of TGCT by histological subtype and race were obtained by the Surveillance, Epidemiology, and End Results (SEER) Program. SEER has collected and published data on cancer incidence from population-based cancer registries since 1973. The population covered by SEER is comparable to the general US population in regards to poverty and education. 41 Approximately 26% of the US population is currently covered by the 17 SEER registries (SEER). In order to look at long term trends, the analysis was restricted to the original 9 SEER registries which have provided data since 1973. Included are the following nine registries: Atlanta, Connecticut, Detroit, Hawaii, Iowa, New Mexico, San Francisco-Oakland, Seattle-Pudget Sound, and Utah. Histological subtype was defined by the International Statistical Classification of Diseases for Oncology, Third Edition (ICD-O-3). ICD codes are primarily used in cancer registries to be able to note the site and histology of all neoplasms. Classical seminoma and nonseminoma were included in this analysis. Seminomas which are postulated to originate from relatively undifferentiated germ cells, grow at a slow pace and usually stay localized. Nonseminomas on the other hand, are postulated to originate from more mature germ cells and more frequently present as aggressive tumors. Table 1 lists the specific subtypes of TGCTs with ICD codes and their classifications for the purpose of this analysis. 6 Table 1: Histological subtypes from SEER9 of testicular cancer by race for ages 0-85 years. Cases: 1973-2003 SEER9 ICD# Seminomas White Black Other Unknown Total % 9060 Dysgerminoma 428 13 11 4 456 4.50 9061 Seminoma,NOS 8,158 187 374 55 8,774 86.70 9062 Seminoma, Anaplastic 576 9 17 2 604 5.90 9063 Spermatocytic Seminoma 107 4 3 1 115 1.10 9064 Germinoma 148 13 7 1 169 1.80 Total 9,417 226 412 63 10,118 Non-Seminomas 9065 Germ Cell Tumor 29 0 6 1 36 0.47 9070 Embryonal Carcinoma, NOS 2,676 39 74 28 2,817 36.88 9071 Yolk Sac Tumor 192 7 20 3 222 2.91 9072 Polyembryoma 0 0 0 0 0 0.00 9080 Teratoma,malignant, NOS 481 11 26 2 520 6.80 9081 Teratocarcinoma 1,502 23 41 8 1,574 20.61 9082 Malignant teratoma, undifferentiated 8 0 3 0 11 0.14 9083 Malignant teratoma, intermediate 4 0 1 0 5 0.07 9084 Teratoma with Malignant Transformation 2 0 0 0 2 0.03 9100 Choriocarcinoma,NOS 149 7 7 1 164 2.15 9101 Choriocarcinoma + other germ cell elements 470 5 18 3 496 6.49 9102 Malignant teratoma, trophoblastic 1 0 1 0 2 0.03 9085 Mixed Germ Cell 1,651 49 78 11 1,789 23.42 Total 7,165 141 275 57 7,638 7 As noted earlier, about 95% of testicular tumors are germ cell in origin. Germ cells develop early in life at the embryonic stage and later form the reproductive cells found in both men and women. Seminoma subtypes include dysgerminoma, seminoma nos, spermatocytic seminoma, anaplastic seminoma and germinoma. Although only ovarian germ cell tumors are currently classified as dysgerminomas or germinomas, some pathologist in the past also classified some testicular germ cell tumors as these histological types. Consequently, we included all known male germinal tumors in the analysis. Nonseminoma subtypes consist of germ cell tumor, embryonic carcinoma nos, yolk sac tumor, polyembryoma, teratoma malignant nos, teratocarcinoma, malignant teratoma undifferentiated, malignant teratoma intermediate, teratoma with malignant transformation, choriocarcinoma nos, choriocarcinoma with other germ cell elements, and malignant teratoma trophoblastic. For the purpose of this analysis, mixed germ cell tumors were grouped together with nonseminomas because prior to 1990, these tumors were spread across nonseminoma subtypes and only after 1990, were they considered a separate category of germinal tumors (Percy, 1992). For this reason, it is difficult to examine the trends of incidence for individual subtypes of nonseminomas since mixed germ cell tumors may have been mistakenly classified as another entity. Furthermore, since spermatocytic seminoma peaks at an age of 65 it was excluded from this analysis. 8 2.2 Age-Period-Cohort Analysis Many investigations of cancer epidemiology involve analyzing temporal trends by age and year of diagnosis. This type of analysis summarizes the focus of cross-sectional data and the limits of its interpretations since; it cuts through an accumulation of birth cohorts at a specific point in time. Failing to consider longitudinal trends such as those from birth cohorts can lead to overlooking information that may provide insight into the etiology of a specific disease. For example, a cross-sectional analysis conducted on US females demonstrated that breast cancer mortality rates among younger women were decreasing while among older women the calendar period trends were increasing. 46 These results indicated that there were possibly different causes of breast cancer among pre- versus postmenopausal women, reflecting perhaps a difference in carcinogenic exposures. The results also suggested that breast cancer occurring in these two age groups may respond differently to treatment, or some combination of differences in etiology and response to treatment leading to the difference noted in calendar period trends. However, a longitudinal analysis on the same dataset revealed that the noted differences among younger and older womens mortality rates was due to a change in birth cohort trends. Breast cancer mortality rates in US women were actually increasing for women born before the 1920s. However, these rates moderated for women born between 1925 and 1940. 46 The moderation in mortality rates seen in more recent cohorts suggests that there has been a change in childbearing patterns following World War II. 45 This example helps to illustrate the significance of longitudinal analysis in making deductions about the behavior of a disease, i.e. determine if a disease is equally distributed across all ages or if 9 a disease appears to be more frequent amongst a group of individuals that share a common time reference (e.g. birth year). From the example described above, the importance of understanding how to interpret and estimate both period and birth cohort effects is undeniable. As a result, the age-period-cohort (APC) model has been introduced as a means of interpreting both cross-sectional and longitudinal analysis simultaneously. The model is defined as follows: Y ij = a i +B j +γ c where c = A a + p, A = the total number of age intervals, i = specific age interval, j = specific calendar period (or year of diagnosis), and Y = logarithmic disease rates for a given age and period. Parameter coefficients a, B, and γ measure the effects of age, period, and birth cohort, respectively. The APC model allows the effect of age, period, and cohort to be evaluated alongside each other. Specific to our analysis, the age effects will help to determine the patterns of testicular germ cell tumors (TGCTs) across all age groups, i.e. determine the rate at which the incidence of TGCTs is increasing (or decreasing) by age. The period effects will describe a change in risk over time that affect all age groups concurrently, whereas the cohort effects will describe the variations of incidence over time among US males that share the temporal characteristic of birth year. Understanding how to interpret the results of an APC model, however, can be difficult. Demonstrating the pattern of risk associated with successive birth cohorts is impossible to do since the linear trends common to both a cohort and the period of diagnosis cannot be separated. Such a variation in time that can be equally described by either birth cohort or period of 10 diagnosis is termed as a drift. 9,10 In order to properly control for the temporal variation present in APC models, a non-drift model must be implemented as part of the analysis and is defined as follows Y ij =a i +B j (j-j o )+(B c + B j )(c-c o ) where Y ij is the logarithmic disease rates for the i th age interval and j th period. Similarly, p o and c o correspond to a reference period and cohort, respectively. Since a birth cohort can be defined by means of calendar period and age, the non-drift APC model does not provide unique parameter estimates of the age, period, and cohort effects. Various parameterizations can be utilized to perform APC analysis and all depend on how much transfer of drift is made onto the age components from that of calendar year of birth year. Several methods have been proposed to present both period and cohort effects. The method implemented in this paper is one discussed by Clayton and Schiffler (1987) which constrains the first and last birth cohort to be zero. The non-drift effects of adjacent periods or cohorts can be expressed in terms of contrasts between the relative risks of these components. These contrasts are referred to as second differences since the effect of a period or cohort is measured by the difference in two differences. For example, suppose we wanted to compare direction and magnitude change from c=1 and c=2 to c=2 and c=3, our second difference could be expressed as (B 3 -B 2 )-(B 2 -B 1 )=B 3 -2B 2 +B 1. If a plot were generated of second differences, a zero value would indicate that the log- risk versus cohort is a straight line, or that there is no change in rates seen in adjacent eras, while a concave or convex relationship indicates an acceleration or deceleration of rates respectively. Although parameter estimates in an APC model may not be unique, 11 the differences in slopes are unique and are independent of transformation. Therefore, it is the difference in slopes that captures the period and cohort effects. A detailed derivation of the APC model is provided in the Appendix for further consideration. 2.3 Statistical Analysis The SEER*Stat statistical software was used to calculate incidence rates among white and black males from the SEER9 program. Rates were age-adjusted to the 2000 US population and consisted of ages 0-85 years. In order to calculate the percent changes in incidence, the difference in incidence between consecutive years of diagnosis is calculated then divided by the previous years incidence in the following fashion: % change = 1 1 2 Incidence Incidence Incidence − *100 where Incidence 2 = incidence rate in most recent year and Incidence 1 = incidence rate in previous year. The percent change can either be positive or negative and reflects the direction in which the incidence is moving in, i.e., incidence is either increasing or decreasing. In order to calculate a 95% confidence interval (CI) the following equation was implemented: % change ± 1.96 * se(%change) (Census Bureau, 2003) 12 where the se(% change) is defined as follows: se(% change) = 100 * * 2 1 2 1 2 2 2 2 1 2 Incidence SE Incidence SE Incidence Incidence + An age-period-cohort model was fit to determine the effects of age, year of diagnosis and birth cohort on incidence of testicular cancers. Age and year of diagnosis were divided into 5-year groupings in order to provide a more stable estimate of age specific trends by year of diagnosis. 25 Birth cohorts were created in the manner presented by Clayton and Schiffler (1987). The numbers of TGCTs by age, year of diagnosis and birth cohort, were generated separately for seminomas and nonseminomas and for blacks and whites. Seminomas were modeled in 12 age intervals (15-19 through 70-74 years), 6 year of diagnosis intervals (1973-1978 through 1999-2000) and 17 birth cohort intervals (1899-1908 through 1979-1988) each 10 years in length. The numbers of cases of TGCTs under age 15 or over age 74 were too low to fit in our model. One year of diagnosis interval, 1973-1978, consisted of 6 years while all the others consisted of 5 years. For the nonseminomas, 10 age intervals (15-19 through 60-64 years), 6 year of diagnosis intervals (1973-1978 through 1999-2000) and 15 birth cohort intervals (1909-1918 through 1979-1988) were modeled. To reference any birth cohort in the text, the fifth year in the interval is specified. For instance, the 1913 birth cohort identifies birth years 1909-1918. All analysis was conducted using the PROC GENMOD procedure in SAS v. 9.1 13 Chapter 3: RESULTS 3.1 Descriptive Analysis Over the 1973-2003 period, the average age-specific incidence of testicular germ cell tumors (TGCTs) varies with age, peaking at 30-34 years (Fig. 1). Figure 1: Age-Specific Incidence of Testicular Germ Cell Tumors by Histological Subtype for in 1973-2003 Amongst all Cases from SEER9 The height of the peak noted in Figure 1 is 12 cases of TGCTs per 100,000 males. Examining the age-specific incidence of TGCT by histological subtype suggests that seminomas and nonseminomas are different in nature. Seminomas do not develop until 15-19 years of age where their incidence starts to rapidly increase until age 35-39 years where this cancer peaks at 7.4 cases per 100,000 males. The incidence of this histological subtype starts to gradually decrease after this point until age 75-79 years where a small 0 2 4 6 8 10 12 14 0-4 5-9 10-14 15-19 20-24 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+ Age Incidence per 100,000 TGCT Seminomas Nonseminomas 14 peak is noted. This bi-modal figure is a direct reflection of the incidence of spermatocytic seminomas known to occur much later in life in comparison to all other seminomas. On the other hand, nonseminomas occur as early as 0-4 years and are not seen again until age 15-19 year of age and thereafter with increasing incidence until peaking at 25-29 years, and their frequency gradually decreases thereafter. At the highest peak, there has been an average of 6.1 cases of nonseminomas per 100,000 males from 1973 to 2003. In the younger age intervals, nonseminomas are more common than seminomas until age 30-34 years where seminomas begin to occur more often (Fig. 1).A 10-year difference is noted in the peak age interval for these histologic subtype. Furthermore, there are about 21% more cases of seminomas at the peak interval in comparison to nonseminomas. Accounting for all malignant cases of TGCTs from 1973-2003, 60% were seminomas (Table 1). Specifically, there have been 10,118 cases of malignant seminomas reported through SEER9 of which classical (NOS) was reported 86.7% of the time. Since, mixed germ cell tumors were not identified by a specific ICD code until after1991, the changes in the proportions of nonseminomas by histological subtype seen in Table 1 may be inflated by inclusion of mixed germ cell tumors in tallies for these years. It may appear that the proportion in some histologic subtypes declined after 1991; however, after that time, they are included in the mixed germ cell category. Among seminomas, incidence rates have been progressively increasing in all age groups between the 15-19 and 55-59 age intervals (Fig. 2). The peak age of incidence is between 30-39 years. From 60-74 years of age, incidence rates have been consistent, although some differences in older ages reflect apparently decreasing incidence of 15 spermacytic seminoma. Among nonseminomas, there has been less variation in incidence as seen from Figure 3 where the peak age is between 20-29 years. Figure 2: Age-Specific Incidence of Seminomas by Year of Diagnosis: SEER9 1973-1978 to 1999-2003 0 1 2 3 4 5 6 7 8 9 10 0-4 5-9 10-14 15-19 20-4 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+ Age Incidence per 100,000 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 16 Figure 3: Age-Specific Incidence of Nonseminomas by Year of Diagnosis: SEER9 1973-1978 to 1999-2003 Age-adjusted incidence rates by 5-year calendar period show a linear trend in TGCTs. The incidence of TGCTs rose by 50% (95% CI: 40.2-59.8) from 1973-1978 and 1999-2003 amongst all men living in the United States (Table 2). Examining calendar period trends by histological subtype, we find that at each period, seminomas were more common at every instant. Nonseminomas may prove to be more aggressive in their behavior, however less common to occur. Examining the incidence of TGCTs from 1973 to 2003 amongst white and black US males reveals distinct patterns both in the magnitude and frequency of this malignancys occurrence. Table 1 shows that cases of TGCTs are more common among white men than any other ethnicity. Among SEER9, white men have been responsible for 93.1% of all seminomas and 93.8% of all nonseminomas. Exploring the calendar 0 1 2 3 4 5 6 7 8 0-4 5-9 10-14 15-19 20-4 25-29 30-34 35-39 40-44 45-49 50-54 55-59 60-64 65-69 70-74 75-79 80-84 85+ Age Incidence per 100,000 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 17 trends per 5-year interval amongst white and black males shows that both white seminomas and nonseminomas were more common for every period from 1973-2003 (Figure 4). Table 2: Percent change in incidence rates of testicular germ cell tumors: SEER9, 1973-1978 to 1999-2003 Percent Change in Rate 95% CI Interval TGCT Seminoma Nonseminoma White Men 1973-1978 - - - 1973-1978 to 1979-1983 +17.5 (9.9-30.0) +18.18 (4.4-32.0) +16.67 (-0.1-33.4) 1979-1983 to 1984-1988 +12.7 (6.5-19.1) +15.38 (3.9-26.9) +4.76 (-8.8-18.3) 1984-1988 to 1989-1993 +11.32 (5.8-16.9) +13.3 (3.5-23.2) +9.09 (-4.1-22.3) 1989-1993 to 1994-1998 +1.69 (-3.0-6.4) +5.88 (-2.5-14.3) -4.17 (-7.1-15.5) 1994-1998 to 1999-2003 +6.67 (1.9-11.4) +8.3 (0.3-16.4) +8.7 (-3.9-21.3) 1973-1978 to 1999-2003 +60.0 (50.8-76.4) +77.27 (59.1-95.4) +38.9 (20.3-57.5) 1984-1988 to 1999-2003 +20.8 (15-26.6) +30 (19.3-40.7) +13.6 (0.2-27.1) Black Men 1973-1978 - - - 1973-1978 to 1979-1983 -11.11 (-33.3-55.5) -33.3 (-5.9-72.6) 0 (-92.4-92.4) 1979-1983 to 1984-1988 0 (-34.6-34.6) +25 (-53.4-103) 0 (-92.4-92.4) 1984-1988 to 1989-1993 -12.50 (-20.1-45.1) -20 (-30.2-70.2) 0 (-92.4-92.4) 1989-1993 to 1994-1998 +57.14 (5.0-109.3 ) +100 (-9.6-209.6) 0 (-92.4-92.4) 1994-1998 to 1999-2003 +18.18 (-9.4-45.8) 0 (-34.6-34.6) +66.7 (-60.3-193.7) 1973-1978 to 1999-2003 +44.4 (-22.2-111) +33.3 (-21.1-87.7) +66.7 (-60.3-193.7) 1984-1988 to 1999-2003 +62.5 (15.8-109.2) +60 (-14-134) +66.7 (-60.3-193.7) 18 Figure 4: Calendar period trends of the incidence of testicular germ cell tumors by histological subtypes amongst white and black SEER9 men from 1973-1978 through 1999- 2003. In the last 30 years, white males TGCT incidence rose 60% (95% CI: 50.8-76.4) from 1973-1978 to 1999-2003. Among seminomas, the incidence among white males has been increasing steadily, with an overall increase of 77% from 1973 to 2003. A significant increase of 8.3% (95% CI: 0.3-16.4) is noted in the most recent calendar period. Nonseminomas among white males were steadily increasing until 1989-1993. From 1989-1993 to 1994-1998 a decrease of 4.2% (95% CI: -7.1-15.5) in incidence is observed. However, in the last calendar period an 8.7% (95% CI: -3.9-21.3) increase in 0 1 2 3 4 5 6 7 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 Calendar Period Incidence per 100,000 White TGCT Black TGCT White Seminoma Black Seminoma White Nonseminoma Black Nonseminoma 19 incidence was noted. Overall, there has been a 38.9% (95% CI: 20.3-57.5) increase in nonseminomas among white men in the last three decades (Table 2). Among black men on the other hand, there has been a significant increase of TGCT occurrence from 1984- 1988 to 1999-2003 of 62.5% (95% CI: 15.8-109.2). From Figure 4, an increase of nonseminomas is noted in the last calendar period among black nonseminomas. 3.2 Age-Drift, Age-Period, Age-Cohort Analysis As described above, there are significant differences between race-specific rates of TGCT overall and specific histological types, with incidence of TGCTs being more common amongst white males compared with black males. As a result, the focus of the following models is on US white incidence rates. A model that would try to fit the incidence of any other race would result in unreliable parameter estimates due to significantly lower number of cases available for us to evaluate. The counts for which the following models are based are depicted in Tables 3 and 8. Furthermore, Tables 4 and 9 describe the incidence of TGCTs by birth cohort. 20 Table 3: Age- specific incidence rates of US white seminomas (per 100,000) from SEER9 for period 1973-2003. Number of cases for which rates are based are in parenthesis. Age/Period 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 15-19 0.2 (9) 0.3 (12) 0.2 (7) 0.3 (9) 0.4 (13) 0.6 (22) 20-24 1.7 (75) 2.2 (91) 2.7 (104) 2.2 (74) 2.8 (91) 3.6 (127) 5-29 4.8 (202) 4.9 (202) 6.2 (266) 7.2 (290) 7.7 (288) 7.3 (260) 30-34 4.6 (162) 6.6 (247) 8.2 (341) 9.8 (432) 9.6 (408) 10.2 (397) 35-39 5.1 (145) 5.9 (180) 7.9 (287) 8.8 (363) 9.9 (437) 10.7 (443) 40-44 3.7 (97) 5.2 (127) 5.9 (174) 7.5 (274) 8.0 (325) 9.0 (385) 45-49 3.3 (89) 4.6 (99) 4.2 (98) 4.6 (132) 5.5 (194) 5.7 (223) 50-54 2.8 (78) 2.6 (59) 2.8 (58) 2.9 (64) 3.1 (87) 3.3 (114) 55-59 1.8 (43) 1.8 (38) 2.0 (42) 2.0 (39) 1.6 (34) 1.9 (50) 60-64 1.7 (33) 1.6 (31) 1.2 (22) 1.9 (33) 1.4 (24) 1.3 (23) 65-69 0.8 (12) 1.2 (18) 1.1 (16) 1.1 (17) 1.5 (21) 1.5 (21) 70-74 0.9 (9) 1.1 (12) 0.7 (8) 0.7 (9) 0.7 (8) 0.6 (8) Table 8: Age-specific incidence rates of US white nonseminomas (per 100,000) from SEER9 for period 1973-2003. Number of cases for which rates are based are in parenthesis. Age/Period 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 15-19 2.2 (106) 2.9 (113) 2.7 (93) 3.4 (107) 3 (103) 3.1 (110) 20-24 5.8 (261) 7.2 (302) 7.5 (290) 7.7 (269) 6.2 (201) 8.3 (289) 5-29 5.1 (215) 7.4 (306) 7.4 (321) 7.7 (313) 8.2 (309) 7.2 (256) 30-34 4 (141) 4.8 (183) 6.1 (254) 6.6 (290) 6.3 (267) 6.2 (245) 35-39 2.6 (74) 2.7 (84) 3.6 (134) 3.8 (157) 3.8 (167) 3.9 (163) 40-44 1.9 (49) 1.5 (37) 1.7 (51) 2.2 (83) 2.2 (88) 2.9 (123) 45-49 1.6 (43) 1.1 (24) 0.8 (19) 1.5 (44) 1.5 (52) 1.6 (64) 50-54 0.8 (23) 0.7 (16) 0.7 (14) 0.7 (16) 0.9 (25) 0.9 (32) 55-59 0.6 (15) 0.4 (10) 0.5 (10) 0.4 (8) 0.4 (9) 0.3 (7) 60-64 0.5 (10) 0.3 (5) 0.2 (4) 0.3 (5) 0.3 (5) 0.3 (5) 21 Table 4 Age-Specific incidence of white seminomas rearranged in a longitudinal series by central year of birth. Birth Year Age 1903 1908 1913 1918 1923 1928 1933 1938 1943 1948 1953 1958 1963 1968 1973 1978 1983 15-19 0.2 0.3 0.2 0.3 0.4 0.6 20-24 1.7 2.2 2.7 2.2 2.8 3.6 25-29 4.8 4.9 6.2 7.2 7.7 7.3 30-34 4.6 6.6 8.2 9.8 9.6 10.2 35-39 5.1 5.9 7.9 8.8 9.9 10.7 40-44 3.7 5.2 5.9 7.5 8.0 9.0 45-49 3.3 4.6 4.2 4.6 5.5 5.7 50-54 2.8 2.6 2.8 2.9 3.1 3..3 55-59 1.8 1.8 2.0 2.0 1.6 1.9 60-64 1.7 1.6 1.2 1.9 1.4 1.3 65-69 0.8 1.2 1.1 1.1 1.5 1.5 70-74 0.9 1.1 0.7 0.7 0.7 0.6 Table 9 Age-Specific incidence of white nonseminomas rearranged in a longitudinal series by central year of birth. Birth Year Age 1913 1918 1923 1928 1933 1938 1943 1948 1953 1958 1963 1968 1973 1978 1983 15-19 2.2 2.9 2.7 3.4 3 3.1 20-24 5.8 7.2 7.5 7.7 6.2 8.3 25-29 5.1 7.4 7.4 7.7 8.2 7.2 30-34 4 4.8 6.1 6.6 6.3 6.2 35-39 2.6 2.7 3.6 3.8 3.8 3.9 40-44 1.9 1.5 1.7 2.2 2.2 2.9 45-49 1.6 1.1 0.8 1.5 1.5 1.6 50-54 0.8 0.7 0.7 0.7 0.9 0.9 55-59 0.6 0.4 0.5 0.4 0.4 0.3 60-64 0.5 0.3 0.2 0.3 0.3 0.3 22 An age-drift model was fit to the incidence of seminomas from 1973 to 2003 amongst white US men to demonstrate that temporal variations can be predicted by either period or cohort effects. Specifically, variables period and cohort were modeled as ordinal variables to illustrate that the predicted changes in incidence are identical. Table 5 helps depict the additive age effects from the two age-drift models and that the linear trend coefficients for both period and cohort are equal. Next, to examine the period and cohort effects, separate models were fit amongst white seminomas from 1973 to 2003 (Table 6). In this fashion, both a cross-sectional and longitudinal analysis may be performed to determine the behavior of this malignancy across all age groups and birth cohorts. In Table 6, the age effects are denoted by a a in both the period and cohort model. Period and cohort effects are denoted by B p and B c , respectively. In both models, we can see that the incidence of white seminomas peaks between the ages of 30-39 years. Furthermore, we can see that incidence is increasing for every 5-year period and for birth cohorts after 1938. An analysis of the goodness of fit of these models: age-only, age + drift, age + period, age + cohort, age-period-cohort, concludes that the model that fits the data for white seminomas the best is the full age- period-cohort model (Deviance=35.8, df=40) (Figure 5). 23 Table 5: Age effects in US white males among Seminomas adjusted for period (in period drift model) and for cohort (in cohort drift model). Both models have a deviance of 118.6 with 59degrees of freedom and an estimated drift of 0.12 for both period and birth cohort. Age Period Drift Model Cohort Drift Model 15-19 -13.05 -14.31 20-24 -11.01 -12.16 25-29 -10.09 -11.13 30-34 -9.83 -10.75 35-39 -9.84 -10.65 40-44 -10.05 -10.74 45-49 -10.42 -11.0 50-54 -10.89 -11.35 55-59 -11.35 -11.69 60-64 -11.57 -11.80 65-69 -11.85 -11.97 70-74 -12.30 -12.30 Figure 5: Suggested order of analysis for temporal data of white seminomas. Age Deviance=459.8, d.f.=60 Age + Drift Deviance=118.6, d.f.=59 Age +Cohort Deviance=54.4, d.f.=45 Age + Period Deviance=102.6, d.f.=55 Age+Period+Cohort Deviance=35.8, d.f.=40 24 Table 6: Age, period, and cohort effects in US white males 1973-2003 amongst Seminomas. Additive effects, deviances and degrees of freedom (d.f.) for the age-period and age-cohort models estimated using Poisson regression. Age+ Period Age+Cohort (deviance=102.6, d.f.=55) (deviance=54.4, d.f.= 45) Age a a Period B p Age a a Cohort B c 15-19 -13.00 1973-1978 0.00 15-19 -13.67 1899-1908 0.16 20-24 -10.96 1979-1983 0.19 20-24 -11.39 1904-1913 -0.01 25-29 -10.05 1984-1988 0.34 25-29 -10.33 1909-1918 0.1 30-34 -9.79 1989-1993 0.47 30-34 -9.96 1914-1923 0.01 35-39 -9.80 1994-1998 0.54 35-39 -9.87 1919-1928 -0.03 40-44 -10.00 1999-2003 0.61 40-44 -9.94 1924-1933 0.00 45-49 -10.38 45-49 -10.19 1929-1938 -0.001 50-54 -10.84 50-54 -10.55 1934-1943 0.03 55-59 -11.30 55-59 -10.95 1939-1948 0.12 60-64 -11.53 60-64 -11.17 1944-1953 0.37 65-69 -11.81 65-69 -11.44 1949-1958 0.49 70-74 -12.26 70-74 -11.89 1954-1963 0.66 1959-1968 0.75 1964-1973 0.78 1969-1978 0.83 1974-1983 1.17 1979-1988 1.67 3.3 Age-Period-Cohort Analysis 3.3.1 Seminomas The birth cohort effects from the age-period-cohort model are illustrated in Figure 6. The curve depicts a convex shape and is indicative of a change in the incidence of seminomas with regards to year of birth. Prior to 1943, the incidence of seminomas was decreasing with successive birth cohort, after this point there is a statistically significant change in the direction (P < 0.0001) resulting in an increase until 1963. After 1963, the rates start to drop and almost appear to be consistent until 1978 when they start to increase resulting in a change in slope with marginal significance (P=0.09). The period 25 effects illustrated in Figure 7 suggest that the risk over time due to calendar effects have been consistently increasing over time. Figure 6: Maximum likelihood estimates of 10- year birth cohort effects from an age- period-cohort model fit to seminoma incidence data amongst white men aged 15-74 years using SEER9, 1973-2003 -1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 1903 1908 1913 1918 1923 1928 1933 1938 1943 1948 1953 1958 1963 1968 1973 1978 birth year birth cohort effects 26 Figure 7: Maximum likelihood estimates of 5-year calendar period effects from an age- period-cohort model fit to seminoma incidence data amongst white men aged 15-74 years using SEER9, 1973-2003 A table describing the predicted number of cases from this age-period-cohort model exemplifies how well these data were fit (Table 7). Noted in this table are 6 observations with poor fit (|Pearson residuals| >2). These observations, ages 25-29, 40-44, and 65-69 for period 1973-1978, ages 45-49 and 70-74 for period 1979-1983 and age 20-24 for period 1989-1993, predicted higher or lower incidence rates than what was observed. 0 0.1 0.2 0.3 0.4 0.5 0.6 1976 1981 1986 1991 1996 2001 calendar year 27 Table 7: Predicted number of US white seminomas for 1973-2003. Pearson residual for each predicted value appears in parenthesis. Age/Period 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 15-19 10 (-0.19) 9 (1.02) 9 (-0.74) 9 (0.008) 13 (-0.10) 22 (.) 20-24 73 (0.28) 89 (0.22) 95 (1.18) 91 (-2.24) 87 (0.54) 127 (0.10) 25-29 180 (2.18) 217 (-1.37) 281 (-1.27) 292 (-0.15) 274 (1.30) 264 (-0.53) 30-34 170 (-0.84) 257 (-0.85) 327 (1.13) 417 (1.12) 424 (-1.27) 392 (0.47) 35-39 140 (0.58) 180 (-0.008) 289 (-0.17) 365 (-0.14) 444 (-0.51) 437 (0.43) 40-44 118 (-2.75) 123 (0.47) 170 (0.37) 272 (0.13) 319 (0.49) 379 (0.48) 45-49 98 (-1.26) 85 (2.00) 98 (0.05) 132 (0.05) 195 (-0.13) 228 (-0.49) 50-54 67 (1.92) 62 (-0.42) 58 (-0.05) 65 (-0.18) 83 (0.50) 124 (-1.22) 55-59 42 (0.18) 39 (-0.30) 39 (0.56) 37 (0.48) 39 (-0.94) 50 (0.06) 60-64 31 (0.72) 28 (0.69) 28 (-1.47) 28 (1.16) 25 (-0.18) 26 (-0.74) 65-69 16 (-2.02) 19 (-0.22) 19 (-0.83) 19 (-0.43) 17 (1.03) 15 (1.67) 70-74 9 (.) 8 (2.02) 10 (-0.69) 10 (-0.26) 9 (-0.46) 9 (-0.24) 28 3.3.2 Nonseminomas Figure 8 explains the goodness of fit of the different types of analysis performed on the data for white nonseminomas. Similar to seminomas, we see that the full age- period-cohort model is the best fit model. Tables 8 and 9 describe the number of cases for white nonseminomas cross-sectionally and longitudinally, respectively. The birth cohort effects across 14 distinct cohorts are illustrated for this histological subtype in Figure 9. The shape of this figure resembles a convex curve with the most drastic change in slope seen around 1938 (P=0.007). The incidence after this point increases at a faster rate per 5 year cohort than seen amongst white seminomas, These rates appear to plateau in 1968 (P=0.16) and decrease slightly in 1973 (P=0.06), but to start to increase shortly after. Figure 8: Suggested order of analysis for temporal data of white nonseminomas Age Deviance=164.8, d.f.=50 Age + Drift Deviance=95.5, d.f.=49 Age +Cohort Deviance=44, d.f.=36 Age + Period Deviance=73.1, d.f.=45 Age+Period+Cohort Deviance=29.6, d.f.=32 29 Figure 9: Maximum likelihood estimates of 10-year birth cohort effects from an age-period- cohort model fit to nonseminoma incidence data amongst white men aged 15-64 years using SEER9, 1973-2003. An examination of the period effects on this histological subtype reveals a concave curve suggesting a change in the frequency of exposure to environmental factors involved with this malignancy (Figure 10). -1 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0 1913 1918 1923 1928 1933 1938 1943 1948 1953 1958 1963 1968 1973 1978 birth year birth cohort effects 30 Figure 10: Maximum likelihood estimates of 5-year calendar period effects from an age- period-cohort model fit to nonseminomas incidence data amongst white men aged 15-64 using SEER9, 1973-2003 Worth mentioning is that there was one observation in this dataset that had a Pearson residual > 2. This observation created an over estimate of the true number of cases for age 15-19 at period 1989-1993 (Table 10). -0.08 -0.06 -0.04 -0.02 0 0.02 0.04 0.06 0.08 0.1 1976 1981 1986 1991 1996 2001 calendar year period effects 31 Table 10: Predicted number of US white nonseminomas for 1973-2003. Pearson residual for each predicted value appears in parenthesis. Age/Period 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 15-19 117 (-1.35) 108.5 (0.54) 100 (-0.88) 89.2 (2.4) 107.1 (-0.56) 110 (0) 20-24 263.1 (-0.20) 293.6 (0.75) 280.4 (0.86) 272.8 (-0.35) 217.2 (-1.74) 284.9 (0.56) 25-29 226.6(-1.17) 293.4 (1.14) 327.6 (-0.58) 325.9 (-1.14) 288.9 (1.86) 257.6 (-0.16) 30-34 138.5 (0.31) 190.3 (-0.76) 246.2 (0.73) 288.3 (0.15) 262.2 (0.44) 254.3 (-0.89) 35-39 64.6 (1.65) 87.5 (-0.49) 120.9 (1.54) 165 (-0.85) 172 (-0.51) 168.9 (-0.62) 40-44 43.9 (1.12) 40.8 (-0.75) 56.2 (-0.85) 82.3 (0.09) 97.4 (-1.20) 110.3 (1.59) 45-49 37.6 (1.43) 28 (-0.97) 27.3 (-1.88) 38.9 (0.98) 49.4 (0.45) 64.9 (-0.14) 50-54 22 (0.39) 21.2 (-1.5) 16.3 (-0.68) 16.4 (-0.11) 20.7 (1.10) 29.4 (0.60) 55-59 13.5 (0.84) 10.1 (-0.04) 10 (-1.47) 8.1 (-0.02)) 7.2 (0.74) 10.1 (-1.15) 60-64 10 (.) 6.5 (-0.84) 4.9 (-0.51) 5.2 (-0.09) 3.7 (0.76) 3.6 (0.78) 32 3.3.3 US Black Males As mentioned in the previous section, the analysis described above focused on white US males due to the rarity of TGCTs in black men. In Table 11, the incidence of TGCTs amongst black US males from 1973 to 2003 is illustrated by year of diagnosis for all age groups. The number of cases for consecutive calendar periods was 42, 44, 45, 46, 84, and 102, respectively. These numbers are insufficient to propose a formal age-period-cohort analysis to make inferences over time. However, referring to both Figure 4 and Table 3, the incidence of TGCTs has started to increase among black men in the past 10 years. These findings are parallel to those described by McGlynn et al. (2006), who suggest that the incidence of TGCTs among black men began to increase between the 1988 to1992 and 1993 to 1997 time periods. 33 Table 11: Age- specific incidence rates of US black TGCTs (per 100,000) from SEER9 for period 1973-2003. The number of cases for which rates are based are in parenthesis. Age/Period 1973-1978 1979-1983 1984-1988 1989-1993 1994-1998 1999-2003 0-4 0.6 (3) 0.2 (1) 0 (0) 0 (0) 0 (0) 0.3 (2) 5-9 0 (0) 0 (0) 0 (0) 0.2 (1) 0 (0) 0 (0) 10-14 0 (0) 0 (0) 0.2 (1) 0 (0) 0 (0) 0.1 (1) 15-19 0.2 (1) 0.2 (1) 0.7 (4) 0 (0) 0.7 (4) 0.9 (6) 20-24 0.8 (4) 2.0 (11) 1.4 (8) 1.2 (7) 1.3 (7) 2.5 (15) 25-29 2.2 (10) 2.4 (12) 1.4 (8) 1.4 (8) 2.5 (15) 2.0 (12) 30-34 2.0 (7) 2.3 (10) 1.0 (5) 1.7 (10) 2.4 (15) 2.8 (18) 35-39 1.5 (4) 1.0 (3) 2.6 (11) 1.5 (8) 3.5 (21) 3.1(20) 40-44 2.4 (6) 0 (0) 0.6 (2) 1.6 (7) 2.1 (11) 2.0 (12) 45-49 1.2 (3) 0.5 (1) 1.2 (3) 0.3 (1) 1.4 (6) 2.1 (11) 50-54 0.4 (1) 0 (0) 0.5 (1) 0.4 (1) 0.7 (2) 1.0 (4) 55-59 1.0 (2) 0.5 (1) 0 (0) 0.5 (1) 0.9 (2) 1.8 (5) 60-64 0 (0) 1.3 (2) 0.6 (1) 0 (0) 0.6 (1) 0 (0) 65-69 0 (0) 0.8 (1) 0.7 (1) 0.7 (1) 0 (0) 0 (0) 70-74 0 (0) 0 (0) 0 (0) 0.9 (1) 0 (0) 0 (0) 75-79 2.1 (1) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 80-84 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 0 (0) 85+ 0 (0) 6.2 (1) 0 (0) 0 (0) 0 (0) 0 (0) 34 Chapter 4: CONCLUSIONS Our results confirm that there has been a steady increase in the incidence of testicular germ cell tumors (TGCTs) among all men in the United States. The incidence of TGCTs rose 50% (95% CI: 40.2-59.8) between 1973 and 2003. The incidence of seminomas rose 65% (95% CI: 46.1-83.9) while that of nonseminomas rose 31% (95% CI: 11-51.5). When examining the patterns of seminomas and nonseminomas among white and black US males, identical patterns in incidence are noted however, more common among white males. The question as to whether or not the incidence of TGCTs has begun to subside is of medical interest and has been studied by several epidemiological studies. 29,37 Reviewing the SEER9 data, revealed that the incidence of TGCTs has increased 5.9% (95%CI: 0.3-11.5) between 1994 and 2003 among all men in the United States. Specifically, among white US males, TGCTs has increased 6.7% (95% CI: 1.9-11.4); seminomas increased a significant 8.3% (95%CI: 0.3-16.4) and nonseminomas increased 8.7% (95% CI: -3.9-21.3). Amongst black US males, an increase in incidence of 18.2% (95%CI:-9.4-45.8) was noted between 1994 and 2003. This finding may not be statistically significant, however since 1989 the incidence of TGCTs has begun to steadily increase in black US males. The results of this study indicate that seminomas are increasing at a faster rate than nonseminomas. These data and the results of our age-period-cohort analysis support that seminomas and nonseminomas are different manifestations of the same disease. This idea is reinforced by Oosterhuis et al (2005) who claims that the various types of germ 35 cell tumors are one common disease and reveal the potential of its developmental variability. This finding proposes a whole new question as to what are the associated risk factors of TGCTs and why seminomas are more common. There is no evidence to suggest that a possible change in diagnostic patterns occurred as of 1973 that might reflect the common occurrence of seminomas. 25 Several studies have reported histologic specific factors associated with each subtype. For instance, Akre et al. (1996) found that mothers were proned to higher estrogen levels during their pregnancy, due to such circumstances as increased maternal age, may be placing their sons at risk for seminomas. Similarly, an increase in placental weight or a decrease in parity is also known to increase estrogen levels and consequently elevates the risk for seminomas. Cryptorchidism is a well known risk factor of TGCTs in general, however, several studies have found higher risks for seminomas than nonseminomas. Fewer risk factors have been reported to be associated with nonseminomas. These finding may be due to the fact that nonseminomas are a heterogeneous mix of several histological subtypes that do not share common risk factors. Nevertheless, Akre et al. (1996) reported that intrauterine retardation i.e., low birth weight or decreased maternal age was associated with nonseminomas more so than seminomas. Other sources such as Stone et al. (1991) and Coupland et al. (1999) have reported a correlation between trauma and nonseminomas. Moss et al. (1986) reported a higher risk of nonseminomas was associated with early puberty. Another interesting observation that should be noted from this analysis is that although calendar period effects were assessed among seminomas and nonseminomas, the age-period-cohort analysis identified the birth cohort as the dominating effect on 36 disease incidence rates. This observation is consistent with other studies that expand on this. 3,6,20,25,27,43 As a direct consequence, this widespread correlation between birth cohort and TGCTs suggests that an early or a prolonged exposure to some carcinogenic stimuli might be required for the occurrence of testicular cancer. 27 Several studies that have focused on perinatal characteristics have repeatedly demonstrated the importance of in utero development and the onset of TGCTs. Exposure to exogenous steroid hormones while in utero is a perinatal factor known to be associated with the development of TGCTs 12 . Women were usually provided with estrogens and on occasion progestins to determine pregnancy, as supplements, or to avoid abortion. A cohort study demonstrated that men having been exposed to the estrogen analogue diethylstilbestrol were 2 to 3 times more likely to develop testicular cancer. 42 Higher levels of estrogen exposure are amongst firstborn sons compared to subsequent pregnancies. 4 Furthermore, high levels of free estradiol are released during pregnancy amongst hypertensive or obese women. 12 In conclusion, the incidence of testicular cancer does continue to increase, however the rate of increase is gradually decreasing. The occurrence of this disease is more prominent amongst white males than any other racial group however, in the last 10 years incidence amongst black males has started to increase. Future investigations would prove beneficial in determining why the incidence amongst black males is increasing, and what risk factors are shifting the risk towards seminomas to occur more often. 37 Chapter 5: Future Work The Los Angeles Cancer Surveillance Program (CSP), will be used to investigate the incidence of TGCT amongst Hispanic and non-Hispanic males and to determine how these findings correlated to those found in SEER9 amongst white and black males. This analysis will help estimate the risk of TGCT across racial groups with a special emphasis on histological subtypes in the last 30 years, 1973-2003. Incidence rates of TGCT by histological subtype and race for Los Angeles County will be obtained from the Los Angeles CSP, a population based cancer registry that was established in 1972 in order to collect and analyze all newly diagnosed cases of cancer within the county. Furthermore, it is a member of the statewide population-based cancer surveillance system, the California Cancer Registry, SEER, managed by the University of Southern California (USC) Keck School of Medicine as well as the USC/Norris Comprehensive Cancer Center. 38 REFERENCES 1) Akre O, Ekbom A, Hsieh C-C, Trichopoulos D, Adami H-O. Testicular nonseminomas and seminoma in relation to pre-natal characteristics. Journal of the National Cancer Institute. 1996; 88:883-889. 2) Adami H, Bergstrom R, Sparen P, et al. Increasing cancer risk in younger birth cohorts in Sweden. Lancet. 1993; 341:773-7. 3) Bergström R, Adami H, Möhner M, Zatonski W, Storm H, Ekbom A, Tretli S, Teppo l, Akre O, Hakulinen T. Increase in testicular cancer incidence in six European countries: a birth cohort phenomena. Journal of the National Cancer Institute. 1996; 88:727-733. 4) Bernstein L, Depue Rh, RossRk, et al. Higher maternal levels of free estradiol in first compared to second pregnancy: early gestational differences. Journal of the National Cancer Institute. 1986; 76(6):1035-9. 5) Bray F, Sankila R, Ferlay J, et al. Estimates of cancer incidence ant mortality in Europe in 1995. European Journal of Cancer. 2002; 38:99-166. 6) Bray F, Richiardi L, Ekbom A, Pukkala E, Cuninkova M, Moller H. Trends in testicular cancer and mortality in 22 European countries: Continuing increases in incidence and declines in mortality. International Journal of Cancer. 2006; 118:3099-3111. 7) Brown L, Pottern l, Hoover R, et al. Testicular cancer in the United States: Trends in the incidence and mortality. International Journal of Epidemiology. 1986; 15: 164-170. 8) Buetow SA. Epidemiology of testicular cancer. Epidemiology Review. 1995; 17:433-49. 9) Clayton D, Schifflers E. Models for temporal variation in cancer rates. I: Age- Period and Age-Cohort Models. Statistics in Medicine. 1987; 6:449-467. 10) Clayton D, Schifflers E. Models for temporal variation in cancer rates. II: Age-Period-Cohort Models. Statistics in Medicine. 1987; 6:469-481. 11) Cook R. Teaching and promoting testicular self-examination. Nursing Standard. 2000; 14(24):48-51. 12) Cortessis V. Epidemiologic insights into the occurrence and causes of testicular cancer. American Cancer Society Atlas of Clinical Oncology Germ Cell Tumors. Hamilton. BC Decker Inc. 2003; 16-29. 39 13) Coupland CAC, Chilvers CED, Davey G, Pike MC, Oliver RTD, Forman D. Risk factors for testicular germ cell tumours by histologic subtype. British Journal of Cancer. 1999; 80: 1859-1863. 14) Devesa SS, Blot WJ, Stone BJ, Miller BA, Tarone RE, Fraumeni JF Jr. Recent cancer trends in the United States. Journal of the National Cancer Institute 1995;87:175-82. 15) Dearnaley DP, Huddart RA, Horwich A. Regular Review: Managing testicular cancer. British Medical Journal. 2001; 322:1583-1588. 16) Gleason A. Racial disparities in testicular cancer: Impact on Health Promotion. Journal of Transcultural Nursing. 2006;17(1):58-64 17) Henderson B, Ross R, Yu M, Bernstein l. An explanation for the increasing incidence of testis cancer: decreasing age at first full-term pregnancy. Journal of the National Cancer Institute. 1997; 89:818-819. 18) Henley J, Young R, Ulbright T. Malignant sertoli cell tumors of the testis. A study of 13 examples of neoplasm frequently misinterpreted as seminoma. American Journal of Surgical Pathology. 2002; 26(5):541-550. 19) Hoel DG, Davis Dl, Miller Ab, et al. Trends in cancer mortality in 15 industrialized countries, 1969-1986. Journal of the National Cancer Institute. 1992; 84:313-20. 20) Huyghe E, Matsuda T, Thonneav P. Increasing incidence of testicular cancer worldwide: A Review. Journal of Urology. 2003; 170:5-11. 21) Jacobsen R, Bostofte E, Engholm G et al. Fertility and offspring sex ration of men who develop testicular cancer: a record linkage study. Human Reproduction Journal. 2000; 15:1958-61. 22) Klein FA, Melamed MR, Whitmore WF Jr. Intratubular malignant germ cells (carcinoma in situ) accompanying invasive testicular germ cell tumors. Journal of Urology. 1985; 133:413-5. 23) Landis S. H., Murray T., Bolden S, Wingo P.A.: Cancer statistics, 1998. Cancer Journal for Clinicians. 1998; 48:6. 24) McCullagh J, Lewis G. Testicular cancer:epidemiology, assessment and management. Nursing Standard.2005; 19(25):45-53. 25) McGlynn K, Devesa S, Sigurdson A, Brown L, Tsao L, Tarone R. Trends in the incidence of testicular germ cell tumors in the United States. American Cancer Society. 2003; 97:63-70. 40 26) McGlynn K, Devesa S, Graubard L, Castle P. Increasing incidence of testicular germ cell tumors among black men in the United States. Journal of Clinical Oncology. 2005; 23:5757-5761. 27) Mckiernan J, Goluboff E, Liberson G, Golden R, Fisch H. Rising risk of testicular cancer by birth cohort in the United States from 1973 to 1995. Journal of Urology. 1999; 162:361-363. 28) Miller MA, Sharpe RM. Environmental oestrogens and human reproductive cancers. Endocrine-Related Cancer. 1998; 5:69-96. 29) Moller H. Trends in incidence of testicular cancer and prostate cancer in Denmark.Hum Reprod. 2001;16:1007-1011. 30) Moss AR, Osmond D, Bacchetti P, Torti FM, Gurgin V. Hormonal risk factors in testicular cancer. American Journal of Epidemiology. 1986; 1224:39-52. 31) National Cancer Institute. SEER*Stat version 6.2.4: SEER cancer incidence public- use database, 1973-2003. Silver Spring: Journal of the National Cancer Institute, 2006. 32) Oosterhuis J, Looijenga L. Testicular germ-cell tumours in a broader perspective. Nature Publishing Group. 2005; 5:210-220. 33) Østerlind A. Diverging trends in incidence and mortality of testicular cancer in Denmark, 1943-1983. British Journal Cancer: 1986;53 34) Osterlind A, Berthelsen JG, Ablidgaard N, et al. Risk of bilateral testicular germ cell cancer in Denmark:1960-1984.Journal of the National Cancer Institute. 1991; 83(19):1391-5. 35) Percent Changes. http://www.census.gov/acs/www/Downloads/ACS/PercChg.pdf 36) Percy C, Van Holten V, editors. Conversion of neoplasms by topography and morphology from the International classification of diseases for oncology, second edition (IDC-O-2) to International Classification of Diseases for Oncology (ICD-O) (first edition and field trials). NIH publication no. 98-3430. Bethesda: National Institutes of Health, 1992. 37) Pharris-Ciurej ND, Cood LS, Weiss NS. Incidence of testicular cancer in the United States: has the epidemic begun to abate? American Journal of Epidemiology. 1999; 150: 45-46. 38) Pont J, Albrecht W. Fertility after chemotherapy for testicular germ cell cancer. Fertil Steril. 1997: 68:1-5. 41 39) Richie JP. Detection and treatment of testicular cancer. Cancer Journal for Clinicians. 1993; 43:151-175. 40) SAS Institute Inc. SAS version 9.1. Carry. 2002-2003. 41) SEER http://seer.cancer.gov/ 42) Strohsnitter WC, Noller KL, Hoover RN, et al. Cancer risk in men exposed in utero to diethylstilbestrol. Journal of the National Cancer Institute. 2001:93(7):545-51. 43) Stone JM, Cruickshank DG, Sandeman TF, Matthews JP. Laterality, maldescent, trauma, and other clinical factors in the epidemiology of testis cancer in Victoria, Australia. British Journal of Cancer. 1991; 64:132-138. 44) Spitz MR, Sider JG, Pollack ES, Lynch HK, Newll GR. Incidence and descriptive features of testicular cancer among Unites States whites, blacks, and Hispanics, 1973-1983. Cancer. 1986; 58:1785 45) Tarone R, Chu k. Evaluation of birth cohort patterns in population disease rates. American Journal of Epidemiology. 1996; 143:85-91. 46) Tarone RE, CHuKc. Implications of birth cohort patterns in interpreting trends in breast cancer rates. Journal of the National Cancer Institute. 1992; 84:1402-10. 42 APPENDIX A: MODEL DERIVATIONS Log-Linear Drift Model Let c = birth cohort p= calendar period A= number of total age intervals i= i th age interval, and j= j th calendar period interval We can express, c in terms of age and calendar period in the following relationship: c = A+ j i. (1) Using the relationship in (1), a log-linear age-period drift model may be expressed as a log-linear age-cohort model interchangeably as follows: Y ij = a i + B j (j-j o ) (2) = a i + B j (c + i A - c o + i o - A) = a i + B j (c + i A - c o A - i o ) = (a i + B j (i-i o )) + B j (c-c o ) = Y ic (3) where Y is the logarithmic age-specific disease rates, j o the reference period, c o the reference cohort, B j a constant change in the log rates noted between periods, and B c the constant change in the log rates between cohorts. Given a log-linear age-period drift model, we see that it can be expressed as an age-cohort model where B c = B j although, a distinct age relationship given the disease rates is noted. 43 Age-Period-Cohort (Non-drift) Model An age-period-cohort model with non-drift variables can be expressed in the following relationship: Y ij = a i + B j (j-j o )+ B c (c-c o ) (4). However, keeping in mind equation (1) we can also express equation (4) as follows: Y ij =a i + B j (i-i o )+(B c + B j )(c-c o ) (5) or Y ij =a i -B c (i-i o )+(B c + B j )(j-j o ) (6). From the illustrations above, we can see how these models are indistinguishable from each other so long as the net drift, B * =(B c + B j ) remains constant. Given this set of variables, it is simple to demonstrate why infinitely many maximum likelihood estimates of age, period, and birth cohort effect exist for any given data set. Due to the equality seen in i j + c = A, for any given constant, say λ, we have that λ(i j + c A)=0. Therefore, for any given λ and any constants x 1 , x 2 , and x 3 such that x 1 x 2 + x 3 =A, and any given set of a i s, B j s, and B c s a whole new set of parameters can be specified by: a i = a i + λ(i - x 1 ) B j = B j - λ(j x 2 ) B c = B c + λ(c x 3 ) which would create the same expected values that would be expected from either (5) or (6). This can be seen by the following calculation: a i + B j + B c = a i + λ(i - x 1 ) + B j - λ(j x 2 ) + B c + λ(c x 3 ) = a i + B j + B c + λ(i j + c (x 1 x 2 + x 3 ) ) 44 = a i + B j + B c + λ(i j + c A ) = a i + B j + B c . Measuring Birth Cohort Effects Suppose the linear trends of birth cohorts effects differ across distinct eras marked by indices c n and c m such that c m <c n . For those cohorts, c ≤ c m the linear trend of birth cohorts effects can be described by the following relationship: Y m = a m + B m c and for c ≥ c n Y n = a n + B n c. (Tarone et al., 1996) The difference in slopes B m B n may be described by the B c transformation previously discussed as follows: B c = a m + B m c + λ(c x 3 ) = a m λ x 3 + c(B m + λ) for c ≤ c m B c = a n + B n c + λ(c x 3 ) = θ n λ x 3 + c(B n + λ) for c ≥ c n . Under this transformation, the new slopes are B m = (B m + λ) and B n = (B n + λ). These parameter estimates are not unique in that they do not equal the original parameter estimates of B m and B n . However, differences between slopes to determine changes in magnitude or direction are unique independent of transformation. This can be seen in the following equality: B m B n = (B m + λ) (B n + λ) = B m B n.

Abstract (if available)

Abstract In order to address the question of whether or not the incidence of testicular germ cell tumors (TGCTs) has begun to abate, the incidence by histological subtype (seminoma and nonseminoma) among racial groups from 1973-2003 was examined on a national level. Incidence rates of TGCTs were obtained from the Surveillance, Epidemiology, and End Results Program and subjected to age-period-cohort analysis. -- Since 1973 to 2003, the incidence of all TGCTs has risen 50% (95% CI: 40.2 - 59.8). The overall incidence of TGCTs among white men has been consistently increasing in the US since 1973 with black males showing a recent increase in the last 10 years. Analysis of the incidence of TGCTs amongst white men revealed that birth cohort was the dominating risk factor for both seminoma and nonseminoma subtypes.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Bilateral testicular germ cell tumors: distribution of histologic types

PDF

A comparison of three different sources of data in assessing the adolescent and young adults cancer survivors

PDF

The association of prediagnostic metformin use with prostate cancer in the multiethnic cohort study

Asset Metadata

Creator Rodriguez, Janelle (author)

Core Title Trends of testicular cancer within the United States by histological subtype amongst whites and blacks, 1973-2003

School Keck School of Medicine

Degree Master of Science

Degree Program Biostatistics

Publication Date 05/08/2007

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag OAI-PMH Harvest,testicular cancer

Place Name USA (countries)

Language English

Advisor Siegmund, Kimberly (committee chair), Cortessis, Victoria Kristence (committee member), Hamilton, Ann S. (committee member)

Creator Email janeller@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-m486

Unique identifier UC1274653

Identifier etd-Rodriguez-20070508 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-484810 (legacy record id),usctheses-m486 (legacy record id)

Legacy Identifier etd-Rodriguez-20070508.pdf

Dmrecord 484810

Document Type Thesis

Rights Rodriguez, Janelle

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Repository Name Libraries, University of Southern California

Repository Location Los Angeles, California

Repository Email cisadmin@lib.usc.edu