Transfusion History and Cancer Risk in Older Women

Methods

The Iowa Women's Health Study Cohort

Participants in the Iowa Women's Health Study were a random sample of all women ages 55 to 69 years derived from the state of Iowa automobile driver's license list in 1985, which represented approximately 94% of Iowa women in that age group [50]. A total of 41 837 women returned a questionnaire mailed to them in January 1986, which represented a 42.7% response rate. A comparison between responders and nonresponders has been published elsewhere and showed only minor demographic differences by response status [51]. The 5-year cancer rates for major cancer sites in the Iowa Women's Health Study cohort is close to the 1988 rate for cancer in all Iowa women ages 55 to 69 years (Cerhan and Folsom. Unpublished data available on request).

To specifically address the transfusion and cancer risk hypothesis, the questionnaire included an item asking Have you ever received blood or had a blood transfusion? Women who responded yes were also asked their age at first transfusion and the reason for the transfusion. To identify the prevalence of persons with cancer, participants were asked if they had ever been diagnosed by a physician with any form of cancer other than skin cancer. The validity of responses to this cancer question compared with the physician report was found to be good [51].

Follow-up of the Cohort

Vital status and cancer incidence in this cohort were ascertained through 5 years of follow-up (1986 to 1990). Follow-up questionnaires were mailed in October 1987 and August 1989 to assess vital status and address changes. Deaths were ascertained using Iowa death certificates and using the National Death Index. Nonrespondents to the follow-up questionnaires were traced using the National Change of Address files. Through 1990, 1535 deaths were ascertained; vital status was unknown for less than 1% of the cohort.

Incident cancers, except for nonmelanoma skin cancers, were ascertained by the State Health Registry of Iowa, part of the National Cancer Institute's Surveillance, Epidemiology, and End Results program [52]. Field representatives routinely visited hospitals and clinics in and around Iowa. For participants with cancer who were Iowa residents at the time of diagnosis, information including identifiers, demographic data, date of diagnosis, primary site, histologic findings, and extent of disease were recorded. Topographic and morphologic data were coded by the International Classification of Diseases for Oncology. The Iowa Women's Health Study cohort was matched to the registry with combinations of first, last, and maiden names, zip code, birthdate, and social security number.

Data Analysis

Women with the following baseline characteristics (n = 4438) were excluded: a history of any cancer other than skin cancer, a history of mastectomy, a history of cancer chemotherapy, and missing data for all of the transfusion questions. Women with a hysterectomy or oophorectomy (unilateral or bilateral) were excluded from the analyses for uterine and ovarian cancer, respectively. In addition, we excluded women who had a previous cancer documented in the registry (n = 38) and excluded those for whom we were unable to calculate person-years of follow-up (n = 24, see below). The final data set for analysis contained 37 337 women.

Because only cancers of Iowa residents were detectable, each woman was allocated person-years of follow-up from the date of the 1986 baseline questionnaire to one of the following events: 1) date of cancer diagnosis; 2) date of death, if the death occurred in Iowa; 3) date of a move out of Iowa; 4) the midpoint between the date of last contact and date located outside of Iowa [date of move from Iowa not known]; 5) midpoint between date of last contact and date of death, if the death did not occur in Iowa. If none of these occurred, follow-up was through December 1990.

Women were categorized according to their transfusion status as reported at baseline (no, yes, do not know) and according to the time from first transfusion to baseline (<5 years, 5 to 14 years, 15 to 29 years, and 30 or more years). The time categories were chosen before analysis and were designed to include general induction times for various cancers. Potential confounders measured at baseline were also categorized for analysis (see Table 1 for categories) and included age at baseline survey, education, residence, smoking status, number of pregnancies, and body mass index (kg/m²).

Incidence rates were calculated by dividing the number of events by the number of person-years of follow-up. Relative risks and their 95% confidence intervals were computed for the categories of the two transfusion history questions, after adjustment for age [53]. For analysis of specific cancers, person-years were calculated excluding persons with all other cancers. This was done because we hypothesized that transfusion might be related to cancer at more than one site. However, re-analysis of the lymphoma data with the person-years for other cancers included did not alter the relative risk estimates, and therefore they are not reported here. Poisson regression [54] was used to estimate the multivariate relative risk, after controlling for age, education, residence, smoking (only for the lung, breast, pancreas, kidney, and bladder cancer analyses), obesity (only for the breast, endometrial, colon, and kidney cancer analyses), and number of pregnancies (only for the breast, endometrial, ovarian, and colon cancer analyses). The multivariate relative risks for each cancer type were nearly identical to the age-adjusted relative risks listed in Table 2, and therefore only the age-adjusted results are reported. Analyses were done using SAS (SAS Institute, Cary, North Carolina) or EGRET (SERC, Seattle, Washington) software programs.

Results

Approximately 25% of the women (n = 9539) reported at baseline ever having received a blood transfusion (see Table 1). The mean age at first transfusion was 37.3 years [SD = 12.1; range, <1 to 69 years]. The mean time from first transfusion to the baseline survey was 24.3 years (SD = 11.6; range, <1 to 64 years). The most common reasons given for a transfusion were for blood loss from surgery or bleeding (92%) or from decreased blood production due to an illness (5%). Only a small percentage did not know why they had received a transfusion (1%) or gave no reason (1%). Age, education, residence, smoking history, and obesity did not vary by transfusion status (see Table 1); however, women with a history of transfusion had slightly more pregnancies [mean = 4.3 pregnancies compared with 3.8 pregnancies for women who never had a transfusion].

During 174 441 person-years of follow-up of the 37 337 women during a 5-year period, there were 1816 incident cancers. The age-adjusted relative risks for various cancers according to transfusion status are reported in Table 2. Women who did not know their transfusion status were included as a separate category, but their results were uninformative, and therefore they are not reported in Table 2. Compared with women with no history of transfusion, women with a history of transfusion were not at an increased risk for all cancers considered together. However, women with a history of transfusion had an increased risk for non-Hodgkin lymphoma and kidney cancer. In contrast, cancers of the soft tissues, the reproductive organs, the respiratory tract, and the digestive tract were not associated with transfusion history. If the associations between transfusion history and cancer risk found here were causal, then the attributable risk of non-Hodgkin lymphoma and kidney cancer in transfused women due to their having received a transfusion would be 30.5/100 000 per year and 25.1/100 000 per year, respectively.

Of the 68 non-Hodgkin lymphomas (66 in Table 2 plus 2 with unknown transfusion status), 35% were low grade, 50% were intermediate grade, 3% were high grade, and 12% were reported as only diffuse, not otherwise specified. The most commonly reported morphologic type was diffuse large cell (31%). No women had Burkitt lymphoma, and only four women had Hodgkin disease. Thirty percent of the lymphomas were extranodal, and they included the small intestine (n = 4), stomach (n = 3), soft tissues (n = 3), brain (n = 1), and miscellaneous other sites. Intermediate-grade lymphomas were the most common grade in the nontransfused group (71%), and there were no high-grade lymphomas. In the transfused group, 7% (n = 2) of the lymphomas were high grade and 51% were intermediate grade. No specific extranodal site dominated in either the transfused or nontransfused women. When analysis of non-Hodgkin lymphoma was restricted to only the lymph nodes as the primary site, the relative risk for women with a history of transfusion increased to 2.92 (CI, 1.69 to 5.05).

There were 38 carcinomas of the kidney, of which 79% were renal cell, 18% were clear cell, and 3% were transitional cell. No pattern existed between type of kidney cancer and transfusion history.

Women with a history of transfusion were categorized by the time from their first transfusion until the baseline survey (Table 3); results are presented only for those cancers that showed any transfusion-associated risk (P < 0.05). For non-Hodgkin lymphoma, the trends in the relative risks had a similar pattern regardless of whether the extranodal sites were included, although the trends were of greater magnitude when the extranodal sites were excluded. The risk was greatest for transfusions within 5 years of baseline, and the risk decreased with time. Kidney cancer showed a similar pattern to non-Hodgkin lymphoma, although no cases of kidney cancer occurred in women reporting transfusion 5 years before baseline. Leukemias, multiple myeloma, and bladder cancer, which were not related to a history of blood transfusion Table 2[P < 0.05]), showed increased risks at various times before transfusion.

Discussion

Transfusions and Risk for Certain Cancers

A history of transfusion was not a risk factor for the most common types of cancers. However, our data suggest an increased risk for certain types of cancers. The increased risk for non-Hodgkin lymphoma in women who were transfused is consistent with the increased risk for this malignancy found in patients from transplant studies [44, 45, 47], patients receiving immunosuppressive therapy [46, 48], and patients with AIDS [49]. However, our associations were much smaller than the 30-to 70-fold increased risk found in those studies and may reflect misclassification of our exposure variable or, more likely, a weaker degree of immune suppression. Indeed, this difference between a healthy general cohort and groups with specific known disorders is not surprising. In studies of immunosuppressed patients, the most common morphologic findings have included not only intermediate-grade, large-cell diffuse lymphoma but also high-grade immunoblastic lymphoma, and the most common extranodal site has been the brain [44, 45]. Only 7% of the lymphomas in the transfused group were high grade in our study, one of which was a lymphoma of the central nervous system. Alternatively, other factors, especially virus transmission, might also explain our findings. However, our findings are unlikely to be related to the AIDS epidemic because only 12 documented patients have had human immunodeficiency virus infection that was transfusion associated in the state of Iowa since 1982 [55]. The suggestive association between transfusion history and leukemia and, to a lesser degree, transfusion history and bladder cancer that we observed is consistent with data reported for transplant patients [44] and patients receiving immunosuppressive drugs [46, 48], although these associations are less well established compared with the associations found for lymphomas. In addition, the increased risk for bladder cancer has previously been attributed to the local effects of treatment with cyclophosphamide [46]. An increased risk for multiple myeloma with immunosuppression has not been reported before, although the number of patients with this malignancy was small in our study. Finally, a slightly increased risk for renal cancer has been reported [45] in transplant patients, with most of the malignancies occurring in the host kidney. Other kidney transplant registries have not reported this increase [44, 47], but given the rarity of this neoplasm and the small number of patients involved in the transplant registries, the statistical power to find differences would be limited.

Based on previous, although less consistent, findings in immunosuppressed persons, we might also expect to find an increased risk for nonmelanoma skin cancer, melanoma, Kaposi sarcoma, adenocarcinoma of the lung, and hepatobiliary carcinoma. However, the Surveillance, Epidemiology, and End Results program does not collect data on nonmelanoma skin cancer, and there were too few patients with Kaposi sarcoma (n = 1) and hepatocellular carcinoma (n = 4) for analysis. Melanomas and adenocarcinomas of the lung showed no relation to transfusion status. Although we tested many cancer sites, we believed that the suggestive associations found for cancer occurrence in immunosuppressed persons and the increased risk for cancer recurrence in cancer surgery patients transfused during operation justified testing multiple sites. However, a conservative adjustment of the P value by the method of Bonferroni [56] to take into account the 13 specific sites that we tested in Table 2 would decrease the traditional level of significance from P = 0.05 to P = 0.004. The P value for the relative risk (see Table 2) for non-Hodgkin lymphoma was 0.001 and for kidney cancer it was 0.004.

A potential problem exists for interpreting the cancer risk for women first transfused less than 5 years before baseline because they may have required a transfusion due to an occult cancer. We chose not to exclude this group because the transplantation data has shown a short latency period for many of the tumors. A wide range of times occurred between the approximate time of transfusion and the date of cancer diagnosis. The four non-Hodgkin lymphomas were diagnosed about 2.3, 5.1, 5.4, and 8.6 years after the first transfusion; the three leukemias were diagnosed after 5.0, 5.1, and 7.6 years; and the two multiple myelomas were diagnosed after 3.2 and 4.5 years. On this basis, it seems less likely that the affected women had developed such extensive disease that they required a transfusion so long before diagnosis. However, this hypothesis cannot be excluded. In addition, it is possible that the transfusion-cancer association is confounded by the disease for which the transfusion was given, that is an earlier disease, and its treatment may, in fact, be the risk factor for the subsequent cancer. We do not have the necessary data in our study with which to assess this possibility.

Previous Studies

We could find no reports on whether blood transfusion itself is associated with increased cancer risk. It was originally reported that transfused renal transplant patients were at an increased risk for tumors compared with recipients without transfusions [57], but a subsequent analysis showed no difference [58]. Opelz [59] reported in a letter that no increased risk for tumors occurred in the Denver Kidney Transplant-Tumor Registry for transfused patients compared with nontransfused controls; however, no data were provided. Finally, Tartter [60] reported that patients having colorectal cancer surgery were less likely to have had a transfusion compared with an age-and gender-matched control group of patients having hernia or gallbladder operations (9% compared with 17%, P < 0.01). However, the author [60] noted that transfusion may itself be a risk factor for gallstones.

Before discussing interpretations of these findings, the exposure variable requires comment. History of blood transfusion was self-reported and we did not collect other information on the nature of the transfusion. However, the August 1989 follow-up questionnaire asked each participant if she had received any of the following since February 1, 1986: 1) a blood transfusion; 2) other blood products such as plasma or globulin; or 3) intravenous fluids. During the 3.5-year interval, 2.9% of respondents reported receiving a blood transfusion, 1.3% reported receiving other blood products, and 20.2% reported receiving other fluids intravenously; only 0.5% reported receiving all three. Thus these women did make a distinction between blood transfusion, transfusion of blood products, and the use of intravenous fluids.

The annual rate between 1986 and 1989 for receiving a blood transfusion was approximately 8.3 women per 1000 (0.029/3.5 years) in our cohort compared with 14.3 patients per 1000 persons ages 18 to 65 years in the United States in 1980 [61]. The mean units of erythrocytes transfused was 3.1 in the U.S. survey. The corresponding rates for Iowa men and women ages 18 to 65 years in 1980 were 12.8 patients transfused/1000 persons, and a mean of 3.1 units of erythrocytes transfused/patient. A different study, the AIDS Knowledge and Attitudes questionnaire of the National Health Interview Survey, has consistently reported that 7% to 8% of women older than age 50 have received at least one blood transfusion between 1977 and 1985 [62, 63], which is an annual rate of 9.4/1000 persons (0.075/8 years). Thus, our rate of transfusion is compatible with other national data. A lifetime prevalence of 250/1000 persons for having had at least one transfusion for women ages 55 to 69 years in our study also appears consistent with these transfusion rates, although this is difficult to estimate.

Limitations of Study

We do not know the reliability of recall for having received a transfusion, particularly for transfusions that took place several decades ago. However, misclassification of baseline transfusion status (that is, classifying a women as having received a transfusion when in fact no transfusion took place and vice versa) is not likely to be related to future cancer status and thus should not create artificial associations. Indeed, misclassification is much more likely to attenuate observed risk estimates. Misclassification of transfusion status during follow-up also occurred. From the August 1989 questionnaire, we know that 492 women with no transfusion history at baseline subsequently received a transfusion and 83 of these women also developed cancer. However, it is not known when they received their transfusions and, for patients with cancer, whether they received their transfusions before or after diagnosis. Exclusion of the misclassified cancers and person-years slightly increased the estimates of the relative risks in Table 2.

Two other limitations of the transfusion exposure history included not knowing what type of blood was transfused (packed erythrocytes, whole blood, or frozen erythrocytes) and whether the blood transfusion was autologous or homologous. These questions can only be partially addressed by reference to blood surveys conducted in the United States between 1971 and 1988 [61, 64]. Transfusion of erythrocytes and whole blood increased about 60% between 1971 and 1980 and continued to increase until reaching a peak of 12.2 million units in 1986, and then this transfusion rate declined through 1988. The decline was attributed to the AIDS epidemic, which led to unprecedented changes in transfusion practices [64]. Similar patterns were observed for transfusion of platelets and plasma. Between 1971 and 1980, a major reduction occurred in the use of whole blood (from 78% of transfusions involving erythrocytes in 1971 to 19% in 1980), while the transfusion of separated erythrocytes increased six times [61]. Frozen blood contributed only a small percentage (<3%) of the total erythrocytes transfused. Storage of autologous blood increased more than 13 times from 1982 to 1987, although the 1987 figure represented only 3% of the homologous blood collections [64].

The increased use of packed erythrocytes during the 1970s is of interest. Several studies have implicated the leukocyte fractions as a cause of immune suppression [32, 65], which is consistent with the observation that the immune suppression in renal transplant patients is greatest with packed erythrocytes (which also generally contain the leukocyte fraction), less with whole blood, and least with frozen blood [33]. This increased use of packed erythrocytes coincides with a 42% increase in the incidence of non-Hodgkin lymphoma recorded by the Surveillance, Epidemiology, and End Results program registry in white women between 1973 and 1987 [66]. This increase occurred mainly in older women, a group that was not at high risk for human immunodeficiency virus infection. Because the increase seems to be greater than expected due to improved medical technology in diagnosis, intense speculation exists as to the cause of the increase, although no definitive explanation has been provided [67]. In the same time frame, kidney cancer increased 33%, bladder cancer increased 11%, and multiple myeloma increased 7%.

Cause of Transfusion-associated Carcinogenesis

We can only speculate about the mechanisms by which transfusions may be related to carcinogenesis. A viral agent transfused into the patient is an attractive hypothesis and is consistent with data on the Epstein-Barr virus and Burkitt lymphoma link as well as the more recent reports of new retroviruses associated with lymphomas in AIDS patients [68]. However, a recent study [49] of AIDS-associated non-Hodgkin lymphoma concluded that, although infectious agents probably play a part in the development of lymphomas, they could find no single or simple explanation for their role, and the authors concluded that immunosuppression was the main determinant of risk for the malignancies. Although the immunomodulating effects of blood transfusions may play a role in the carcinogenic process, it is not possible to rule out viruses themselves as the cause of immune suppression after transfusion. However, virus as the sole cause of immune suppression seems unlikely because immune suppression after transfusion has been replicated in several animal models [34, 69, 70].

The role of the immune system in the development of cancer is controversial and still poorly understood. Leukemias and lymphomas tend to have antigens that are homogeneous and highly antigenic, and thus a role for the immune system in their development is plausible. In contrast, the more common solid tumors of the internal organs tend to have antigens that are heterogeneous and of variable immunogenicity, so a direct role for the immune system in these cancers seems less likely, although there may be indirect influences. De novo cancers that arise in immunosuppressed persons were originally considered to be strong evidence for the immune surveillance theory as proposed by Thomas [71] and Burnet [72]. However, most of these tumors have been malignancies of the immune system itself, and little evidence exists of an increase in risk for other tumors. An alternative interpretation of these findings is that immune suppression selects for autonomous cells in the immune system and allows them to flourish, whereas the remaining cells are suppressed [73].

The viral and immune suppression hypotheses are not mutually exclusive; they could work together as cofactors in the carcinogenic process. This synergism has been advanced to explain the strong link between Epstein-Barr virus and Burkitt lymphoma in tropical areas, with the cofactor being the high prevalence of malaria [74]. Unknown genetic mechanisms could be involved, such that only susceptible genotypes are at increased cancer risk from a transfusion. None of these mechanisms, however, offers an immediately satisfactory explanation for the increased risk for kidney or bladder cancer.

Conclusion

Our findings, if replicated, would have important implications for clinical medicine, public health, and cancer biology. Blood transfusions provide both benefits and risks, and we have discussed only one potential long-term risk. As with all medical interventions, the physician must weigh current knowledge of risks and benefits for a given patient. Further study of the effect of transfusions on the immune system and cancer risk may lead to new insights into the carcinogenic process and may ultimately offer novel preventive measures.