SEARCH

SEARCH BY CITATION

Keywords:

  • non-Hodgkin lymphoma;
  • prognosis;
  • smoking;
  • social class;
  • autoimmune diseases

Abstract

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. References
  8. Supporting Information

Lifestyle factors and medical history are known to influence risk of non-Hodgkin lymphoma (NHL). Whether these factors affect the prognosis of NHL, especially its subtypes, is unclear. To investigate this, the association between these factors and all-cause and lymphoma-related mortality was assessed in a population-based cohort of 1,523 Swedish NHL patients included in the Scandinavian Lymphoma Etiology study in 1999–2002. Participants contributed time from NHL diagnosis until death or October 1, 2010, with virtually complete follow-up through linkage to the Swedish Cause of Death Register. Hazard ratios (HR) and 95% confidence intervals (CI) were estimated using stratified and multivariable-adjusted Cox regression models. During a median follow-up of 8.8 years, 670 patients (44%) died, with the majority of deaths attributed to lymphoma (86%). Current versus never smoking at diagnosis was associated with increased rate of all-cause death for all NHL (HR = 1.5, 1.2–1.8) and diffuse large B-cell lymphoma (HR = 1.8, 1.2–2.7). Low educational level (HR = 1.3, 1.1–1.7, <9 vs. >12 years) and NHL risk-associated autoimmune disease (HR = 1.4, 1.0–1.8) were associated with death for all NHL combined. However, evidence of an association with lymphoma-related death was limited. Body mass index, recent sunbathing and family history of hematopoietic malignancy were not consistently associated with death after NHL or its specific subtypes. These results add to the evidence that cigarette smoking, socioeconomic status and certain autoimmune diseases affect survival after NHL. Further investigations are needed to determine how these factors should be incorporated into clinical prognostic assessment.

The prognosis for non-Hodgkin lymphoma (NHL) patients has improved in the past decade, especially since the introduction of rituximab.1 Nevertheless, NHL accounted for approximately 4% of all cancer-related deaths in Europe and in the United States in 2008.2, 3 The International Prognostic Index (IPI), first described in 1993, and the Follicular Lymphoma International Prognostic Index (FLIPI), described in 2004, are the primary measures for assessing prognosis for diffuse large B-cell and follicular lymphoma, respectively.4 However, sharper tools are needed to predict outcome and improve individualized treatment. In the search for additional prognostic determinants, factors such as medical history, lifestyle factors5–8 and inherited genetic variation9 are explored with increasing interest.

Information on lifestyle factors and medical history may provide simple and readily available complementary tools for prognostic prediction. Cigarette smoking, alcohol intake, diet and socioeconomic status have been proposed to affect NHL survival, but the results are inconsistent, and evaluation of subtype-specific variation has been limited.6, 7, 10–15 Although we and others have shown an inverse association between ultraviolet (UV) radiation and the risk of NHL,16–19 whether UV radiation influences prognosis is unknown. Along the same lines, family history of hematopoietic malignancy and autoimmune disease history have been associated with increased lymphoma risk,20, 21 but not consistently with prognosis.5, 8, 22–26

These inconclusive answers to intriguing questions about potential NHL prognostic factors gave us cause to investigate whether family history, medical history and lifestyle factors were associated with overall or lymphoma-specific mortality in a population-based cohort of 1,523 incident NHL patients diagnosed in Sweden in 1999–2002, mainly in the prerituximab era.

Material and Methods

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. References
  8. Supporting Information

Study population

Swedish NHL cases from the Scandinavian Lymphoma Etiology (SCALE) study were included in the present analysis. SCALE is a population-based case–control study of malignant lymphoma in Sweden and Denmark and has been described in detail previously.19 Cases with incident and morphologically verified NHL and chronic lymphocytic leukemia according to the International Classification of Diseases (ICD 7 codes 200, 202 and 204.1) were recruited between October 1, 1999, and April 15, 2002, and were 18–74 years old at diagnosis. Multiple myeloma was not included. The patients were identified through a nationwide hospital-based network of physicians and pathologists, with back-up through regional population-based cancer registries based on ICD codes. Upon inclusion, tumor material was reviewed and reclassified according to the 2001 World Health Organization (WHO) classification of hematopoietic malignancies.27 Individuals with a medical history of hematopoietic malignancy, organ transplantation, or HIV infection and those who did not speak Danish or Swedish were excluded. The Danish component of SCALE was excluded from our analysis owing to limited information on follow-up, as were cases with chronic lymphocytic leukemia (CLL) because of a lack of information on CLL-specific prognostic factors. The participation rates of eligible cases in Sweden, excluding CLL, were 84% for NHL overall and diffuse large B-cell lymphoma (DLBCL), and 94% for FL. Early death was the most common cause of nonparticipation among all NHL and DLBCL cases (6% of eligible subjects in each group), but was an uncommon cause of nonparticipation among FL cases (1% of all eligible FL cases). In our analysis, NHL subtypes with more than 90 cases were considered separately. The group of T-cell NHL consisted mostly of peripheral T-cell lymphoma types (73%) but also included mycosis fungoides (12%) and other rare subtypes. The study was approved by the Ethical Review Board in Stockholm and all study subjects gave informed consent.

Prognostic factors

Data on pre-existing autoimmune disease, lifestyle and other factors of interest were collected by telephone interview with a standardized, computer-aided questionnaire. Cigarette smoking history included ever smoking (defined as daily tobacco smoking for at least 1 year), age at start and stop and average number of cigarettes smoked per day. Individuals were classified as current smokers if they smoked at NHL diagnosis or had stopped <1 year before, and past smokers if they had stopped at least 1 year prior to the NHL diagnosis. We further categorized smoking intensity (1–9, 10–19, ≥20 cigarettes per day) and smoking duration (1–19, 20–39 and ≥40 years) among ever-smokers. Cigarette-years (none, <200, 200 to <400 and ≥400) were calculated by multiplying smoking intensity by duration. Body mass index (BMI) was categorized as <25, 25 to <30 or ≥30 kg/m2 and educational attainment (as a proxy for socioeconomic status) as ≤9, 10–12 or >12 years. UV radiation exposure was determined by assessing past and recent recreational sun exposure, as well as skin sensitivity to sun and exposure to artificial UV radiation.19 For our analysis, we focused on summer sunbathing frequency in the past 5–10 years, defined as never, ≤1, 2–3 or ≥4 times per week. The remaining UV exposure variables were considered in sensitivity analyses. A composite variable summarizing history of autoimmune disease was defined based on the self-report of physician-diagnosed rheumatoid arthritis (RA) (n = 64, 64% of autoimmune diseases comprising the composite variable), celiac disease (n = 20, 20%), primary Sjögren's syndrome (n = 10, 10%) or systemic lupus erythematosus (SLE) (n = 7, 7%); these four diagnoses represent autoimmune disorders consistently associated with increased NHL risk.22, 28 Validated information on occurrence of hematopoietic malignancies in first-degree relatives was obtained through linkage with the nationwide Multi-generation Register and Swedish Cancer Register, as described in detail elsewhere.20 Information on primary extranodal disease and Ann Arbor stage of nodal disease at diagnosis was obtained for 89% of the individuals from all six regional lymphoma quality registers contributing data to the national Swedish lymphoma quality register.29

Follow-up

Using the unique national registration number assigned to all Swedish residents, we linked all case-patients to the Swedish Cause of Death Register to obtain virtually complete follow-up from the interview date through October 1, 2010, or the date of death, whichever came first. Lymphoma-related death was defined as having lymphoma as the main underlying or contributory cause of death in the Cause of Death Register.30

Statistical analysis

The distributions of demographic, potential prognostic and NHL-specific characteristics were summarized for overall NHL and its four most common subtypes (DLBCL, follicular, mantle cell and T-cell lymphomas) and compared across subtypes with Chi-square tests or Fisher's exact tests as appropriate. The survival probability by smoking and education was assessed for all NHL and DLBCL cases using the Kaplan–Meier estimator. We used Cox proportional hazards models to estimate hazard ratios (HR) of all-cause mortality for potential prognostic factors of interest for all NHL and by subtype, with time since interview (in days) as the time scale. First, survival was modeled as a function of spread (primary extranodal vs. not) and Ann Arbor stage (1–2 vs. 3–4) separately adjusting for age at diagnosis (<55, 55–64 or 65+ years) and sex. The remaining models were stratified by age and stage (primary extranodal disease or nodal disease at Ann Arbor stages 1–2 or 3–4) at diagnosis and adjusted for male sex, cigarette smoking, years of education and BMI, because they were hypothesized a priori to be important prognostic factors and potential confounders. Models were further adjusted individually for recent sunbathing, autoimmune disease and family history of hematopoietic malignancy. The proportional hazards assumption was assessed by using the Wald test for statistical interactions with time included in the model; no violations were found. Missing values were included in the models using the missing indicator method as well as using imputation if missingness appeared to be informative by the first method. Lymphoma-related mortality was investigated as a secondary endpoint using Cox models in which deaths from a nonlymphoma-related cause were censored at the date of death. As the censoring of a competing event may introduce bias akin to informative censoring, we defined nonlymphoma death as the competing risk and reran the multivariable-adjusted models using competing risks regression31 in sensitivity analyses. This alternative approach allowed us to estimate the sub-HR and corresponding 95% confidence intervals; robust standard errors were the default. A p-value for linear trend was calculated treating the exposure variables as ordinal instead of categorical in regression models. The significance level was set at <0.05 and all tests were two-sided. SAS 9.2 and STATA 11 were used for the analyses.

Results

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. References
  8. Supporting Information

Distribution of patient characteristics and deaths

There were 1,523 NHL cases with a median age of 60 years (mean, 58.6 years) at diagnosis, and 59% were male. The most common NHL subtypes were DLBCL (n = 513, 34%) and follicular lymphoma (n = 364, 24%). We found no significant differences in education, BMI, smoking status, recent sunbathing and family history of hematopoietic malignancy by NHL subtype; however, there were significant differences in sex, tumor stage and autoimmune disease by subtype (Table 1). During a median follow-up of 8.5 years (total 10,261 person-years), 44% of the study population (n = 676) died. Approximately 86% (n = 573) of the observed deaths were classified as lymphoma related (Supporting Information Table 1). Other than lymphoma, the leading causes of death were cancer (n = 28), ischemic heart disease (n = 8) and stroke (n = 8). In 25 patients (4% of all deaths) the cause of death was missing (data not shown).

Table 1. Distribution of potential prognostic factors by NHL subtype, presented as count (%)
inline image

Associations with all-cause and lymphoma-related mortality

As expected, disseminated disease at diagnosis (Ann Arbor stages 3 and 4) and male sex was associated with an increased rate of overall and lymphoma-related death among all NHL, DLBCL and follicular lymphoma patients (Table 2).

Table 2. Multivariable-adjusted HR for all-cause mortality among patients with all NHL or NHL subtypes, presented as HRs and 95% CIs
inline image

In univariate analyses, cumulative survival for all NHL was slightly lower toward the end of follow-up among current smokers compared to past and never smokers (Fig. 1a), and clearly lower for DLBCL (Fig. 1b). In multivariable-adjusted analyses, current smokers with all NHL or DLBCL had a statistically significantly higher rate of death from any cause compared to never smokers (all NHL HR 1.5, 1.2–1.8; DLBCL HR 1.8, 1.2–2.7; Table 2). Overall, high smoking intensity, long smoking duration and high number of cigarette-years were significantly associated with an increased rate of all-cause death in all NHL compared to nonsmokers, and, if anything, this pattern was also suggested among follicular lymphoma but not DLBCL patients (Table 3). The point estimates for current smoking at diagnosis and lymphoma-specific death were also elevated, but not statistically significant, among individuals with all NHL combined and DLBCL (all NHL HR 1.3, 1.0–1.6; DLBCL HR 1.5, 0.9–2.3; Supporting Information Table 2). Smoking intensity, duration and cigarette-years were not associated with lymphoma-related mortality in all NHL or any of the subtypes examined (data not shown).

thumbnail image

Figure 1. Cumulative overall survival by smoking status at diagnosis in patients with (a) non-Hodgkin lymphoma (NHL) or (b) DLBCL. (c) Cumulative survival by years of completed education in patients with NHL.

Download figure to PowerPoint

Table 3. Distribution of patients and multivariable-adjusted HR for smoking intensity, duration and cigarette-years and all-cause mortality, presented as number of deaths from any cause/number of cases and HRs and 95% CI1
inline image

Low educational attainment was associated with reduced overall survival for all NHL when assessed using the Kaplan–Meier estimator (Fig. 1c) and in multivariable-adjusted analyses including adjustment for smoking (HR 1.3, 1.1–1.6; Table 2). For lymphoma-related death, this association appeared elevated (HR 1.3, 1.0–1.6), with a statistically significant linear trend (p-trend = 0.047; Supporting Information Table 2).

Obesity (≥30.0 vs. <25 kg/m2) and frequency of sunbathing within 5–10 years of diagnosis were not significantly associated with death from any or lymphoma-related causes (Table 2 and Supporting Information Table 2). Results were similarly null for other UV exposure covariates, including sun vacations abroad, sun lamp use and sunburn history (data not shown).

For all NHL combined, self-reported history of autoimmune disease was associated with an increased rate of all-cause death (HR = 1.4, 1.0–1.8, p = 0.03; Table 2), but not significantly with lymphoma-related death (HR = 1.3, 1.0–1.8, p = 0.08; Supporting Information Table 2). No such association was evident for NHL subtypes. Family history of hematopoietic malignancy was not associated with all-cause or lymphoma-related death for all NHL or most subtypes.

In all analyses, missingness was not informative, except for smoking intensity and cigarette-years, where missingness was associated with increased HRs for all outcomes. However, there were only five missing observations (0.3%), and in a sensitivity analysis, using imputation to assign the five observations each of the different values within each variable, the results did not change. Competing risks analysis did not appreciably change the results for lymphoma-related mortality, suggesting that bias owing to competing risks was minimal (data not shown).

Discussion

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. References
  8. Supporting Information

Current smoking, as well as high smoking intensity, long duration and a high number of cigarette-years, compared to never smoking, were associated with an increased rate of all-cause mortality in our NHL study. Three previous smaller studies have reported that a high number of pack-years7, 10 or cigarettes smoked per day13 significantly increased both overall and lymphoma-related mortality in NHL patients. Our study had less statistical power than the previous studies owing to fewer current smokers (15 vs. 197 and 28%10) and ever smokers (51 vs. 60%13) as well as fewer patients smoking ≥20 cigarettes/day (13 vs. 217, 1910 and 24%13), which might explain the different results.

With regard to NHL subtypes, Battaglioli et al.10 found a significantly increased rate of all-cause death associated with long smoking duration (>39 years) for DLBCL, whereas Geyer et al.7 observed significantly increased rates of overall and lymphoma-related death associated with current, long-term (≥20 years) smoking and high pack-years (>17), primarily among patients with follicular lymphoma. In our study, current smoking at diagnosis, but not high cumulative smoking exposure (smoking duration and cigarette-years), was associated with increased mortality among patients with DLBCL, whereas the converse was suggested among patients with follicular lymphoma. These results are concordant with the idea that individuals with the aggressive DLBCL subtype need to be in good physical condition to tolerate the intensive chemotherapy aimed to cure the disease. Smokers are more likely to suffer from comorbid conditions, such as chronic lung and heart disease, than nonsmokers,32–34 and intensive treatment schedules may aggravate these conditions and reduce overall survival.35 For indolent lymphomas, such as most follicular lymphomas, initial treatment is often less intense or not needed in spite of widespread disease stage at diagnosis, which could explain the reason why current smoking status is less influential in this group.

We found that lower educational attainment, as an indicator of lower socioeconomic status, was positively associated with all-cause mortality for patients with NHL. For lymphoma-related death, we observed a significant linear trend. Two previous European studies demonstrated a worse outcome for NHL cases with low educational attainment12 or in deprived residential areas.11 Among these cases, a more advanced NHL disease was found at diagnosis,6, 11 reflected by lower performance status, higher IPI/FLIPI score, the presence of B-symptoms (fever, night sweats and weight loss) and divergent hemoglobin or lactate dehydrogenase, which could partly explain the lower survival in this group. Hence, NHL prognosis could potentially be improved in the lower socioeconomic classes by earlier detection of the NHL. Socioeconomic status may also influence NHL treatment. This ought not to be the case in Sweden, however, where all patients are offered government-financed universal health care. However, individuals with lower socioeconomic status have more comorbidities,36 which could reduce treatment tolerability, ultimately affecting survival.35

We observed an increased rate of all-cause death associated with autoimmune disease for all NHL combined but not for specific subtypes. The use of self-reported diagnoses may have introduced some misclassification, but we expect that it would be nondifferential. Comparing NHL patients with and without RA in the Nebraska Lymphoma Study Group, those with RA had a reduced rate of lymphoma-related mortality but were more likely to die from other comorbid conditions, and overall survival was not significantly different.26 However, two register-based studies conducted in Sweden suggested that patients with RA and NHL8 or DLBCL specifically22 had a worse prognosis than lymphoma patients without RA. One study of the impact of SLE on NHL outcome found no difference,25 whereas another study found an adverse prognosis.23 Along with other studied exposures, autoimmunity could reduce survival through comorbidities.35, 37–38 Finally, NHL tumor biology may also differ between patients with and without an autoimmune disease, since in the former, the disease evolves in a dysregulated immune system,39 which may have implications for prognosis.40

Two register-based studies explored the relationship between NHL survival and a family history of NHL24 or any lymphoma,5 and neither found a difference in prognosis among familial compared to sporadic NHL cases. However, in one of the studies,24 a nonsignificant familial concordance of NHL prognosis was noted. Indeed, genetic susceptibility studies indicate that NHL prognosis may in part be inherited through germline single nucleotide polymorphisms.9, 41, 42

We detected no statistically significant associations between NHL prognosis and either BMI or UV radiation. Although not statistically significant, the elevated HR among obese individuals (BMI ≥30) with DLBCL in our study does not contradict the results of a US study, showing a reduced overall and lymphoma-related survival among obese cases with all NHL.7 A lower prevalence of obesity in our study population (11 vs. 38% in the US study) might explain the difference. Somewhat contradictory to our findings was a recent report of an association between vitamin D insufficiency and worse lymphoma prognosis in a US study.43 The lack of an association with past UV-related behaviors here could relate to an imprecise correlation with vitamin D status at NHL diagnosis.

The strengths of our study include its population-based design, large size and long and virtually complete follow-up. Limitations include the lack of information on IPI/FLIPI score, performance status, laboratory parameters and treatment for multivariable adjustment. However, it is uncertain to what extent they are important confounders. The majority of the patients were treated according to the national written guidelines, which define primary and often also secondary and subsequent treatments for each subtype and stage/IPI/FLIPI score. Rituximab was usually not part of initial therapy during the time period but many may have received this treatment in combination with chemotherapy upon relapse. Although it seems improbable that our findings with regard to smoking and socioeconomic status would be invalid for patients treated with modern combined modality regimens including rituximab, we cannot exclude the possibility that the prognosis in autoimmune-associated NHL could be altered, as rituximab has a documented effect on prognosis both in autoimmune disease (mainly RA)44 and in NHL.1 Finally, although our study represents one of the largest investigations of the impact of lifestyle factors on NHL survival to date, we nevertheless had limited power for subtype-specific analyses, especially for the rarer NHL subtypes. Thus, any of the null associations, particularly with NHL subtypes or for lymphoma-related survival, may have been owing to insufficient power, just as any of the positive associations may have been owing to chance in view of the large number of comparisons made. Furthermore, some subtypes, such as T-cell lymphomas, are more heterogeneous than others.

Conclusions

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. References
  8. Supporting Information

In conclusion, our study adds to the evidence that cigarette smoking, low socioeconomic status and having an NHL risk-related autoimmune disease at NHL diagnosis have an adverse impact on overall survival among patients with NHL, perhaps partly through comorbidity. It is worth noting that as comorbidity is an exclusion criterion in many clinical trials providing the basis for treatment recommendations, less is known about its effect on outcome. Whether smoking cessation improves NHL outcome is not established. We assessed smoking status at diagnosis and defined former smoking as having quit at least 1 year earlier; therefore, our findings encourage such an investigation. The reason why low socioeconomic status or an autoimmune disease is associated with a worse outcome for patients needs further investigation before these factors can be used in clinical prognostic assessments.

References

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. References
  8. Supporting Information

Supporting Information

  1. Top of page
  2. Abstract
  3. Material and Methods
  4. Results
  5. Discussion
  6. Conclusions
  7. References
  8. Supporting Information

Additional Supporting Information may be found in the online version of this article.

FilenameFormatSizeDescription
IJC_27944_sm_SuppTab1.rtf130KSupporting Information Table 1.
IJC_27944_sm_SuppTab2.rtf142KSupporting Information Table 2.

Please note: Wiley Blackwell is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.