Improved population level survival in younger Hodgkin lymphoma patients in Germany in the early 21st century

Authors

  • Dianne Pulte,

    Corresponding author
    1. Division of Clinical Epidemiology and Aging Research, German Cancer Research Centre, Heidelberg, Germany
    2. Division of Hematology, Department of Medicine, Thomas Jefferson University, Philadelphia, PA, USA
    • Correspondence: Dianne Pulte, Division of Hematology, Department of Medicine, Thomas Jefferson University, 1015 Chestnut Street Suite 1321, Philadelphia, PA 19107, USA.

      E-mail: pultedi@gmail.com

    Search for more papers by this author
  • Lina Jansen,

    1. Division of Clinical Epidemiology and Aging Research, German Cancer Research Centre, Heidelberg, Germany
    Search for more papers by this author
  • Adam Gondos,

    1. Division of Clinical Epidemiology and Aging Research, German Cancer Research Centre, Heidelberg, Germany
    Search for more papers by this author
  • Katharina Emrich ,

    1. Cancer Registry of Rhineland-Palatinate Institute for Medical Biostatistics, Epidemiology and Informatics, University Medical Centre, Johannes Gutenberg University Mainz, Mainz, Germany
    Search for more papers by this author
  • Bernd Holleczek,

    1. Saarland Cancer Registry, Saarbrücken, Germany
    Search for more papers by this author
  • Alexander Katalinic,

    1. Cancer Registry of Schleswig-Holstein, Lübeck, Germany
    Search for more papers by this author
  • Hermann Brenner,

    1. Division of Clinical Epidemiology and Aging Research, German Cancer Research Centre, Heidelberg, Germany
    Search for more papers by this author
  • the GEKID Cancer Survival Working Group


Summary

Treatment for Hodgkin lymphoma (HL) is more aggressive in Germany than in the United States (US) and differences in treatment may lead to differences in population level survival. Patients diagnosed with HL in 11 German states in 1997–2006 were included in the analyses and were compared to similar analyses from patients in the Surveillance, Epidemiology, and End Results database in the US. Period analysis was used to calculate 5-year relative survival for the time period of 2002–2006 overall and by gender, age and histology. Overall 5-year relative survival for patients with HL in Germany was 84·3%, compared to 80·6% for the US. Survival was highest in patients aged 15–29 years at 97·9% and decreased with age to 57·5% at age 60 +  Survival for men and women, respectively, was 84·7% and 84·1% in Germany and 78·2% and 83·6% in the US. 5-year relative survival for patients diagnosed with HL in Germany was close to 100% for younger patients. Survival of HL patients in the US was lower than in Germany overall, but was comparable in older patients and in women. Population-based studies with longer follow-up are still needed to examine effects of late toxicity on long term survival.

Hodgkin lymphoma (HL) is one of the few cancers that can be reliably cured with chemotherapy alone. Prior to the introduction of combination chemotherapy, patients with stage III-IV HL had a 5-year survival of <10% (DeVita, 1981). Beginning in the 1960s, combination chemotherapy with various regimens has drastically improved survival (Frel et al, 1966; Santoro et al, 1987). The use of haematopoietic stem cell transplant further improved survival in patients with resistant disease (Armitage et al, 1991). In 2003, patients treated with a novel combination chemotherapy, BEACOPP (bleomycin, etoposide, doxorubicin, cyclophosphamide, vinblastine, procarbazine and prednisone) showed even greater survival in a clinical trial (Diehl et al, 2003) and has become the standard first line treatment in Germany. However, survival in clinical trials may not reflect survival on the population level (Pulte et al, 2011). Thus, there is a need to demonstrate that improvements in survival in clinical trials have translated to improvements on the population level.

Survival in HL has improved on the population level in Germany and other countries (Verdacchi et al, 2007; Brenner et al, 2008, 2009; Gondos et al, 2009). Until recently, population-based cancer survival analysis in Germany relied primarily on data from the Saarland Cancer Registry, which covers only 1·3% of the total population(Gondos et al, 2009) making detailed analysis of survival for relatively rare conditions such as HL difficult. Previous studies have demonstrated improved survival in HL in Germany, but have lacked the power to determine whether this improvement is present for potential sub-groups of patients, including patients at different ages, of both genders, and with various sub-types of HL.

The aim of this study was to provide up-to-date and detailed cancer survival estimates of German HL patients based on data from 11 population-based cancer registries, covering 33 million inhabitants, both overall and by specific patient and disease characteristics. Additionally, the survival expectations of patients in Germany were compared to those of patients with HL in the United States (US).

Methods

Data sources

A detailed description of the cancer registries from which data were obtained has been published previously (Hiripi et al, 2012). Briefly, data extracted from 11 cancer registries throughout Germany, representing a total base population of 33 million people, were included (Table 1). It should be noted that some registries provided information for only a portion of the years included (Table 1). Patients aged 15 years or older with a primary diagnosis of HL (International Classification of Diseases (ICD)-10 codes 81·0–81·9) in 1997–2006 and with mortality follow-up up to December 2006 were included. For some registries, data were available starting from later years only. A total of 5835 patients were initially identified. Patients reported to the cancer registry by death certificate or autopsy only (535 cases, 9·2%) were excluded, leaving 5300 patients for analysis.

Table 1. Patients from the US and Germany with Hodgkin lymphoma diagnosed in 1997–2006 by registry.
RegistryUnderlying population in 2006 (Million)Years of Registration includedCases diagnosed% DCO (excluded)Cases in the analysisMedian age at diagnosis (years)Microscopically confirmed cases (%)
  1. DCO, cases identified by death certificate only.

Bavaria8·132002–20069158·783537100
Brandenburg2·551997–20066559·35943999·3
Bremen0·661998–20061317·612139100
Hamburg1·751997–20064654·744340100
Mecklenburg-Vorpommern1·691997–20063887·535941100
Lower Saxony7·982001–2006105217·08734298·2
North Rhine-Westphalia2·621997–20044112·939945100
Rhineland-Palatinate0·521997–200610811·19649100
Saarland1·041997–20062283·122143100
Saxony4·251997–200610277·69493899·7
Schleswig-Holstein1·851999–20064559·741046100
Total33·04 58359·253004099·6
SEER13381997–200610,1920·610,13638 

In order to compare population level survival for HL in Germany with survival in the United States (US) data from the Surveillance, Epidemiology, and End Results (SEER13) database were analysed (http://seer.cancer.gov/data/). The same inclusion criteria as for patients from the German cancer registries were applied for the same time period. From the SEER13 database, 10 192 patients were identified, of whom 56 (0·6%) patients reported to the cancer registry by death certificate or autopsy only were excluded, leaving 10 136 patients for analysis. The SEER13 database includes data from 13 regional cancer centres in the US, covering a population of about 39 million people. Centres are chosen for inclusion based on their high quality and epidemiologically interesting population groups. The SEER population is considered to be similar to the general US population with respect to most sociodemographic characteristics, although it may be more affluent than average and may have slightly higher than average survival for some cancers (Mariotto et al, 2002).

Statistical methods

Five-year relative survival estimates for the time period 2002–06 were calculated using period analysis (Brenner et al, 2004) with age adjustment according to the International Cancer Survival Standards (Corazziari et al, 2004). Period analysis, first introduced in 1996 (Brenner & Gefeller, 1996) provides more up-to-date survival estimates than traditional cohort-based analysis. In particular, it has been shown by extensive empirical evaluation, that period estimates of 5-year relative survival for a given period quite closely predict 5-year relative survival later observed for patients during the period of interest (Brenner & Hakulinen, 2002; Brenner et al, 2002a). Because survival in HL varies with age and gender, we examined survival by major age groups (15–29, 30–39, 40–59, 60 + years) and by gender. Differences in survival between men and women, as well as between patients in Germany and the US, were tested for statistical significance, overall and by single age groups, using model-based period analysis, with a P-value of <0·05 being considered statistically significant (Brenner & Hakulinen, 2006). Subgroup analyses by histological type were also performed for nodular lymphocyte predominant HL (NLPHL, C81·0) nodular sclerosing classic HL (NSCHL, C81·1) mixed cellularity classic HL (MCHL, C81·2) lymphocyte-depleted HL (LDHL, C81·3) and HL, other or not otherwise specified (HL-NOS, C81·7 and C81·9) the most common subtypes of HL. Additionally, model-based period analysis (Brenner & Hakulinen, 2006) was also employed to estimate most recent changes in 5-year relative survival within the 2002–2006 period.

Relative survival was calculated as the ratio of actual survival to expected survival. Expected survival was estimated according to the Ederer II method (Ederer & Heise, 1959) using national life tables stratified by age, sex, and calendar year obtained from the German Federal Statistical Office. Relative survival for the US patients were calculated using US sex, age, calendar year and race-specific life tables published by the Centers for Disease Control (CDC) (Arias, 2002). Because stage is not available for haematological malignancies in the SEER and German database, stage could not be included in the model.

All calculations were carried out using SAS software (version 9·2) using macros developed for standard and modelled period analysis (Brenner et al, 2002b; Brenner & Hakulinen, 2006).

Results

A total of 5835 patients with HL were identified in the available databases in Germany. Between 98·2% and 100% of cases were confirmed microscopically (Table 1). Median age at diagnosis was 40 years overall, ranging from 37 to 49 years in different registries. Median age for patients in the US was 38 years at diagnosis.

Overall, 5-year relative survival in 2002–2006 was 84·3% for patients diagnosed with HL in Germany (Table 2). Survival ranged from 97·9% for ages 15–29 years to 57·5% for patients age 60 + years and was very similar for men (84·7%) and for women (84·1%). When histological subgroup was considered, survival was highest for patients with NLPHL and lowest for LDHL, although fewer than 100 cases of the latter were observed, leading to a large standard error of the survival estimate for the latter group of patients.

Table 2. Five-year relative survival of patients with Hodgkin lymphoma diagnosed 1997–2006 from Germany and the US by age, gender, and subtype.
 GermanyUS
Factor n RSSE n RSSEDiff P
  1. N, number of cases; Diff, difference between point estimates for Germany and US; RS, point estimate of relative survival; SE, standard error; ICD 10, International Classification of Diseases, version 2010; NLPHL, nodular lymphocyte predominant Hodgkin lymphoma; NSHL, nodular sclerosing Hodgkin lymphoma; MCHL, mixed cellularty Hodgkin lymphoma; LDHL, lymphocyte-depleted Hodgkin lymphoma; HL-NOS, other and unspecified Hodgkin lymphoma.

Age (years)
15–29154197·90·5318693·10·74·8<0·0001
30–39103595·80·9216391·40·94·40·0036
40–59128388·31·4267982·71·15·60·0015
60+144157·52·2210854·61·92·90·1043
Sex
Male294584·71·1547378·20·96·5<0·0001
Female235584·10·9466383·60·80·50·6679
ICD 10
NLPHL54189·31·969990·22·2−0·90·3143
NSHL227386·61·4624582·50·94·10·0004
MCHL132284·11·3140278·81·75·30·0783
LDHL9267·37·112755·76·411·60·9695
HL-NOS107480·81·5166374·11·56·70·0030
Total530084·30·71013680·60·63·7<0·0001

Survival for patients with HL in the US was lower than survival for patients in Germany overall (80·6% vs. 84·3%, respectively) in men and women, and in each age group analysed (Table 2). However, the differences were small and not statistically significant for patients aged 60 years or older or for women. When histological subtype was analysed, survival was statistically significantly better in Germany for patients with NSHL and HL-NOS. Compared to Germany, survival was lower for men (78·2%) than for women (83·6%) with HL in the US.

When survival was analysed by age and gender, survival was higher in Germany than the US for men at each age group. In contrast, in women, survival was significantly higher than in the US in age group 15–29 only (Table 3).

Table 3. Five-year relative survival of patients with Hodgkin lymphoma diagnosed in 1997–2006 in Germany and the US by age and gender.
 GermanyUSDiff P
RSSERSSE 
Age (years)
  1. a

    Age-standardized estimates.

  2. Diff, difference between point estimates for Germany and US; RS, point estimate of relative survival; SE, standard error.

Men
15–2997·60·892·81·04·80·0046
30–3995·81·288·71·47·10·0016
40–5987·31·880·11·57·20·0011
60+62·03·051·42·510·60·0037
Alla84·71·178·20·96·5<0·0001
Women
15–2998·20·793·40·94·80·0019
30–3995·71·494·31·11·40·5057
40–5990·12·186·71·63·40·3289
60+52·53·158·32·8−5·80·4569
Alla84·10·983·60·80·50·6679

Table 4 provides more detailed analyses of 5-year relative survival by histological subtypes and age. Survival was highest for NLPHL, for which survival in German patients aged 15–29 years was essentially the same as that of the general population, and relative survival was >90% except for age group 60 + , for which 5-year relative survival was much lower (72·2%). A similar pattern was observed for NSHL and MCHL; 5-year relative survival above or close to 90% in age groups 15–29, 30–39 and 40–59 years, and much lower survival (62·8% and 56·2%, respectively) in age group 60 + . For HL-NOS, the survival decrease began at a younger age, with 40- to 59-year olds having a 5-year relative survival of 81·2%. Again, survival was lower in the US for each subtype except for NLPHL in patients age 60 +  for whom survival was higher in the US. In all cases except NLPHL, the difference was significant for men, but not for women.

Table 4. Five-year survival of patients with Hodgkin lymphoma diagnosed in 1997–2006 in Germany and the US by age, sex, and histological subtype.
 GermanyUSDifference
NLPHL n RSSE n RSSEDiff P
  1. a

    P-value could not be calculated as model did not converge.

  2. Diff, Difference; NLPHL, nodular lymphocyte predominant Hodgkin lymphoma; NSHL, nodular sclerosing Hodgkin lymphoma; MCHL, mixed cellularity Hodgkin lymphoma, HL-NOS, other and unspecified Hodgkin lymphoma; Diff, difference between point estimates for Germany and US; RS, point estimate of relative survival; SE, standard error.

Age (years)
15–2991100·30·016597·81·92·5 a
30–3912494·33·114690·23·84·10·4027
40–5915898·22·023393·22·85·00·3157
60+16872·25·615581·56·2−9·30·0988
Sex
Male36088·12·846989·43·0−1·30·7027
Female18192·42·123089·83·62·60·7264
NSHL
Age (years)
15–2989397·40·8246792·70·84·70·0013
30–3953196·81·2153093·31·03·50·0550
40–5948590·32·1146086·71·43·60·1449
60+36462·84·678858·43·04·40·0627
Sex
Male114988·02·0304180·81·37·20·0040
Female112485·01·9320484·71·20·30·0553
MCHL
Age (years)
15–2929198·41·121593·72·54·70·1092
30–3919495·42·020688·83·56·60·1315
40–5936587·52·747177·72·99·80·0140
60+47256·23·751055·03·91·20·8711
Sex
Male83885·22·089876·22·09·00·0015
Female48484·11·850482·22·81·90·6962
HL-NOS
Age (years)
15–2925698·90·932295·41·83·50·1437
30–3917694·92·526283·33·311·60·0202
40–5925581·23·847773·23·080·0574
60+38549·64·360244·53·45·10·0551
Sex
Male54281·02·198470·42·110·60·0003
Female53280·12·267979·32·00·80·4002

Because of concern that the results could be affected by differences in the racial makeup of Germany versus the US, we examined survival for patients in the SEER database whose race was listed as white. When the analysis was restricted in this manner, the US survival estimates increased by <1% in each age group, with no substantial change of the differences observed between Germany and the US.

In order to determine changes in survival in very recent years, a model-based period analysis of survival trends in 2002–2006 was performed. Survival was higher in 2006 compared with 2002 for all age groups except for 60+ (Table 5). However, the difference was statistically significant only for ages 30–39 years, for whom survival increased from 92·5% to 97·9%. It should be noted that survival for ages 15–29 years was 97·3% in 2002, making further increases in survival difficult to detect. When individual subtypes of HL were examined, survival improved for NLPHL, from 83·9% in 2002 to 93·1% in 2006, and remained essentially unchanged for other subtypes. However, case numbers for specific subtypes are quite small, making detection of subtle changes difficult.

Table 5. Model based estimates of trends in five-year relative survival of patients with Hodgkin lymphoma from Germany in 2002 and 2006 by age and histological subgroup.
 20022006Difference
RSSERSSEDiff P
  1. a

    Age-standardized estimates.

  2. Diff, Difference; NLPHL, nodular lymphocyte predominant Hodgkin lymphoma; NSHL, nodular sclerosing Hodgkin lymphoma; MCHL, mixed cellularity Hodgkin lymphoma; HL-NOS, other and unspecified Hodgkin lymphoma.

Age Group (years)
15–2997·31·298·30·8+1·00·5490
30–3992·52·497·91·0+5·40·0464
40–5986·52·889·72·0+3·20·4023
60+60·83·654·93·3−5·90·2808
Alla84·21·284·21·00·00·9864
Histology
NLPHL83·93·393·12·3+9·20·0408
NSHL87·52·384·52·0−3·00·3782
MCHL84·62·183·51·9−1·10·7183
HL-NOS82·12·580·02·3−2·00·5779

Discussion

HL is one of the few cancers that can be truly cured with chemotherapy. Combination chemotherapy with or without radiotherapy has been the standard treatment for stage III–IV HL since the 1960s and has been used in earlier stage HL since the 1990s. However, the specific type of therapy has changed over time and both improvements in chemotherapy and use of haematopoietic stem cell transplantation have led to improved survival in clinical trials and on the population level during the latter half of the 20th century and beginning of the 21st century (Frel et al, 1966; DeVita, 1981; Santoro et al, 1987; Armitage et al, 1991; Diehl et al, 2003).

Germany has been the site of a number of practice changing clinical trials (Diehl et al, 2003; Behringer & Diehl, 2005) and has had high population level overall survival for HL (Hiripi et al, 2012). This history of groundbreaking care is reflected in the high survival for patients with HL observed in this study. However, our study identified areas of concern. First, survival in older patients remains far lower than survival for younger patients. Additionally, survival in 2006 was no higher and trended towards lower than in 2002 for patients age 60 + , suggesting that there has not been significant progress in improving survival for older patients with HL in the past several years. Previous studies have suggested that, while cancer survival has improved for older patients, it has not improved to the same extent as for younger patients, particularly in cancers for which major increases in survival were observed (Gondos et al, 2007). The current results demonstrate that this trend extends to HL. This finding again underscores the need for further research specifically designed to examine the efficacy and safety of cancer treatment options in older patients.

In general, survival for patients with HL was higher in Germany than in the US, with statistically significant differences more often occurring in younger age groups and in men. Although we cannot definitively state the reasons for these differences, several possible explanations may be hypothesized. First, initial treatment in Germany may be more aggressive, with regimens such as BEACOPP being used more often in Germany whereas ABVD (doxorubicin, bleomycin, vinblastine and dacarbazine) and ABVD-like regimens are still often the initial treatment approach in the US. BEACOPP has been associated with higher disease-free and overall survival in clinical trials (Diehl et al, 2003). In the pivotal trial of BEACOPP versus COPP-ABVD, a complete remission rate of 96% and 5-year survival of 91% were seen for BEACOPP therapy versus 85% and 83%, respectively, for COPP-ABVD (Diehl et al, 2003). However, there is concern about increased late toxicity with BEACOPP, including a higher incidence of leukaemia and other tumours later in life, which may negate some of the early survival advantage (Scholz et al, 2011). At this time, longer term survival analysis is not possible in the GEKID database, but we plan to continue to monitor survival in HL patients in order to determine whether the higher survival at 5 years is sustained over a longer time period.

Additionally, lower survival in patients with HL without health insurance has been documented (Yung et al, 2011) particularly among patients who may have lost their insurance around the time of diagnosis (Koroukian et al, 2012). In the US, younger adults are especially likely to lack health insurance (Collins et al, 2007; Martin et al, 2007) and only people over 65 years of age are guaranteed health insurance via the Medicare program. In contrast, in Germany, essentially all citizens are guaranteed health insurance. Thus, differences in access to medical care and delays due to financial constraints may play a role in relatively low survival in HL patients in the US.

The median age at diagnosis with HL was lower in the US than in Germany. Although we cannot rule out some differences in the aetiology of the disease in each country, this observation may primarily reflect the much younger age structure of the US population compared to the German population (difference in median age of almost 7 years) and should not have affected age-specific and age-standardized survival estimates.

Finally, human immunodeficiency virus (HIV) infection rates are higher in the US, particularly among men, compared with Germany (Robert Koch Institut, 2010; Centers for Disease Control and Prevention (CDC) (2008). HIV prevalence in the US is around 480 per 100 000 versus <100 per 100 000 in Germany (Robert Koch Institut, 2010; Centers for Disease Control and Prevention (CDC) 2008). HIV infection increases the risk of HL and decreases survival, especially if it is poorly controlled. Thus, the higher rates of HIV infection in the US may play a role in the observed differences. However, HIV prevalence in the white population in the US is about 130 per 100 000 (Centers for Disease Control and Prevention (CDC) 2008) making it similar to the prevalence in Germany and survival for HL in young people in Germany continues to be higher than in the US when analysis was restricted to people of white race (data not shown). This suggests that HIV infection plays only a minor role in the differences in survival observed.

Male gender was associated with lower 5-year survival in the US but not in Germany. Male gender has been considered a poor prognostic indicator under the International Prognostic Score (IPS) (Hasenclever & Diehl, 1998). However, it should be noted that the IPS was validated in patients treated prior to 1992 and, thus, prior to the use of BEACOPP and other newer therapies and has been validated for advanced disease only. Thus, the similarity of survival in both genders in Germany suggests that gender may be less of a prognostic factor with newer treatment options, but further investigation is needed.

In interpreting our results, several limitations must be considered. First, despite the large number of registries covering a very large population base, the relative rarity of some forms of HL limits our ability to estimate survival precisely for these histologies. Second, in the absence of a national death index, most cancer registries rely on record linkage with vital statistics from the state that they cover and may miss deaths among patients who move out of the state. Nevertheless, previous validation studies have suggested potential overestimation of survival due to deaths missed by migration to be very small (Hiripi et al, 2012). Third, we could not include stage at diagnosis in our model as the SEER and German databases do not include information on stage for haematological malignancies. Given that there is no standard screening test for HL in use in either country, one might expect similar stage distribution. However, the lack of universal health insurance in the US may lead to some delay in diagnosis, especially in younger patients. Fourth, methods of diagnostic confirmation of HL have changed over time. It is possible that some changes in the diagnostic methods (i.e. more frequent use of cytology or flow cytometry) may have altered the mix of cases over time. However, any changes that might have occurred should have affected diagnostic confirmation more or less equally in Germany and the US.

Finally, the number of death certificate only (DCO) cases was significantly higher in the German databases compared with the SEER database, particularly among newer registries which are still in the build-up phase, and this may lead to higher apparent survival rates in Germany. Model calculations based on age-specific DCO rates suggest that survival rates might have been overestimated by up to several percent units in Germany in age groups 40–59 and 60 + years due to exclusion of high proportions of DCO cases (Brenner & Holleczek, 2011) which might partly explain the apparent survival differences between Germany and the US in these age groups. By contrast, DCO proportions, and hence potential bias, are relatively small for age groups 15–29 and 30–39 years and could not (age group 15–29 years) or very unlikely (age group 30–39 years) account for the inter-country differences in these age groups. Conversely, as mentioned earlier, there is some evidence that survival estimates from the SEER database may be slightly higher than survival in the US population in general (Mariotto et al, 2002) which might have led to some underestimation of survival differences between both countries.

In summary, the 5-year survival for HL is generally quite high in Germany and there appears to be some improvement over time for younger patients, even in the short period available for analysis. Much lower survival was persistently observed for older patients, suggesting a need for greater concentration on treatment of HL in this group. Further population-based survival analyses with longer-term follow-up of patients are needed to examine possible effects of late toxicity on survival.

Acknowledgements

Members of the GEKID Cancer Survival Working Group: Karla Geiss, Martin Meyer (Cancer Registry of Bavaria) Andrea Eberle, Sabine Luttmann (Cancer Registry of Bremen) Roland Stabenow (Cancer Registry of Berlin and the New Federal States) Stefan Hentschel, Alice Nennecke (Cancer Registry of Hamburg); Joachim Kieschke, Eunice Sirri (Cancer Registry of Lower Saxony) Bernd Holleczek (Saarland Cancer Registry) Katharina Emrich (Cancer Registry of Rhineland-Palatinate) Hiltraud Kajüter, Volkmar Mattauch (Cancer Registry of North Rhine-Westphalia) Alexander Katalinic (Cancer Registry of Schleswig-Holstein) Klaus Kraywinkel (Robert Koch Institute, Berlin) Hermann Brenner, Adam Gondos, Lina Jansen (DKFZ).

This work was supported in part by a grant from the German Cancer Aid (Deutsche Krebshilfe, no. 108257) and a visiting scientist grant from the German Cancer Research Centre (DKFZ).

Author contributions

DP: contributed to the design of the experiment and analysis of the results and wrote the manuscript. LJ performed the analyses, contributed to the interpretation of the results, and critically reviewed the manuscript. AG contributed to the organization of the database and the analysis of the results and critically reviewed the manuscript. BH, AK, KE contributed to the organization of the database and critically reviewed the manuscript. HB organized the GEKID database, designed the experiment, contributed to the interpretation of the results, and critically reviewed the manuscript.

Ancillary