Diagnostic value of JAK2 V617F somatic mutation for myeloproliferative cancer in 49 488 individuals from the general population

Authors

  • Camilla Nielsen,

    1. Department of Clinical Biochemistry, Herlev Hospital, Copenhagen University Hospital, Herlev, Denmark
    2. The Copenhagen General Study Population, Herlev Hospital, Copenhagen University Hospital, Herlev, Denmark
    3. Faculty of Health Sciences, University of Copenhagen, Herlev, Denmark
    Search for more papers by this author
  • Henrik S. Birgens,

    1. Faculty of Health Sciences, University of Copenhagen, Herlev, Denmark
    2. Department of Haematology, Herlev Hospital, Copenhagen University Hospital, Herlev, Denmark
    Search for more papers by this author
  • Børge G. Nordestgaard,

    1. Department of Clinical Biochemistry, Herlev Hospital, Copenhagen University Hospital, Herlev, Denmark
    2. The Copenhagen General Study Population, Herlev Hospital, Copenhagen University Hospital, Herlev, Denmark
    3. Faculty of Health Sciences, University of Copenhagen, Herlev, Denmark
    Search for more papers by this author
  • Stig E. Bojesen

    Corresponding author
    1. The Copenhagen General Study Population, Herlev Hospital, Copenhagen University Hospital, Herlev, Denmark
    2. Faculty of Health Sciences, University of Copenhagen, Herlev, Denmark
    • Department of Clinical Biochemistry, Herlev Hospital, Copenhagen University Hospital, Herlev, Denmark
    Search for more papers by this author

Correspondence: Stig E. Bojesen, Department of Clinical Biochemistry, 54M1, Herlev Hospital, Copenhagen University Hospital, Herlev Ringvej 75, 2730 Herlev, Denmark.

E-mail: Stig.Egil.Bojesen@regionh.dk

Summary

The JAK2 V617F somatic mutation is present in the majority of patients with myeloproliferative cancer (polycythaemia vera, essential thrombocytosis, and primary myelofibrosis). However, the diagnostic value of the JAK2 V617F somatic mutation for myeloproliferative cancer in the general population is unknown. We examined this question in 49 488 individuals from the Copenhagen General Population Study. We also examined the association between JAK2 V617F somatic mutation, rs10974944 germline genotype, haematological phenotype, any cancer, haematological cancer, myeloproliferative cancer, ischaemic heart disease, and venous thromboembolism. The JAK2 V617F somatic mutation was present in 0·1% (n = 68), increasing across rs10974944 germline genotypes (P-trend = 0·001). JAK2 V617F somatic mutation positives versus negatives had higher erythrocyte (P = 2 × 10−5), thrombocyte (P = 2 × 10−16), and leucocyte (P = 4 × 10−9) counts, and had 2·7-/2·5-fold risk of cancer (prevalent/incident), 44-/28-fold risk of haematological cancer, 221-/97-fold risk of myeloproliferative cancer, 2·2-/1·2-fold risk of ischaemic heart disease, and 3·1-/1·0-fold risk of venous thromboembolism. By combining conventional haematological parameters with a test for the JAK2 V617F somatic mutation, myelo;?>proliferative cancer could be identified or ruled out with a sensitivity of 47–100% and a specificity of 98–100%. In conclusion, in the general population the JAK2 V617F somatic mutation has a high diagnostic value for myeloproliferative cancer when combined with conventional haematological parameters.

JAK2 V617F somatic mutation status is now used routinely in the diagnostic and monitoring procedures for myeloproliferative cancers (Tefferi & Vardiman, 2008), i.e., polycythaemia vera, essential thrombocythaemia, and primary myelofibrosis. However, the diagnostic value of a positive JAK2 V617F somatic mutation test for myeloproliferative cancer is unknown when used in the general population.

Patients with myeloproliferative cancer usually present with abnormal haematological parameters (Tefferi & Vardiman, 2008), however, whether JAK2 V617F somatic mutation positive subjects found in the general population without overt signs of myeloproliferative cancer (Sidon et al, 2006; Xu et al, 2007; de Stefano et al, 2007; Martinaud et al, 2010; Nielsen et al, 2011), also have abnormal haematological parameters is presently unknown. Although the JAK2 V617F somatic mutation is associated with higher risk of thrombosis in patients with polycythaemia vera and essential thrombocythaemia (Barosi et al, 2007; Vannucchi et al, 2007; Basquiera et al, 2009; Tefferi et al, 2009), it is yet unknown if this association also applies to JAK2 V617F somatic mutation positives in the general population. Finally, genome-wide analyses identified a haplotype of the JAK2 locus spanning over the rs10974944 germline polymorphism that might predispose to the development of the JAK2 V617F somatic mutation, and consequently to myeloproliferative cancer (Jones et al, 2009; Kilpivaara et al, 2009; Olcaydu et al, 2009a,b). However, the relationship between rs10974944 germline genotype and haematological parameters, morbidity, and mortality in the general population is also unclear at present.

This study examined the association between JAK2 V617F somatic mutation, rs10974944 germline polymorphism, and haematological phenotype in 49 488 individuals from the Danish general population, the Copenhagen General Population Study. Furthermore, we measured risk of any cancer, haematological cancer, myeloproliferative cancer, ischaemic heart disease and venous thromboembolism as a function of JAK2 V617F somatic mutation status as well as rs10974944 germline genotype in the general population. Finally, we evaluated the diagnostic value of the test for the JAK2 V617F somatic mutation with respect to myeloproliferative cancer in all individuals from the general population, and in relevant subgroups with elevated haematological parameters.

Participants and methods

Study population

The Copenhagen General Population Study was initiated in 2003 and is still recruiting individuals (Nordestgaard et al, 2007; Frikke-Schmidt et al, 2008; Zacho et al, 2008). Of eligible white individuals of Danish descent in the selected geographical areas, 25% of 20–39-year-olds and 100% of over 39-year-olds were invited and 46% participated. We included the first 49 488 participants with available DNA samples. Participants gave written informed consent. The study was approved by a Danish ethical committee (H-KF 01-144/01) and by Herlev Hospital, Copenhagen University Hospital.

Covariates

Participants filled in a self-administered questionnaire concerning present and past life-style and health status. This was completed by an examiner on the day of attendance, prior to physical examination and blood sampling. Current and cumulative tobacco consumption and alcohol consumption were identified from information in the questionnaire, while body mass index was calculated as measured weight (kg) divided by measured height (m) squared.

Genotyping

The rs10974944 germline genotype is inherited from parents and is present from conception of the individual. The polymorphism was chosen as a marker of the JAK2 haplotype designated 46/1, associated with risk of developing the JAK2 V617F somatic mutation (Jones et al, 2009; Kilpivaara et al, 2009; Olcaydu et al, 2009a). Genotyping was performed as reported in the supporting information.

Somatic mutation detection assays

The JAK2 V617F mutation arises in some haematopoietic stem cells at some point of time after conception in certain individuals. The JAK2 V617F somatic mutation was first detected in DNA isolated from peripheral leucocytes from all 49 488 participants using a polymerase chain reaction (PCR)-based Taqman screening assay developed by our own group (Nielsen et al, 2011). The 1% of the participants (n = 493) with the highest signal from the screening assay were then tested using a highly sensitive real-time quantitative PCR assay, based on a previously published assay (Larsen et al, 2007) that quantitated the mutation burden. Individuals with a mutation burden of 0·8% or above were categorized as positive for the JAK2 V617F somatic mutation as this was the lower detection limit of the assay. All other individuals were categorized as negative. Additional details on the quantitative PCR assay are provided in the supporting information.

Haematological phenotype

Haematological parameters were measured with a flow cytometer-based haematology analyser in the routine laboratory of Herlev Hospital, Copenhagen University Hospital as described in the supporting information.

Endpoints

All registry information was collected using each participant's unique Central Person Registry number and follow-up was 100% complete, i.e., we did not lose track of even a single individual. Dates of death until 18 August 2010 were obtained from the national Danish Civil Registration System (Pedersen et al, 2006).

Diagnoses of cancers from 1943 until 31 December 2009 were obtained from the national Danish Cancer Registry (Storm, 1991; Storm et al, 1997). Diagnoses were classified according to the World Health Organization (WHO) International Classification of Diseases (ICD) (World Health Organization, 1952, 2004), seventh revision (ICD-7 codes 200–205;207;209;404;503–504;514;900;903–904;914 for haematological cancer and 207;209 for myeloproliferative cancer) until 31 December 1977 or 10th revision (ICD-10 codes C81;C84-C85;C88;C90–96;D45–47;D60–61;D70–77;D85 for haematological cancer and C94;D45;D47;D75 for myeloproliferative cancer) thereafter. Thus, the endpoint ‘myeloproliferative cancer’ is included in the endpoint ‘haematological cancer’, which in turn is included in the endpoint ‘any cancer’.

Remaining endpoints were collected from the national Danish Patients Registry and/or the national Danish Causes of Death Registry. Ischaemic heart disease (ICD-8 410–414 and ICD10 I20-I25) was fatal or non-fatal myocardial infarction or characteristic symptoms of angina pectoris, including revascularization procedures; myocardial infarction was diagnosed based on the presence of characteristic chest pain, electrocardiographic changes, and/or elevated cardiac biomarkers following the changes in diagnostic criteria over time (Thygesen et al, 2007). Venous thromboembolism (ICD-8 450–451;671;673 and ICD10 I26;I80;O22;O87;O88) was pulmonary embolism or deep venous thrombosis (Frederiksen et al, 2004).

Statistical analyses

The statistical software package stata, release 11·1 was used for all analyses. We used Pearson's chi-square test for categorical data, and Kruskal–Wallis test and Cuzick's Wilcoxon rank-sum test for continuous data. All statistical tests were two-sided and P values below 0·05 were considered significant, except when correcting for multiple testing.

In the cross-sectional design, which analysed prevalent disease diagnosed prior to blood sampling, we used logistic regression to calculate odds ratios. All analyses were adjusted for sex, age (continuous), current tobacco consumption (0, 0·1–20, or above 20 cigarettes/day or equivalent), alcohol consumption (0, 0·1–168, or above 168 g/week for women; and 0, 0·1–252, or above 252 g/week for men), and body mass index (below 18·5, 18·5–24·9, 25·0–29·9, 30·0–40·0, or above 40 kg/m2), all measured at the time of blood sampling.

In the prospective design, which analysed incident events diagnosed after blood sampling, we used Cox proportional hazard regression models and log-rank statistics with time after blood sampling as the time scale to calculate hazard ratios and test for differences in cumulative incidence. All analyses were adjusted as mentioned above. For all end-points, follow-up began at the time of blood sampling. Participants with diagnosis prior to the time of blood sampling were excluded from the specific analysis. For the endpoints any cancer, haematological cancer and myeloproliferative cancer, follow-up ended at first incident diagnosis, death, emigration, or 31 December 2009, whichever came first. For the endpoints ischaemic heart disease, myocardial infarction, venous thromboembolism, pulmonary embolism, and deep venous thrombosis, follow-up ended at first incident diagnosis, death, emigration, or 10 May 2011, whichever came first.

The diagnostic value of the test for the JAK2 V617F somatic mutation with respect to myeloproliferative cancer was evaluated calculating sensitivity, specificity, positive predictive value, and negative predictive value. Sensitivity was the percentage of individuals with myeloproliferative cancer who tested positive for the JAK2 V617F somatic mutation. Specificity was the percentage of individuals without myeloproliferative cancer who tested negative for the JAK2 V617F somatic mutation. Positive predictive value was the percentage of individuals with a positive JAK2 V617F somatic mutation test who suffered from myeloproliferative cancer, while negative predictive value was the percentage of individuals with a negative test, who did not suffer from myeloproliferative cancer. The cut-points for the parameters age, erythrocyte count, platelet count, leucocyte count, and erythrocyte volume fraction were chosen with the aim of dividing individuals into clinically relevant substrata, based on experience from daily clinical practice, and were chosen a priori. The diagnostic value for the JAK2 V617F somatic mutation test was examined individually for each substratum, while ignoring the other parameters. All these parameters are well-known risk factors in patients with myeloproliferative cancer (Landolfi et al, 2006).

Results

JAK2 V617F somatic mutation and rs10974944 germline polymorphism

Of 49 488 individuals from the general population, 68 individuals were positive for the JAK2 V617F somatic mutation. This corresponded to a prevalence of 0·1% in this sample of the general population with a median age of 56 years at the time of blood sampling. Figure 1 shows an increase in the number of positive individuals with lower levels of mutation down to 5–10% which then drops off at 0·8–5%. The fraction of JAK2 V617F somatic mutation positive individuals increased across rs10974944 genotype with 21 positives of 26 021 (= 0·08%) CC non-carriers, 39 of 19 806 (= 0·19%) CG heterozygotes, and 8 of 3661 (= 0·22%) GG homozygotes (P-trend = 0·001, Supplementary Figure 1). Allele frequencies of the rs10974944 germline polymorphism were 73% for the C-allele and 27% for the G-allele. We found 53% non-carriers (CC), 40% heterozygotes (CG), and 7% homozygotes (GG), and these genotype frequencies were in accordance with the Hardy–Weinberg equilibrium (P = 0·85).

Figure 1.

Number of individuals by JAK2 V617F mutation burden among the 68 JAK2 V617F somatic mutation positive individuals. These were identified among 49 488 individuals from the Danish general population.

Baseline characteristics of individuals by JAK2 V617F somatic mutation status are presented in Supplementary Table 1. Among all individuals, presence of the mutation was positively associated with male sex (P = 0·01), increasing age (P = 4 × 10−5), and with increasing current smoking consumption (P = 0·006). By rs10974944 genotype there was no difference in age, current and cumulative tobacco consumption, alcohol consumption, or body mass index (data not shown).

Haematological phenotype

JAK2 V617F somatic mutation positive versus negative individuals had higher erythrocyte (P = 2 × 10−5), platelet (P = 2 × 10−16), and leucocyte (P = 4 × 10−9) counts; the difference in leucocytes was due to neutrophils (P = 3 × 10−13), eosinophils (P = 0·004), monocytes (P = 0·02), and basophils (P = 1 × 10−9) (Table 1). Mean corpuscular haemoglobin concentration and mean corpuscular volume were lower (P = 0·02 and P = 0·005), while erythrocyte volume fraction, haemoglobin concentration, and lymphocyte count did not differ. Exclusion of the 10 JAK2 V617F somatic mutation positive individuals diagnosed with myeloproliferative cancer prior to blood sampling gave similar estimates for haematological parameters (data not shown). Likewise, exclusion of the 23 JAK2 V617F somatic mutation positive individuals diagnosed with any cancer also gave similar estimates for haematological parameters (data not shown).

Table 1. Haematological parameters by JAK2 V617F somatic mutation and rs10974944 germline genotype status
 JAK2 V617F somatic mutationrs10974944 germline genotype
NegativesPositivesP valueNon-carriers (CC)Heterozygotes (CG)Homozygotes (GG)P value
  1. Values are mean (2·5%–97·5% percentile) for continuous variables.

  2. MCHC, mean corpuscular haemoglobin concentration; MCV, mean corpuscular volume.

  3. P-values are from Kruskal–Wallis test (JAK2 V617F mutation) and Cuzick's trend test (rs10974944 genotype), and corrected for 12 comparisons using the Bonferroni method (P values < 0·05 were multiplied by 12).

Individuals, n49 42068 26 02119 8063661 
Erythrocyte count, 1012/l4·60 (3·82–5·38)4·94 (3·39–6·48)2 × 10−54·59 (3·82–5·38)4·61 (3·82–5·39)4·62 (3·84–5·39)0·001
Platelet count, 109/l278·4 (144·3–412·4)473·6 (27·2–919·9)2 × 10−16279·7 (145·0–414·4)277·8 (140·5–415·1)275·6 (141·9–409·2)4 × 10−5
Leucocyte count, 109/l7·41 (2·86–11·96)9·53 (2·89–16·18)4 × 10−97·40 (3·19–11·62)7·42 (2·31–12·53)7·43 (3·73–11·12)0·37
Neutrophils, 109/l4·34 (1·54–7·14)6·34 (0·77–11·91)3 × 10−134·33 (1·57–7·09)4·35 (1·49–7·21)4·37 (1·51–7·22)0·33
Eosinophils, 109/l0·20 (0–0·49)0·29 (0–0·73)0·0040·19 (0–0·51)0·20 (0–0·47)0·21 (0–0·48)0·006
Monocytes, 109/l0·41 (0·12–0·71)0·46 (0·18–0·74)0·020·413 (0·128–0·698)0·412 (0·099–0·725)0·414 (0·112–0·717)0·83
Basophils, 109/l0·040 (0–0·116)0·076 (0–0·191)1 × 10−90·040 (0–0·085)0·041 (0–0·147)0·041 (0–0·083)0·22
Lymphocytes, 109/l2·25 (0–5·22)2·06 (0·49–3·64)0·402·25 (0–4·74)2·25 (0–5·91)2·23 (0·63–3·84)0·68
Erythrocyte volume fraction,%40·9 (34·5–47·3)41·9 (32·9–50·7)0·2340·9 (34·5–47·3)40·9 (34·6–47·4)41·0 (34·6–47·4)0·07
MCHC, g/l348 (326–371)341 (304–377)0·023482 (3259–3706)3482 (3257–3706)3482 (3255–3708)0·97
MCV, fl89·1 (80·1–98·1)85·8 (66·5–105·1)0·00589·1 (80·1–98·1)89·1 (80·0–98·1)88·9 (79·8–98·0)0·06
Haemoglobin, g/l142 (118–166)142 (109–176)0·95142 (118–166)142 (119–166)143 (119–167)0·17

For the rs10974944 germline genotype, the erythrocyte count increased slightly across genotypes from non-carriers to heterozygotes and homozygotes (P-trend = 0·001), as did eosinophils (P = 0·006), while platelet count decreased slightly (P = 4 × 10−5)(Table 1). Other haematological parameters did not differ by rs10974944 germline genotype.

Among the 68 JAK2 V617F somatic mutation positives, 13 had an erythrocyte count above the reference interval, 43 had platelet counts above the reference interval, and 35 had leucocyte counts above the reference interval. For the rest of the haematological parameters, the majority were within the reference interval (Table 2).

Table 2. JAK2 V617F somatic mutation positives and haematological parameters
 According to reference interval, n
ParameterReference intervalBelowWithinAbove
  1. F = female, M = male.

  2. a

    67 individuals.

Erythrocyte count, 1012/lF: 3·94–5·16, M: 4·25–5·7184713
Platelet count, 109/lF: 165–385, M: 145–35022343
Leucocyte count, 109/l3·5–8·823135
Neutrophilsa, 109/l1·8–6·9514323
Eosinophils, 109/l<0·455414
Monocytes, 109/l<1·10680
Basophils, 109/l<0·20644
Lymphocytes, 109/l0·70–4·802651
Erythrocyte volume fraction,%F: 35·0–46·0, M: 40·0–50·06584
MCHC, g/l317–35814495
MCV, fl82–9814495
Haemoglobin, g/lF: 118–153, M: 134–16911516

JAK2 V617F somatic mutation burden and myeloid haematological parameters

In the 68 individuals positive for the JAK2 V617F somatic mutation, the mutation burden ranged from 0·8% to 50%. The majority of the positive individuals (87%) had a mutation burden below 12% (Fig 1). Erythrocyte, platelet, and leucocyte counts increased with increasing JAK2 V617F mutation burden (P-trend = 0·02, 0·04, and 0·01, respectively) (Supplementary Table 2).

Prevalent morbidity

As we used both a cross-sectional design to measure risk at blood sampling as well as a prospective design to measure time to event, we were able to distinguish between prevalent and incident morbidity. Of the 68 individuals positive for the JAK2 V617F somatic mutation, 23 had a cancer diagnosis before blood sampling, resulting in a multifactorially adjusted odds ratio for any cancer of 2·7 (95% confidence interval [CI]: 1·6–4·6) (Table 3). Ten JAK2 V617F somatic mutation positive individuals had a diagnosis of haematological cancer at blood sampling, all of them myeloproliferative cancer. Exclusion of these 10 individuals led to a multifactorially adjusted odds ratio for any cancer of 1·3 (0·7–2·6) accounting for the cancers outlined in Supplementary Table 3. Multifactorially adjusted odds ratios for haematological cancer and myeloproliferative cancer in JAK2 V617F somatic mutation positives versus negatives were 44 (22–90) and 221 (100–487). For ischaemic heart disease and myocardial infarction corresponding odds ratios were 2·2 (1·1–4·4) and 2·6 (1·1–6·3). Corresponding odds ratio for venous thromboembolism was 3·1 (1·3–7·9), and for deep venous thrombosis 4·6 (1·7–10·9). The rs10974944 germline genotype did not associate with risk of prevalent disease (data not shown).

Table 3. Prevalent morbidity in the general population according to JAK2 V617F somatic mutation status
 JAK2 V617F somatic mutation status
EndpointsNegatives (n = 49 420) cases/controlsPositives (n = 68) cases/controlsOdds ratio (95% CI)
  1. Odds ratios were adjusted for sex, age, tobacco consumption, alcohol consumption, and body mass index at the time of blood sampling. CI = confidence interval.

Any cancer6969/42 45123/452·7 (1·6–4·6)
Haematological cancer139/49 28110/5844 (22–90)
Myeloproliferative cancer32/49 38810/58221 (100–487)
Ischaemic heart disease2689/46 73111/572·2 (1·1–4·4)
Myocardial infarction1091/48 3296/622·6 (1·1–6·3)
Venous thromboembolism989/48 4315/633·1 (1·3–7·9)
Pulmonary embolism344/49 0760/68
Deep venous thrombosis729/48 6915/634·6 (1·7–10·9)

Incident morbidity

The median follow-up time was 4·9 years, ranging from 0·003 to 7·5 years. Individuals positive for the JAK2 V617F somatic mutation had a higher cumulative incidence of any cancer compared to mutation negatives (log-rank, P = 1 × 10−5, Fig 2A), corresponding to a multifactorially adjusted hazard ratio of 2·45 (1·22–4·92) (Table 4). Corresponding values for haematological cancer were log-rank P = 3 × 10−63 (Fig 2B) and a hazard ratio of 27·56 (11·98–63·44), and for myeloproliferative cancer log-rank P = 4 × 10−91(Fig 2C) and a hazard ratio of 97·12 (27·64–341·28) (Table 4). In JAK2 V617F somatic mutation positives versus negatives the corresponding hazard ratios were 1·24 (0·39–3·86) for ischaemic heart disease, 0·81 (0·11–5·74) for myocardial infarction, and 1·04 (0·15–7·39) for venous thromboembolism. JAK2 V617F somatic mutation positives versus negatives had a lower cumulative survival (log-rank P = 2 × 10−5, Fig 2D), and a multifactorially adjusted hazard ratio for early death of 1·66 (0·86–3·21). Finally, the rs10974944 germline polymorphism did not associate with incident disease or mortality (data not shown).

Figure 2.

The JAK2 V617F somatic mutation, cancer risk, and survival in the Danish general population as a function of time after blood sampling. (A) cumulative incidence of any cancer, (B) cumulative incidence of haematological cancer, (C) cumulative incidence of myeloproliferative cancer, (D) cumulative survival. These are based on Kaplan–Meier estimates. The numbers at risk at blood sampling (time = 0) vary due to varying numbers of participants with disease prior to the time of blood sampling.

Table 4. Incident morbidity and mortality in the general population according to JAK2 V617F somatic mutation status
 JAK2 V617F positives versus negatives
Unadjusted analysisMultifactoriallya adjusted analysis
EndpointsAll participants, nJAK2 V617F positives, nEvents in JAK2 V617F positives, nHazard ratio (95%CI)P valueHazard ratio (95%CI)P value
  1. For endpoints other than death, some participants had diagnosis prior to the time of blood sampling and were therefore excluded from the analysis.

  2. CI = confidence interval.

  3. a

    Analyses were adjusted for sex, age, tobacco consumption, alcohol consumption, and body mass index at the time of blood sampling.

Any cancer42 4964584·12 (2·06–8·25)6 × 10−52·45 (1·22–4·92)0·01
Haematological cancer49 33958651·28 (22·59–116·36)5 × 10−2127·56 (11·98–63·44)6 × 10−15
Myeloproliferative cancer49 446583158·72 (47·27–532·99)2 × 10−1697·12 (27·64–341·28)9 × 10−13
Ischaemic heart disease46 7885731·92 (0·62–5·96)0·261·24 (0·39–3·86)0·71
Myocardial infarction48 3916211·45 (0·20–10·34)0·710·81 (0·11–5·74)0·83
Venous thromboembolism48 4946311·46 (0·21–10·39)0·711·04 (0·15–7·39)0·97
Pulmonary embolism49 1446812·85 (0·39–20·29)0·301·91 (0·27–13·62)0·52
Deep venous thrombosis48 754630
Death49 4886893·81 (1·98–7·34)6 × 10−51·66 (0·86–3·21)0·13

Diagnostic value of JAK2 V617F testing

Table 5 shows the diagnostic value of a positive JAK2 V617F test for prevalent myeloproliferative cancer or such cancer diagnosed within 2 years after blood sampling. For the whole study population of 49 488 individuals, the sensitivity was 23%, specificity 100%, positive predictive value 18%, and the negative predictive value was 100%. We also measured the diagnostic value of a positive JAK2 V617F test in individual clinically relevant substrata of the population: specificity and negative predictive value were equal to or close to 100% regardless of the stratification, while sensitivity and positive predictive value improved by pre-selecting individuals with erythrocyte count above 5·5 × 1012/l, or platelet count above 450 × 109/l, or leucocyte count above 15 × 109/l, or erythrocyte volume fraction above 50%, whereas pre-selection of individuals older than 70 years of age diminished the diagnostic value of the test. Thus, by combining a single conventional haematological parameter with a test for the JAK2 V617F somatic mutation, myeloproliferative cancer could be identified or ruled out with a sensitivity of 47–100% and a specificity of 98–100%.

Table 5. Diagnostic value of a positive JAK2 V617F somatic mutation test for myeloproliferative cancer
  n Sensitivity (%)Specificity (%)Positive predictive value (%)Negative predictive value (%)
  1. Included are prevalent myeloproliferative cancer and such cancer diagnosed within 2 years after blood sampling.

  2. The diagnostic value for the JAK2 V617F somatic mutation test was examined individually for each substratum, while ignoring the other parameters.

All participants49 4882310018100
Age >70 years795851006100
Erythrocyte count >5·5 × 1012/l665679943100
Platelet count >450 × 109/l85647982299
Leucocyte count >15 × 109/l14460985099
Erythrocyte volume fraction >50%141100100100100

Discussion

In this cohort study of 49 488 individuals from the Danish general population, we demonstrated that the JAK2 V617F somatic mutation has a high diagnostic value for myeloproliferative cancer when combined with conventional haematological parameters. Using such combinations, specificity of a positive JAK2 V617F somatic mutation test for myeloproliferative cancer was 98–100% and sensitivity 47–100%, reaching clinically useful levels in individuals from the general population with high myeloid cell counts (i.e. elevated erythrocytes, platelets, and non-lymphocytic peripheral leucocytes). These are novel observations.

Another novel observation is that, compared to JAK2 V617F somatic mutation negative subjects, individuals in the general population with a positive JAK2 V617F somatic mutation had higher erythrocyte, platelet, and leucocyte counts, increasing with increasing JAK2 V617F mutation burden. We also observed that JAK2 V617F somatic mutation positives had 2·7-/2·5-fold risks of prevalent/incident cancer, 44-/28-fold risks of haematological cancer, 221-/97-fold risks of myeloproliferative cancer, supported by our own previous findings (Nielsen et al, 2011), as well as 2·2-/1·2-fold risks of ischaemic heart disease and 3·1-/1·0-fold risks of venous thromboembolism, the latter supported by previous studies (Vannucchi et al, 2007; Basquiera et al, 2009; Tefferi et al, 2009). Finally, 68 participants were positive for the JAK2 V617F somatic mutation, corresponding to 0·1% of the general population, with increasing prevalence of the JAK2 V617F somatic mutation across rs10974944 germline genotypes. This is also supported by previous reports (Jones et al, 2009; Kilpivaara et al, 2009; Olcaydu et al, 2009a,b). Likewise, we found an association of the JAK2 V617F somatic mutation with age, which implies that status may change over time. However, this is a single cross-sectional measurement and the value of the marker may be improved if a sequential monitoring approach, with the appropriate analyses, was performed.

Taken together, these results support the following model of the natural, sequential events leading to myeloproliferative cancer in the general population. First, the rs10974944 germline genotype only slightly affects myeloid cell counts in the peripheral blood, but it predisposes to the development of the JAK2 V617F somatic mutation, however, only in very few individuals. Second, the JAK2 V617F somatic mutation positive individuals show clear signs of hyperproliferation of all three myeloid lineages, but only a fraction of these individuals develop overt clinical myeloproliferative cancer or other diseases. Third, this transformation might in turn be a result of yet undiscovered mutational events in the JAK2 V617F mutated stem cell of the bone marrow (de Stefano et al, 2007; Martinaud et al, 2010; Nielsen et al, 2011; Xu et al, 2007).

This hypothetical sequence led us to evaluate the diagnostic value of the test for the JAK2 V617F somatic mutation with respect to myeloproliferative cancer, and whether the diagnostic value could be increased by pre-selecting individuals with easily measured signs of myeloid hyperproliferation. We observed that the specificity and negative predictive value of the test were extremely high, and by pre-selecting individuals with very high cell counts, sensitivity and positive predictive value improved considerably. Interestingly, pre-selecting according to age above 70 years diminished the value of the mutation test. Since approximately 50% of patients with essential thrombocythaemia and primary myelofibrosis do not have the JAK2 V617F somatic mutation (Tefferi & Vardiman, 2008), the negative predictive value of this assay is slightly inflated.

With the inclusion of 49 488 individuals from the Danish general population, the strength of this study is the considerable statistical power. Furthermore, the utility of endpoints collected from the Danish health registries provided 100% follow-up. These registry data, however, are not collected with the specific intent of studying myeloproliferative cancer, and therefore the clinical precision might be somewhat limited: any visit to a general practitioner not resulting in a histological diagnosis is not registered in the Danish Cancer Registry used in the present study. The sensitivity and specificity analysis of participants with haematological abnormalities suggests that many individuals positive for the JAK2 V617F somatic mutation had evidence of a myeloproliferative cancer at the time that the allele burden was measured. As a result, some of the incidence of diagnosis is probably the result of lag between the first documentation of a blood count abnormality and the eventual diagnosis of a myeloproliferative cancer. This potential misclassification of sub-clinical patients into healthy participants in our study would however, only tend to reduce associations and thus cannot explain our positive results. Likewise, despite the inclusion of standard curves and controls with mutation burden of 3% in each batch of the highly sensitive real-time quantitative PCR assay, we cannot completely exclude the possibility that some individuals with mutation burdens <5% were not detected. Again, this potential misclassification of JAK2 V617F somatic mutation positive participants into negatives in our study would however, only tend to reduce associations and thus cannot explain our positive results. Finally, as we studied Whites only, our results may not necessarily apply to other ethnical groups.

Compared to our previous study of another sample from the general population (Nielsen et al, 2011), the present study found a slightly lower prevalence of the JAK2 V617F somatic mutation (0·2 vs. 0·1%) despite employing a more sensitive assay. This apparent difference is, however, not statistically different (P = 0·40), and the population in our former study was slightly older than the present population (median age 59 vs. 56 years), and likewise, the participation ratio was higher for the former than the present study (58% vs. 47%). Given that the frequency of the mutation tends to increase with age and the earlier study (Nielsen et al, 2011) probably included more participants with reduced health, the real difference between the two studies is probably miniscule.

The present study did not describe the clinical phenotype of the 68 JAK2 V617F somatic mutation positives in depth, nor were further additional molecular and/or follow-up tests performed. This, however, will be the aim of future studies.

In conclusion, this study of 49 488 individuals of the Danish general population showed a prevalence of the JAK2 V617F somatic mutation of 0·1% increasing across rs10974944 germline genotypes. We also demonstrated that JAK2 V617F somatic mutation positives versus negatives had a particular haematological phenotype, consisting of higher erythrocyte, platelet, and leucocyte counts that increased with increasing JAK2 V617F mutation burden. Specificity and negative predictive value of a positive JAK2 V617F somatic mutation test for myeloproliferative cancer was 98–100%, whereas sensitivity and positive predictive value only reached clinically useful levels in pre-selected subgroups of the general population with high myeloid cell counts. Our results also suggest a sequential model for myeloproliferative cancer where the rs10974944 germline genotype predisposes to development of the JAK2 V617F somatic mutation in a tiny subset of the carriers of the variant genotype. In turn, this results in hypercellularity of all three myeloid lineages, but only in clinical disease in a subset of these individuals. Thus, by combining these easily measured, peripheral cell counts with the test for the JAK2 V617F somatic mutation, prevalent and incident myeloproliferative cancer can be identified or ruled out with high sensitivity and specificity.

Acknowledgements

The authors would like to thank the staff and participants of the Copenhagen General Population Study for their important contribution and their willingness to participate. We also thank Prof. Dr. W. Fiedler, Universitätsklinikum, Hamburg-Eppendorf, Germany, for providing us with the UKE1 cell line.

This work was supported by the University of Copenhagen; the Foundation of Anders Hasselbalch against Leukaemia; Herlev Hospital, Copenhagen University Hospital; Copenhagen County Foundation; the Chief Physician Johan Boserup and Lise Boserup Foundation; the Danish Cancer Research Foundation, and the Danish Cancer Society. The sponsors had no role in the design and conduct of the study, collection, management, analysis, and interpretation of the data, nor in preparation, review, and approval of the manuscript.

Authorship contributions

CN, HSB, BGN, SEB planned the study. BGN and SEB were responsible for collecting participants. CN was responsible for the laboratory work, supervised by SEB and BGN. CN and SEB analysed data. All authors interpreted data. CN, HSB, BGN, and SEB wrote the report. CN prepared the figures. All authors contributed to the final approved version of this report.

Ancillary