Prognosis and prognostic factors for patients with chronic myeloid leukemia: Nontransplant therapy☆

Article Outline

• Abstract
• The new CML score
• The new CML score and the Sokal score
• Complete hematologic response and prognosis
• Conclusions
• References
• Copyright

Abstract 

Reliable knowledge about an individual's prognosis is needed to select the appropriate treatment for patients with chronic myeloid leukemia (CML). The New CML score using age, spleen size, blast cell count, eosinophil count, basophil count, and platelet count shows good discrimination for survival (96, 65, or 42 months, P ≤ .0001) and has been thoroughly validated. Careful analyses indicate that the New CML score is considerably more precise in identifying high-risk patients than the Sokal score. Achievement of complete hematologic response (CHR) up to 9 months shows a distinct impact on survival, which, however, depends on the baseline prognosis. Ten-year survival probabilities for low- and intermediate-risk patients with a CHR were 0.51 (95% confidence interval [CI], 0.42 to 0.60) and 0.23 (95% CI, 0.15 to 0.31) and without a CHR were 0.26 (95% CI, 0.16 to 0.37) and 0.12 (95% CI, 0.04 to 0.20). In high-risk patients CHR had no impact on prognosis. Therapeutic options were widened by the approval of imatinib for the treatment of CML. However, it will still take 2 or more years to know whether the high rates of CHR and cytogenetic complete remission (CCR) achieved with imatinib translate into a clinically relevant survival advantage for all patients. Semin Hematol 40:4-12. Copyright 2003, Elsevier Science (USA). All rights reserved.

Chronic myeloid leukemia (CML) is still a life-threatening disease and the impact of modern therapeutic approaches on survival time is far from satisfactory. In 1924, the median survival was about 31 months for untreated CML¹³; 75 years later, we have achieved just a doubling of the median life expectancy—not taking into account lead-time bias due to earlier diagnosis. This is not very good news for young patients with CML and the majority of patients are, in fact, diagnosed below the age of 65 years.

Currently, there are three treatment options for patients with CML. Interferon alfa (IFN) has been available since the mid-1980s and increases survival considerably, especially in low-risk patients.⁸ However, due to the need for regular injections and the occurrence of adverse reactions, IFN is considered cumbersome by many patients. The tyrosine kinase inhibitor imatinib mesylate became generally available in 2001. Short-term treatment results have been very impressive: the proportions of complete hematological remission (CHR) and complete cytogenetic remission (CCR) were 95% and 41%, respectively.¹¹, ¹⁷ Imatinib is orally administered and has a promising safety profile. However, long-term results are not yet available, and it is not known whether pretreatment with imatinib adversely influences the success of bone marrow transplantation (BMT). BMT is considered the only curative approach, but the still high transplantation-related mortality rate of about 25% to 30% in the first year⁴ requires careful balance of the risks for the patient of BMT, with special consideration of the prognosis if the patient is responding to conservative treatments like IFN or imatinib. Thus, reliable knowledge about individualized prognosis is needed to select the appropriate treatment that best suits the requirements of the patient and character of the disease.

Back to Article Outline

The new CML score 

Because it is obviously important to define prognostic factors, many CML study groups have worked in this area.⁸ The Sokal score¹⁸ achieved widespread use in the 1990s as a prognostic discriminator for survival in patients treated with chemotherapy (mainly busulfan and hydroxyurea), but it performed less well in patients treated with IFN.⁶, ¹⁴, ¹⁵ Thus the International Collaborative CML Prognostic Factors Project Group developed and validated the New CML score, collecting and analyzing individual data of some 1,300 eligible patients from 14 prospective studies.⁷ The New CML score, occasionally called the Euro score, uses age, spleen size (measured from the left costal margin), blast cell count, platelet count, eosinophil count, and basophil count, all variables measured at the time of diagnosis, resulting in the New CML score formula: New CML score = (0.6666 × age [0 when age < 50 years; 1, otherwise] + 0.0420 × spleen size [cm below costal margin] + 0.0584 × blasts [%] + 0.0413 × eosinophils [%] + 0.2039 × basophils [0 when basophils < 3%; 1, otherwise] + 1.0956 × platelet count [0 when platelets < 1,500 × 10⁹/L; 1, otherwise]) × 1,000.

The score value of individual patients can be calculated by accessing the website www.pharmacoepi.de.

Three distinct prognostic groups were identified.⁷ The low-risk group (score values ≤ 780), comprising about 41% of patients, showed a median survival time of 96 months and a 9-year survival probability of 41% (95% confidence interval [CI], 0.31 to 0.51). The intermediate-risk group (score values > 780 and ≤ 1,480), comprising about 45% of all patients, had a median survival time of 65 months and a 9-year survival probability of 0.16% (95% CI, 0.09 to 0.23). The high-risk group (score values ≥ 1,480), containing 14% of all patients, presented a median survival time of 42 months; no patient had been observed for as long as 9 years.

The current survival status is displayed in Fig 1; the median time of observation for patients still “at risk” is 78 months.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 1. 

  New CML score applied to 1,414 patients with early-stage CML treated with IFN. m.s., median survival time. Confidence intervals for probability of survival are given at 3, 6, and 10 years for each group.

Since its publication, the validity of the New CML score has been confirmed more than once by independent study groups.³, ²⁵

Back to Article Outline

The new CML score and the Sokal score 

With regard to the Sokal score, investigators of the Collaborative CML Prognostic Factors Project Group⁷ confirmed earlier findings.⁶, ¹⁴, ¹⁵ In a large sample of 1,201 patients treated with an IFN-based therapy, Kaplan-Meier curves for intermediate-risk and high-risk patients could not be distinguished. Despite these results, some physicians still apply the Sokal score to samples with IFN-treated CML patients when interested in risk groups prognosticating different survival probabilities. In comparison to the New CML score, one of their main arguments in favor of the Sokal score is that it identifies a larger percentage of high-risk patients.

A cross-sample validation of 493 patients who were not included in the learning sample when the New CML score was developed was performed to investigate differences in risk group allocations of the two methods.⁸ The Sokal score classified 111 of these 493 patients (23%) as high-risk. In contrast, only 45 patients (9%) were high-risk according to the New CML score. The 63 of the 111 patients at high risk according to the Sokal score but intermediate risk according to the New CML score had a median survival of 67 months, whereas the median survival of the 38 patients with high risk according to both scores was 43 months, or 24 months less. The survival probabilities of the 63 and the 38 patients (both high risk according to Sokal score) were statistically significantly different (log-rank test, P = .0105). Adding the 10 patients who were low risk according to the New CML score, the actual survival probabilities suggested an obvious misclassification of 73 of 111 patients (66%) by the Sokal score. The conclusion from this empirical result was that the New CML score reflects reality more closely than does the Sokal score; currently, only a small percentage (9% to 15%) of patients appear to be at high risk with regard to an IFN-based therapy.

Pfirrmann and Hasford¹⁶ offered an explanation for the misleading high-risk group classification in patients of the Italian Study Group on CML.² In this sample, the high-risk groups defined by both scores had a median survival of 45 months. However, 15 of 62 patients identified as high risk by the Sokal score had a survival time longer than 60 months, while this applied to only four of 25 with the New CML score. This example demonstrates the limitation of “median survival time”: it describes “the moment in time when survival probability was 0.5” but does not necessarily have any bearing on subsequent survival probabilities for patients who are still under observation after this time point. Apart from empirical results, methodical reasoning also mitigates against further use of the Sokal score. Because the Sokal score was developed from a sample of chemotherapy-treated patients, it does not adequately discriminate among IFN-treated patients. Moreover, Sokal et al¹⁸ apparently defined the boundaries of their risk groups somewhat arbitrarily.¹⁶ They “divided into three subgroups of roughly similar size, using hazard ratios of 0.8 and 1.2 as boundaries,” probably to obtain subgroups of similar size and, perhaps, for boundaries that were equidistant from the hazard ratio 1.0.¹⁸ In contrast, the boundaries of the New CML score were gained by the “minimal P value” approach,¹ a statistical procedure that looks for boundaries that maximize the differences between the resulting groups and the relevant outcome variable. Thus, these boundaries are derived from the data and not from the investigators' assumptions.⁷ The definition of risk groups that reflect the “real” percentages of patients who should be classified at high risk or low risk is much less likely using the Sokal score than by the statistical approach of the New CML score.⁷ In addition, Sokal's assignment of mean values for missing items “may have introduced a slight bias against prognostic discrimination.”¹⁸

In conclusion, we recommend that the use of the Sokal score to identify significantly different risk groups with regard to survival should be discontinued. A patient wrongly allocated to the high-risk group by the Sokal scoring system may not receive IFN and may thus forfeit the chance to become a hematologic or even cytogenetic responder, both of which could for IFN-treated patients be associated with a significantly longer survival.

Back to Article Outline

Complete hematologic response and prognosis 

To allow for an even more detailed assessment of prognosis building on the New CML score, we analyzed the impact of achieving a CHR on survival. CHR was defined as a white blood cell count ≤ 10 × 10⁹/L, a normalized platelet count ≤ 450 × 10⁹/L, no splenomegaly, blasts + promyelocytes + myelocytes + meta-myelocytes = 0%, and no CML-related symptoms.²⁰ The sample for this analysis consisted of patients from nine studies whose baseline and survival data had already been used as part of the data collected for the extraction and validation of the New CML score⁷: Austrian CML III²¹ and V,²² Benelux Study Group,²³ German CML I⁹ and II,¹⁰ Italian CML 86,²⁴ Kouseisho Study Group,¹⁴ Spanish Study Group,¹⁹ and French CML 88.⁵ Investigators from these studies and from the trial of Mahon et al¹² had the opportunity to provide updated survival times and hematologic response information defined by Talpaz et al.²⁰ Inclusion and exclusion criteria were the same as previously published,⁷ and values for the baseline characteristics remained comparable.

The New CML score was available for 1,039 patients in the 10 studies. A comparison of the survival probabilities for the 1,009 patients for whom data on hematologic response under IFN therapy was also available (median survival, 74 months) with the 30 patients with no information on hematologic response (median survival, 45 months) indicated that these data were not missing at random (log-rank test, P = .0033; stratified according to New CML score, P = .0022). Hematologic data had obviously not been obtained in some cases of early progression. Since the 30 patients amounted to only 2.9% of the 1,039, survival probabilities with or without these patients were very similar, and our further restriction to the 1,009 patients with complete data did not lead to a serious bias.

Among the 1,009 patients, 651 (65%) attained a CHR. The first CHR was observed within the first 6 months after the start of IFN therapy in 418 patients (64% of 651), within the first 9 months in 544 (84%), and within the first year in 583 (90%). Figure 2A depicts the survival probabilities of all 1,009 patients to the end of the sixth month.

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 2. 

  Patients with data on the New CML score and on hematologic response (n = 1,009). During the first (A) 6, (B) 9, and (C) 12 months after start of IFN therapy, survival probabilities of all patients were calculated together. For the (A) 997, (B) 966, and (C) 940 patients still under observation after 6, 9, and 12 months, respectively, 2 Kaplan-Meier curves were plotted from the landmark. The assignment of a patient to either curve depended on whether a CHR was recorded within the first 6, 9, or 12 months of IFN therapy, or not. Confidence intervals for probability of survival are given at 3, 6, and 10 years for each group. Log-rank test between “CHR” and “No CHR” for survival probabilities after 6, 9, and 12 months: P ≤ .0001 for all.

Twelve patients (five with CHR, seven without CHR; two deaths in each group) were observed for less than 6 months; the remaining 997 patients were then stratified for CHR (yes or no). Median survival for complete responders was 85 months and for patients without CHR, 65 months (log-rank test, P ≤ .0001). The same investigations were performed for the landmarks 9 and 12 months (Fig. 2, Fig. 2, respectively). Compared to the result at 6 months, CHR was recorded for 126 additional patients up to the end of the ninth month, but the observation time of 31 patients (14 with CHR after 6 months and 17 without CHR after 6 months) ended during the third quarter of year 1. Of the latter 17 patients, six received a BMT in first chronic phase, four had censored survival times for no treatment-related reasons, and seven died during the 3-month interval. These seven patients were the ones who might have been considered at disadvantage due to the physicians' decision to continue IFN treatment, although they had not achieved a CHR within the first 6 months. However, we believe the amount of additional information on CHR outweighs the drawback of the seven deaths. Compared to the earlier landmark, median survival differed by 6 more months (Fig 2B) and the chi-square statistic was doubled (log-rank test, P ≤ .0001). Grouping together the survival probabilities of 126 patients who gained, the 10 without CHR who were censored, and the seven who died while waiting 3 more months (65 of 143 died), the median survival was 88 months, slightly longer than the 86-month median survival of the 399 patients (169 died) who had already achieved a complete remission within 6 months. The comparison of the two survival curves suggests that earlier achievement of a CHR within 6 months is not associated with a better prognosis. During the last quarter of the first treatment year, a first CHR was observed for 39 patients while for 20 patients (12 received a BMT, one had the last follow-up, seven died) observation time ended. The median survival time of all 59 patients was 82 months (22 died) and 87 months for the 519 patients with a CHR before 9 months (226 died). Table 1 gives an overview of the stratification according to CHR (yes/no) at each landmark. In summary, the benefit of waiting for 9 months instead of 6 months was assessed to be medically very valuable. Compared to the results after 9 months, waiting for 1 year added much less information and was considered to be too long to wait for hematologic remission—true particularly now because not only hematologic but also cytogenetic examinations are performed and imatinib is available.

Table 1. Complete hematologic response after 6, 9, and 12 months

Applying the New CML score to the investigated 1,009 patients, Kaplan-Meier curves and median survival times of low-risk (n = 446, 153 died; median survival, 100 months), intermediate-risk (n = 452, 270 died; median survival, 68 months), and high-risk group (n = 111, 85 died; median survival, 43 months) patients were very similar to those for all 1,414 patients. Tables 2 through 4 show the results on survival for each combination between the risk group evaluated at diagnosis and the hematologic response status at 6, 9, and 12 months, respectively.

Table 2. New CML score and complete hematologic response up to month 6

Table 4. New CML score and complete hematologic response up to month 12

For the reasons outlined above, we focused on the landmark 9 months. Whereas all 446 low-risk patients had a 10-year survival probability of 0.41 and all patients with CHR up to 9 months one of 0.34 (Table 1), the combination of both resulted in a group of 262 patients with a 10-year survival of 0.51 (Table 3). In comparison to 163 low-risk patients without CHR after 9 months, the survival probabilities of the 262 were statistically significantly higher (median survival, 123 months v 76 months; log-rank test, P ≤ .0001, Fig 3).

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 3. 

  Patients with low risk at diagnosis according to the New CML score (n = 446) and with data on hematologic response. During the first 9 months after start of IFN therapy, survival probabilities of all patients were calculated together. For the 425 patients still under observation after 9 months, 2 Kaplan-Meier curves were plotted from this landmark. The assignment of a patient to either curve depended on whether a CHR was recorded within the first 9 months of IFN therapy or not. Confidence intervals for probability of survival are given at 3, 6, and 10 years for each group. Log-rank test between “CHR” and “No CHR” for survival probabilities after 9 months: P ≤ .0001.

The same significance level was attained when 224 responders (median survival, 76 months) and 211 nonresponders (median survival, 57 months) within the intermediate-risk group were considered (Table 3 and Fig 4).

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 4. 

  Patients with intermediate risk at diagnosis according to the New CML score (n = 452) and with data on hematologic response. During the first 9 months after start of IFN therapy, survival probabilities of all patients were calculated together. For the 435 patients still under observation after 9 months, 2 Kaplan-Meier curves were plotted from this landmark. The assignment of a patient to either curve depended on whether a CHR was recorded within the first 9 months of IFN therapy, or not. Confidence intervals for probability of survival are given at 3, 6, and 10 years for each group. Log-rank test between “CHR” and “No CHR” for survival probabilities after 9 months: P ≤ .0001.

No significant difference was found within the high-risk patients (Fig 5).

• 
  • View Large Image
  • Download to PowerPoint

• Fig. 5. 

  Patients with high risk at diagnosis according to the New CML score (n = 111) and with data on hematologic response. During the first 9 months after start of IFN therapy, survival probabilities of all patients were calculated together. For the 106 patients still under observation after 9 months, 2 Kaplan-Meier curves were plotted from this landmark. The assignment of a patient to either curve depended on whether a CHR was recorded within the first 9 months of IFN therapy, or not. Confidence intervals for probability of survival are given at 3, 6, and 10 years for each group. Log-rank test between “CHR” and “No CHR” for survival probabilities after 9 months: P = .0945.

Since the time-dependent variable “CHR (yes or no) up to 9 (and also 6 and 12) months” was twice able to differentiate two groups with statistically different survival probabilities within one baseline-risk group, within the low-risk as well as within the intermediate-risk group, we concluded that CHR provided independent additional prognostic information of clinical relevance. However, in high-risk patients, achievement of CHR did not impact survival (Fig 5).

Table 3. New CML score and complete hematologic response up to month 9

Back to Article Outline

Conclusions 

The New CML score has shown good discrimination of survival and has been thoroughly validated. The comparatively small fraction of high-risk patients identified with the New CML score seems to represent the true frequency. The analyses show the achievement of CHR to have considerable impact on the prognosis of chronic-phase CML treated with IFN, depending on the baseline prognosis.

Since imatinib has received approval for the treatment of CML, therapeutic options have widened and physicians' and patients' preferences have changed dramatically. As the first phase I study with imatinib only started in June 1998, long-term results are not yet available. In addition, the baseline data needed for the New CML score were not recorded in the imatinib studies, partly due to the fact that initially only patients in blast crisis or in chronic phase resistant to IFN were recruited. Thus, the prognostic profile of the patients treated with imatinib has not been known. Therefore, it is very difficult—if not impossible—to base therapeutic decisions on sound empiric data. Previous experience has made us aware that the validity of prognostic models depends on specified treatments, and the value of surrogate markers like CHR or CCR may depend on baseline risk. However, it will still take 2 or more years before we know whether the high rates of CHR and CCR achieved with imatinib translate into a clinically relevant survival advantage for all patients.

Back to Article Outline

References 

1. Altman DG, Lausen B, Sauerbrei W, et al.  Dangers of using “optimal” cutpoints in the evaluation of prognostic factors. J Natl Cancer Inst. 1994;86:829–835
   • View In Article
   • MEDLINE
   • CrossRef
2. Bonifazi F, De Vivo A, Rosti G, et al.  Testing Sokal's and the new prognostic score for chronic myeloid leukaemia treated with α-interferon. Br J Haematol. 2000;111:587–595
   • View In Article
   • MEDLINE
   • CrossRef
3. Bonifazi F, De Vivo A, Rosti G, et al.  Chronic myeloid leukemia and interferon-alpha: A study of complete cytogenetic responders. Blood. 2001;98:3074–3081
   • View In Article
   • MEDLINE
   • CrossRef
4. Gratwohl A, Hermans J, Goldman JM, et al.  Risk assessment for patients with chronic myeloid leukemia before allogeneic blood or marrow transplantation. Lancet. 1998;352:1087–1092
   • View In Article
   • Abstract
   • Full Text
   • Full-Text PDF (85 KB)
   • CrossRef
5. Guilhot F, Chastang C, Michallet M, et al.  Interferon alfa-2b combined with cytarabine versus interferon alone in chronic myelogenous leukemia. French Chronic Myeloid Leukemia Study Group. N Engl J Med. 1997;337:223–229
   • View In Article
   • MEDLINE
   • CrossRef
6. Hasford J, Ansari H, Pfirrmann M, et al.  Analysis and validation of prognostic factors for CML. Bone Marrow Transplant. 1996;17(suppl 3):S49–S54
   • View In Article
7. Hasford J, Pfirrmann M, Hehlmann R, et al.  A new prognostic score for survival of patients with chronic myeloid leukemia treated with interferon alfa. J Natl Cancer Inst. 1998;90:850–858
   • View In Article
   • MEDLINE
   • CrossRef
8. Hasford J, Pfirrmann M, Hehlmann R, et al.  Prognostic factors. In:  Carella A,  Daley G,  Eaves C, et al. editor. Chronic Myeloid Leukemia—Biology and Treatment. London, UK: Martin Dunitz; 2001;p. 205–223
   • View In Article
9. Hehlmann R, Heimpel H, Hasford J, et al.  Randomized comparison of interferon-alpha with busulfan and hydroxyurea in chronic myelogenous leukemia. The German CML Study Group. Blood. 1994;84:4064–4077
   • View In Article
   • MEDLINE
10. Hehlmann R, Heimpel H, Hossfeld DK, et al.  Randomized study of the combination of hydroxyurea and interferon alpha versus hydroxyurea monotherapy during the chronic phase of chronic myelogenous leukemia (CML Study II). The German Study Group. Bone Marrow Transplant. 1996;17(suppl 3):S21–S24
    • View In Article
11. Kantarjian H, Sawyers C, Hochhaus A, et al.  Hematologic and cytogenetic responses to imatinib mesylate in chronic myelogenous leukemia. N Engl J Med. 2002;346:645–652
    • View In Article
    • CrossRef
12. Mahon FX, Fabéres C, Pueyo S, et al.  Response at three months is a good predictive factor for newly diagnosed chronic myeloid leukemia patients. Blood. 1998;92:4059–4065
    • View In Article
    • MEDLINE
13. Minot GR, Buckman TE, Isaacs R. Chronic myelogenous leukemia. J Am Med Assoc. 1924;82:1489–1494
    • View In Article
14. Ohnishi K, Ohno R, Tomonaga M, et al.  A randomized trial comparing interferon-alpha with busulfan for newly diagnosed chronic myelogenous leukemia in chronic phase. Blood. 1995;86:906–916
    • View In Article
    • MEDLINE
15. Ozer H, George SL, Schiffer CA, et al.  Prolonged subcutaneous administration of recombinant alpha 2b interferon in patients with previously untreated Philadelphia chromosome-positive chronic-phase chronic myelogenous leukemia. Effect on remission duration and survival: Cancer and Leukemia Group B Study 8593. Blood. 1993;82:2975–2984
    • View In Article
    • MEDLINE
16. Pfirrmann M, Hasford J. Testing Sokal's and the new prognostic score for chronic myeloid leukaemia treated with α-interferon: Comments. Br J Haematol. 2001;114:241–242 (letter)
    • View In Article
    • MEDLINE
    • CrossRef
17. Savage DG, Antman KH. Imatinib mesylate—A new oral targeted therapy. N Engl J Med. 2002;346:683–693
    • View In Article
    • CrossRef
18. Sokal JE, Cox EB, Baccarani M, et al.  Prognostic discrimination in “good-risk” chronic granulocytic leukemia. Blood. 1984;63:789–799
    • View In Article
    • MEDLINE
19. Steegmann JL, Odriozola J, Rodriguez-Salvans F, et al.  Stage, percentage of basophils at diagnosis, hematologic response within six months, cytogenetic response in the first year: the main prognostic variables affecting outcome in patients with chronic myeloid leukemia in chronic phase treated with interferon-alpha. Results of the CML89 trial of the Spanish Collaborative Group on interferon-alpha-2a and CML. Haematologica. 1999;84:978–987
    • View In Article
    • MEDLINE
20. Talpaz M, Kantarjian HM, McCredie KB, et al.  Clinical investigation of human alpha interferon in chronic myelogenous leukemia. Blood. 1987;69:1280–1288
    • View In Article
    • MEDLINE
21. Thaler J, Gastl G, Fluckinger T, et al.  Treatment of chronic myelogenous leukemia with interferon alfa-2c: Response rate and toxicity in a phase II multicenter study. The Austrian Biological Response Modifier (BRM) Study Group. Semin Hematol. 1993;30(suppl 3):S17–S19
    • View In Article
22. Thaler J, Hilbe W. Comparative analysis of two consecutive phase II studies with IFN-alpha and IFN-alpha + ara-C in untreated chronic-phase CML patients. Austrian CML Study Group. Bone Marrow Transplant. 1996;17(suppl 3):S25–S28
    • View In Article
23. The Benelux CML Study Group: Randomized study on hydroxyurea alone versus hydroxyurea combined with low-dose interferon-alfa-2b for chronic myeloid leukemia. Blood. 1998;91:2272–2713
    • View In Article
    • MEDLINE
24. The Italian Cooperative Study Group on Chronic Myeloid Leukemia . Interferon alfa-2a as compared with conventional chemotherapy for the treatment of chronic myeloid leukemia. N Engl J Med. 1994;330:820–825
    • View In Article
    • MEDLINE
    • CrossRef
25. Thomas MJ, Irving JEA, Lennard AL, et al.  Validation of the Hasford score in a demographic study in chronic granulocytic leukemia. J Clin Pathol. 2001;54:491–493
    • View In Article
    • MEDLINE
    • CrossRef