This Article Abstract Full Text (PDF) eLetters: Submit a response to this article Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services E-mail this article link to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints/Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Littlewood, T. J. Articles by Perkins, A. Search for Related Content PubMed PubMed Citation Articles by Littlewood, T. J. Articles by Perkins, A.

Baseline and Early Treatment Factors are not Clinically Useful for Predicting Individual Response to Erythropoietin in Anemic Cancer Patients

^a John Radcliffe Hospital, Headington, Oxford, UK; ^b Medical Affairs, Ortho-Biotech, High Wycombe, UK; ^c Goulds, Staverton, Totnes, Devon, UK

Correspondence: Timothy J. Littlewood, M.D., John Radcliffe Hospital, Headington, OX3 9DU Oxford, UK. Telephone: 44-0-1865-220364; Fax: 44-0-1865-221778; e-mail: Tim.Littlewood{at}btinternet.com

	LEARNING OBJECTIVES

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 
After completing this course, the reader will be able to:

1. Recognize that anemia is common in patients with malignant disease.
   
2. Appreciate that treatment with recombinant erythropoietin can increase the hemoglobin concentration by 2g/dL or more in 70% of patients at the usual doses and frequency.
   
3. Recognize that factors that predict for a hemoglobin response lack sensitivity and specificity.
   

	ABSTRACT

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 
Recombinant human erythropoietin (rHuEPO) is an effective treatment for anemia in patients with cancer, and recent studies show that over two-thirds of patients can be expected to respond with a large increase (>2 g/dl) in hemoglobin concentration. However, it would be helpful to identify likely responders and nonresponders before initiating treatment. Previous studies have suggested that high pretreatment endogenous erythropoietin levels are associated with a lower response to erythropoietin, especially in certain patient groups, such as patients with hematological malignancies, nonchemotherapy patients, or patients with myelodysplastic syndrome. Various algorithms have therefore been developed to predict patient response to rHuEPO using baseline serum erythropoietin levels and other baseline factors. We performed an analysis of data pooled from four randomized clinical trials of 604 patients with nonmyeloid malignancies, examining the clinical usefulness of pretreatment and early treatment factors for predicting response to erythropoietin. The analysis confirms several other reports that the most predictive models combined pretreatment and early treatment factors, including change in hemoglobin at 4 weeks, but even these models did not increase sensitivity above 85% (total response in unselected patients was 68.1%), while specificity remained poor. We conclude that clinically useful prediction of response to erythropoietin is not possible using baseline or early response variables because of poor sensitivity and specificity of prediction compared with generally accepted clinical tests.

Key Words. Erythropoietin • Erythropoiesis • Anemia • Cancer • Hemoglobin

	INTRODUCTION

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 
Anemia occurs frequently in cancer patients, either as a result of the disease or its therapy, and can have a major impact on quality of life, which often is underestimated by clinicians [1]. Anemia may also reduce the effectiveness of anticancer therapy [2]. Recombinant human erythropoietin (rHuEPO; epoetin alfa) is an effective treatment for anemia in many cancer patients, but response rates are variable [3, 4]. Individual studies have shown response rates of 32%-85%, and overall, about 70% of patients will have an increase in hemoglobin (Hb) concentration of >2 g/dl [3, 5, 6].

In order to direct the potential benefit of rHuEPO treatment to those most likely to have the greatest response, it would be useful to be able to identify nonresponders before the start of treatment or within the first few weeks of treatment. In cancer patients, the serum EPO level is often not appropriately elevated in the presence of anemia [7], and previous studies have suggested that poor response to rHuEPO in cancer patients may be related to high pretreatment serum EPO concentrations [8–10]. This has led to the development of a number of predictive algorithms [11–13]. Unfortunately, the applicability of the findings is limited by small sample sizes and/or analysis of data from specific tumor types.

In several studies, pretreatment and early treatment factors have been examined for their usefulness in predicting a response to EPO, often in combination with pretreatment serum EPO concentrations. Several authors have suggested that initial reticulocyte and Hb response to a short course of EPO treatment (2 to 4 weeks) is the most accurate predictor of eventual response [4, 6, 14, 15]. Various measures of iron status have also been put forward as potential predictors of response. To our knowledge, no study has yet verified these proposed algorithms in large studies enrolling a heterogeneous group of cancer patients, the type of population that would most likely be encountered in standard practice.

The use of epoetin alfa has now been reported in several large, published randomized trials of patients with a range of malignancies, and a substantial dataset is available to reassess the relationship between treatment response and baseline and early response variables that may predict response. We have analyzed data pooled from four studies to determine the relationship, if any, between a large number of pre- and early treatment factors and response to epoetin alfa treatment [3, 16, 17] (Abstract/Ortho-Biotech data on file for the remaining study).

	SUBJECTS AND METHODS

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 
Study Population
Data were pooled from the four pivotal, randomized, double-blind, placebo-controlled multicenter studies used to register epoetin alfa in the U.S. and Europe. The results of all but one of these studies are reported individually elsewhere [3, 16, 17]. A total of 604 patients, treated with 150 IU/kg rHuEPO administered subcutaneously three times per week, was included in these studies. Of these, 561 patients were clinically evaluable and were included in the actual pooled analysis. In all but one study, the dose was increased to 300 IU/kg after 4 weeks if response was unsatisfactory (usually defined as an increase in Hb concentration <1 g/dl); in the remaining study, the dose was not increased but could be titrated down to achieve a hematocrit of 38%-40%.

The pooled population comprised patients aged 18-92 years (median, 62 years): 361 (64%) were women; 292 (52%) had solid tumors; 217 (39%) had bone marrow involvement; and 190 (34%) were transfusion dependent at baseline. The most common types of cancer were breast cancer (23%), myeloma (20%), and lymphoma (16%). Treatment efficacy was defined by abrogation of transfusion requirements or change in Hb concentration (or hematocrit in the U.S. study).

Measurement of Pre- and Early Treatment Tests
In each study, Hb concentration and other hematological parameters were measured locally before the start of therapy and early during the course of treatment (2 and/or 4 weeks after the start of the trials). To eliminate intrastudy center-to-center variability in serum EPO measurements, serum EPO concentrations for each study were measured at a single centralized laboratory using an enzyme-linked immunosorbent assay method. All other laboratory examinations were performed locally according to good clinical practice.

Statistical Analysis
Stepwise logistic regression (PROC LOGISTIC function of SAS statistical software) was used to determine which of the baseline and early response factors in Table 1 were statistically associated with response. Two models were examined: one in which only baseline and demographic factors were incorporated into the model, and one that also included early response variables.

Variables were selected for inclusion in the analysis if they had been present in other studies that addressed EPO response or measured a value associated with hematological status, iron status, or chemotherapy treatment. Cut-off values were selected to match values that had previously been used in the literature, and in some cases, multiple cut-off values were examined. Single-factor sensitivities and specificities were examined for all of the above baseline and early response variables and all cut-off values.

Two-factor analyses were evaluated that combined the factors shown to be most predictive from the logistic regression and one-factor analyses. For completeness, three-factor analyses were also evaluated that looked at the predictive value of the most favorable factors from the one- and two-factor analyses. The effect of prior chemotherapy was addressed by looking separately at the predictive ability of several factors in patients who had and had not received prior chemotherapy.

	RESULTS

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 
Overall Response
From the four studies, data suitable for evaluation were available for 561 of the 604 patients in the combined database who received rHuEPO in the four randomized clinical trials.

Overall, 68.1% of patients (382) responded to rHuEPO treatment according to hematological parameters (increase in Hb 2 g/dl or increase in hematocrit 6%), while 79.1% of participants (444) receiving rHuEPO were transfusion free after 4 weeks of treatment.

Logistic Regression
The logistic regression model incorporating only baseline and demographic factors revealed ferritin 400 ng/ml and baseline rHuEPO level 100 mU/ml to be statistically related to Hb response (p = 0.0002 and p = 0.0037, respectively). The logistic regression incorporating baseline and early response variables identified Hb change at 4 weeks as the single most important predictor of response (p < 0.0001). Baseline serum ferritin remained statistically significant in the model including early response factors, while baseline serum EPO was no longer statistically significant. Hence, logistic regression suggested that a model using baseline ferritin levels and change in Hb at week 4 should be most predictive. The sensitivity and specificity of these models are examined below along with the other factors.

Single-Factor Analysis

Baseline Factors
The results of the single-factor analyses are presented in Table 2. When baseline factors were examined for their ability to predict a response to rHuEPO, it was clear that no single factor showed much improvement in sensitivity compared with the overall rate of response (68%) in these patients. Consistent with the logistic regression, baseline EPO level 100 mU/ml and baseline ferritin 400 ng/ml appeared to have the highest sensitivities as predictive factors, but using these factors would only increase the rate of positive response to 73.8% and 71.9%, respectively.

Early Response Variables
The use of early response variables to predict response identified rates of predicted response slightly higher than baseline factors at 2 weeks, and somewhat higher at 4 weeks. Consistent with the logistic regression, change in Hb >1 g/dl by 4 weeks rose above the 80% level. Even for this factor, which is directly linked to the ultimate definition of response, 52% of patients who failed to have a 1-g/dl rise in Hb by 4 weeks still went on to have a 2-g/dl rise in 8 weeks, again showing poor specificity. Youden’s Index for the use of Hb change >1 g/dl by 4 weeks was 66.2% compared with a Youden’s Index of 68.1% for treatment of all patients without any testing.

Two-Factor Analysis
Two-factor analyses were constructed from the best predictive factors in the one-factor and logistic regression data to try to test the sensitivity and specificity of the predictive tests. The results of these tests are shown in Table 3.

The specificity of the two-factor tests was only slightly higher than for the best single factors, with the two-factor tests including Hb at 4 weeks having a specificity of 53.5%-56.4% for patients having neither factor, meaning that 43.6%-46.5% of patients not meeting either criterion would still respond. The two additional groups of patients having "mixed" test results had response rates between the two extremes.

Three-Factor Analysis
To further determine if adding additional factors could result in more predictive accuracy, three-factor analyses were performed for each of the factors measured at baseline, 2 weeks, and 4 weeks (Table 4).

Effect of Prior Chemotherapy
The overall response rates to EPO therapy in patients who had and had not received prior chemotherapy were 67.4% and 70.7%, respectively. Table 5 shows the results for serum EPO level at baseline and Hb change at 4 weeks. There were no meaningful differences between the groups, and poor specificity was noted no matter what subgroup of patients was used for analysis. These analyses indicate that prestudy chemotherapy is not a significant factor affecting the ability to use predictive testing for rHuEPO response.

	DISCUSSION

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

Some studies have, however, suggested that prediction of response is possible. Ludwig et al. [11] stated, based on a study of 80 patients, that "unresponsiveness of the patient is very likely, (predictive power, 93%)" where EPO level is 100 mU/ml and Hb increase is <0.5 g/dl after 2 weeks. However, we found, in evaluating 356 patients, that 48.7% of patients meeting these criteria still responded to rHuEPO. Ludwig et al. also found that serum ferritin 400 ng/ml after 2 weeks of rHuEPO therapy "strongly indicates unresponsiveness (predictive power, 88%)." Yet our evaluation of 116 patients showed that 57.4% of patients meeting this criterion still responded. It is interesting to note that Ludwig et al. identified the same best positive predictive factors as we did (baseline serum EPO level ferritin, and early Hb change), but the variability of results between studies suggests that these tests still lack the validation and precision required in order to be adopted into general practice for decision making.

We have also illustrated the difficulty posed by multiple test outcomes with multifactor models. The more test factors that are used together, the more patients will have conflicting test results that make application of any algorithm very difficult. In most of the two-factor models, roughly half the patients fell into this gray area where test result interpretation is difficult. Most publications in this area, however, have ignored patients with conflicting test results, preferring to concentrate on patients who either had all or none of the positive factors for prediction.

Another set of issues that has largely been ignored but should be the subject of further study is the precision of Hb testing and whether Hb response is the best measure of response for all patients. If an increase of 2.0 g/dl in Hb is used to define response, then errors in the measurement of Hb of just a few tenths of a point may have a profound impact on the outcome of testing. Adding to this is the fact that for a patient experiencing a rapid decline in Hb at the time of initiation of rHuEPO treatment, an increase in Hb of "only" 1.0 g/dl might be considered a good response for that patient. The study by Abels et al. showed that transfusion frequency was highly correlated with decline in Hb in the first two cycles of chemotherapy, suggesting that patients not meeting the standard definition of Hb response do benefit from a prevention of a decline in Hb [16]. Based on both of these issues, the use of rigid tests and cut-offs to define response and the need for continued treatment may be ignoring the more important clinical issues that are specific to individual patients.

It is possible that further investigation may yet uncover groups of patients for whom prediction of response to rHuEPO is still possible. For example, it is still possible that chemotherapy exposure itself may alter the prognostic value of some of the tests, as there is evidence that EPO levels increase after chemotherapy [21–23]. This observation, acting in certain populations, may explain Abels’ observation that pretreatment EPO concentrations were related to rHuEPO response only in those patients who were not receiving chemotherapy [16]. The results of the present study suggest, however, that many factors that have been offered as predictive in the context of one population or study may not be generalizable to other cancer populations. This appears to be true regarding pretreatment chemotherapy, as we found no meaningful association between pretreatment chemotherapy and the results of predictive factor analyses. As a result, more study in this area is required before predictive testing is used on patients in an actual clinical setting.

It could also be argued that better algorithms might exist for certain tumor types but not others. Ludwig et al. [12] examined this question but found that disease-related factors such as tumor type, bone marrow involvement, and type of chemotherapy did not influence rHuEPO response. Other large studies also suggest that the response rates of different tumor types to rHuEPO are fairly constant [3, 5, 6, 17]. Cazzola et al. restricted their analysis to patients with multiple myeloma and non-Hodgkin’s lymphoma and were able to define a "ratio of observed to predicted logarithm of baseline serum erythropoietin level (O/P ratio)" that appears to be more predictive than we found in our analysis, but their study was actually designed as a dose-response study and used daily dosing of rHuEPO, a dosing regimen that is not commonly used in practice or studied in any other trials [24]. In addition, only 55 patients were exposed to therapeutic doses of rHuEPO. Despite the fact that the O/P ratio was their most predictive factor, only 70% of patients meeting the criteria responded (compared with 61% overall), while 27% who did not meet the criteria still responded. This rate of sensitivity does not vary much from those presented in this and other studies and was not much greater than their baseline response rate. In short, these response criteria did not appear to be clinically useful when put into practice despite the fact that they were statistically the "best" criteria for response.

For any predictive algorithm to be clinically useful, it should accurately predict those patients with a high probability of responding, but must also provide a high degree of certainty that those denied treatment would not have responded. Although some of the predictive factors in this study allow somewhat improved selection of responders, none appear to be useful to exclude the likelihood of response. One strategy, presented recently at a meeting of the American Society of Clinical Oncology (2002) separately by Patton et al. and Glaspy et al., is using higher initial doses of EPO in an attempt to increase the rate of response and identify nonresponders more quickly, but more detailed data are needed in order to compare this approach with standard dosing regimes.

	CONCLUSION

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 
Based on previous studies and our own analysis of over 500 patients, we conclude that a variety of common baseline and early response variables are not useful for predicting the response to rHuEPO since they do not markedly improve sensitivity while having very poor specificity. None of the tests approached the 80%-90% levels of sensitivity and specificity that are generally regarded as appropriate for clinically useful predictive tests or screening techniques. The best models contained some early evaluation of response at either 2 or 4 weeks, but even these models, while somewhat more sensitive, suffered from poor specificity, and upwards of 40% of patients still went on to respond to treatment. It is perhaps open to debate what the sensitivity and specificity cut-offs should be for clinically useful tests in different circumstances, but a 40% or greater false-negative rate would appear to be unacceptable.

For the time being, it appears that no simple test can yet substitute for clinical judgment in managing anemia in cancer patients. Although, ideally, the response to all medical treatments could be better predicted in advance, a response rate of 70% in this patient population is quite high compared with the response rate to other treatments. Use of a predictive algorithm with a >50% false-negative rate would potentially deny effective treatment to too many patients, and further study in this area is needed before such tests become useful.

	ACKNOWLEDGMENT

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 
This meta-analysis and the original studies were funded by the R.W. Johnson Pharmaceutical Research Institute. We thank Elizabeth Wager for editorial assistance.

M.Z. and C.P. are employees of Ortho-Biotech. T.L. is a paid consultant for Ortho-Biotech and receives honoraria from Roche and Amgen. A.P. is a consultant statistician for Ortho-Biotech.

	REFERENCES

Top Learning Objectives Abstract Introduction Subjects and Methods Results Discussion Conclusion References

 

1. Vogelzang NJ, Breitbart W, Cella D et al. Patient, caregiver, and oncologist perceptions of cancer-related fatigue: results of a tripart assessment survey. The Fatigue Coalition. Semin Hematol 1997;34(suppl 2):4–12.[Medline]
2. Mercadente S. Pharmacotherapy of anemia in cancer patients. Expert Opin Pharmacother 2001;2:1949–1961.[CrossRef][Medline]
3. Littlewood TJ, Bajetta E, Nortier JW et al. Effects of epoetin alfa on hematological parameters and quality of life in cancer patients receiving nonplatinum chemotherapy: results of a randomized, double-blind, placebo-controlled trial. J Clin Oncol 2001;19:2865–2874.[Abstract/Free Full Text]
4. Beguin Y. A risk-benefit assessment of epoetin in the management of anemia associated with cancer. Drug Saf 1998;19:269–282.[CrossRef][Medline]
5. Glaspy J, Bukowski R, Steinberg D et al. Impact of therapy with epoetin alfa on clinical outcomes in patients with nonmyeloid malignancies during cancer chemotherapy in community oncology practice. J Clin Oncol 1997;15:1218–1234.[Abstract/Free Full Text]
6. Demetri GD, Kris M, Wade J et al. Quality-of-life benefit in chemotherapy patients treated with epoetin alfa is independent of disease response or tumor type: results from a prospective community oncology study. J Clin Oncol 1998;16:3412–3425.[Abstract]
7. Miller CB, Jones RJ, Piantadosi S et al. Decreased erythropoietin response in patients with the anemia of cancer. N Engl J Med 1990;322:1689–1692.[Abstract]
8. Barlogie B, Beck JT. Recombinant human erythropoietin and the anemia of multiple myeloma. STEM CELLS 1993;11:88–94.[Abstract]
9. Ludwig H, Fritz E, Leitgeb C et al. Erythropoietin treatment for chronic anemia of selected hematological malignancies and solid tumours. Ann Oncol 1993;4:161–167.[Abstract/Free Full Text]
10. Hellström-Lindberg E, Ahlgren T, Beguin Y et al. Treatment of anemia in myelodysplastic syndromes with granulocyte colony-stimulating factor plus erythropoietin: results from a randomized phase II study and long-term follow-up of 71 patients. Blood 1998;92:68–75.[Abstract/Free Full Text]
11. Ludwig H, Fritz E, Leitgeb C et al. Prediction of response to erythropoietin treatment in chronic anemia of cancer. Blood 1994;84:1056–1063.[Abstract/Free Full Text]
12. Ludwig H, Fritz E. Anemia of cancer patients: patient selection and patient stratification for epoetin treatment. Semin Oncol 1998;25(suppl 7):35–38.
13. Cazzola M, Ponchio L, Pedrotti C et al. Prediction of response to recombinant human erythropoietin (rHuEpo) in anemia of malignancy. Haematologica 1996;81:434–441.[Abstract/Free Full Text]
14. Thatcher N. Predicting the response to r-HuEPO in anaemic cancer patients undergoing chemotherapy. Erythropoiesis: new dimensions in the treatment of anaemia 1995;6:99–106.
15. Henry DH, Glaspy J. Predicting response to epoetin alfa in anemic cancer patients receiving chemorx. Proc Am Soc Clin Oncol 1997;16:49a.
16. Abels RI, Larholt KM, Krantz KD et al. Recombinant human erythropoietin (rHuEPO) for the treatment of the anemia of cancer. The Oncologist 1996;1:140–150.[Abstract/Free Full Text]
17. Dammacco F, Castoldi G, Rodjer S. Efficacy of epoetin alfa in the treatment of anaemia of multiple myeloma. Br J Haematol 2001;113:172–179.[CrossRef][Medline]
18. Youden WJ. Index for rating diagnostic tests. Cancer 1950;3:32–35.[CrossRef][Medline]
19. Adamson JW, Ludwig H. Predicting the hematopoietic response to recombinant human erythropoietin (epoetin alfa) in the treatment of the anemia of cancer. Oncology 1999;56:46–53.[CrossRef][Medline]
20. Henry DH. Recombinant human erythropoietin treatment of anemic cancer patients. Cancer Prac 1996;4:180–184.
21. Abedi MR, Backman L, Bostrum L. Markedly increased serum erythropoietin levels following conditioning for allogeneic bone marrow transplantation. Bone Marrow Transplant 1990;6:121–126.
22. Birgegård G, Wide L. Marked erythropoietin increase before fall in Hb after treatment with cytostatic drugs suggests mechanism other than anaemia for stimulation. Br J Haematol 1989;72:462–466.[Medline]
23. Saijo Y, Nakai Y, Saito J et al. Changes in serum erythropoietin levels during chemotherapy for lung cancer. Chemotherapy 1992;38:281–285.[Medline]
24. Cazzola M, Messinger D, Battistel V et al. Recombinant human erythropoietin in the anemia associated with multiple myeloma or non-Hodgkin’s lymphoma: dose finding and identification of predictors of response. Blood 1995;86:4446–4453.[Abstract/Free Full Text]

This article has been cited by other articles:

				T. Steinmetz, M. Hellmich, M. Neise, A. Aldaud, C. Lerchenmuller, A. Tsamaloukas, F. Fandel, C. Weiligmann, U. Totzke, and S. Schmitz Prediction of the Responsiveness to Treatment with Erythropoiesis-Stimulating Factors: A Prospective Clinical Study in Patients with Solid Tumors Oncologist, June 1, 2007; 12(6): 748 - 755. [Abstract] [Full Text] [PDF]

				A. Hill, S.J. Richards, R.P. Rother, and P. Hillmen Erythropoietin treatment during complement inhibition with eculizumab in a patient with paroxysmal nocturnal hemoglobinuria Haematologica, March 1, 2007; 92(3): e31 - e33. [Abstract] [Full Text] [PDF]

				R. Boccia, I. A. Malik, V. Raja, S. Kahanic, R. Liu, T. Lillie, D. Tomita, B. Clowney, and P. Silberstein Darbepoetin alfa administered every three weeks is effective for the treatment of chemotherapy-induced anemia. Oncologist, April 1, 2006; 11(4): 409 - 417. [Abstract] [Full Text] [PDF]

				R. Stasi, S. Amadori, T. J. Littlewood, E. Terzoli, A. C. Newland, and D. Provan Management of Cancer-Related Anemia with Erythropoietic Agents: Doubts, Certainties, and Concerns Oncologist, August 1, 2005; 10(7): 539 - 554. [Abstract] [Full Text] [PDF]

				J. Chang, F. Couture, S. Young, K.-L. McWatters, and C. Y. Lau Weekly Epoetin Alfa Maintains Hemoglobin, Improves Quality of Life, and Reduces Transfusion in Breast Cancer Patients Receiving Chemotherapy J. Clin. Oncol., April 20, 2005; 23(12): 2597 - 2605. [Abstract] [Full Text] [PDF]

				F. Couture, A.R. Turner, B. Melosky, L. Xiu, R.K. Plante, C.Y. Lau, and I. Quirt Prior Red Blood Cell Transfusions in Cancer Patients Increase the Risk of Subsequent Transfusions With or Without Recombinant Human Erythropoietin Management Oncologist, January 1, 2005; 10(1): 63 - 71. [Abstract] [Full Text] [PDF]

				D. H. Henry The Evolving Role of Epoetin Alfa in Cancer Therapy Oncologist, February 1, 2004; 9(1): 97 - 107. [Abstract] [Full Text] [PDF]

This Article Abstract Full Text (PDF) eLetters: Submit a response to this article Alert me when this article is cited Alert me when eLetters are posted Alert me if a correction is posted Citation Map Services E-mail this article link to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Reprints/Permissions Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Littlewood, T. J. Articles by Perkins, A. Search for Related Content PubMed PubMed Citation Articles by Littlewood, T. J. Articles by Perkins, A.