1. Darbepoetin alfa (Aranesp®) molecular characteristics and basic research

2. Presentation overview
The evolution of protein therapeutics
Structure and function of recombinant human erythropoietin
Importance of sialic acid content
Discovery and development of darbepoetin alfa
molecular characteristics
implications for clinical use

3. The evolution of protein therapeutics
Recognition that proteins can be useful ‘hormone-like’ therapeutics eg, insulin (1920s)
First purified from animal and human tissues eg, insulin, growth hormone, factor VIII
Recombinant protein therapeutics – Amgen scientists were among the leaders cloning erythropoietin in 1983
Key points
The development of recombinant protein technology to produce drugs for human therapy has been one of the most significant achievements of biotechnology in recent decades.
Therapeutic proteins are often recombinant versions or functional equivalents of naturally occurring molecules in the body. Research in this area has made substantial progress since the 1950s when the structure of DNA was determined.
Many recombinant human proteins have since been generated and have proved invaluable as targeted treatments for a range of diseases, eg, recombinant human erythropoietin for anaemia, insulin for diabetes, interferon alpha-2b for the management of hairy cell leukaemia and recombinant methionyl human granulocyte colony-stimulating factor for the treatment of chemotherapy-induced neutropenia.
Research is now focusing on the relationship between pharmacokinetic and pharmacodynamic properties of molecules, with the aim of engineering proteins that possess enhanced therapeutic characteristics.

4. The evolution of protein therapeutics
Recognition that proteins can be useful ‘hormone-like’ therapeutics eg, insulin (1920s)
First purified from animal and human tissues eg, insulin, growth hormone, factor VIII
Recombinant protein therapeutics – Amgen scientists were among the leaders cloning erythropoietin in 1983
Now a new era where protein therapeutics are modified to enhance their properties as therapeutics eg, darbepoetin alfa (Aranesp®)
Key points
Molecular modifications, such as the creation of sialic acid-containing carbohydrate molecules or the derivitisation of molecules, such as the attachment of polyethylene glycol molecules, can increase the half-life of drugs and allow extended dosing intervals, thus relieving the burden of frequent dosing schedules.
Such advances have already been made in the design and development of haematopoietic growth factors such as darbepoetin alfa (Aranesp®).

5. Darbepoetin alfa
Darbepoetin alfa is a biochemically distinct recombinant erythropoietic protein that stimulates the production of red blood cells
The discovery of darbepoetin alfa resulted from basic research into the structure and function of rHuEPO and its attached carbohydrate
The longer serum residence time and greater biological activity of darbepoetin alfa results from the addition of two extra sialic acid-containing carbohydrate side chains
Key points
Amgen hypothesised that if it could create a novel super-sialated erythropoiesis stimulating protein, then that novel protein might last longer in the circulation, and exhibit increased potency. Such an enhanced erythropoietic agent might offer patients the benefit of less frequent dosing.
Darbepoetin alfa is an erythropoietic protein that was, therefore, rationally designed to contain an increased number of carbohydrate chains and sialic acid molecules compared with recombinant human erythropoietin (rHuEPO).
The increased sialic acid content of darbepoetin alfa results in a greater serum residence time and increased in vivo biological activity compared with rHuEPO.1
Reference
1. Egrie JC, et al. Br J Cancer. 2001;84(suppl 1):3-10.

6. Structure of EPO bound to an EPO receptor
Carbohydrate side chains
rHuEPO has four carbohydrate side chains
rHuEPO has a theoretical maximum of 14 sialic acids
Key point
Recombinant human erythropoietin (rHuEPO) has four carbohydrate side chains, with a maximum number of 14 sialic acid residues, which are distal to the receptor-binding sites.
Background
Structural determination of rHuEPO has been possible through techniques such as nuclear magnetic resonance spectroscopy and X-ray crystallography. This slide illustrates the overall topography of rHuEPO when bound to the erythropoietin (EPO) receptor.
The carbohydrate addition sites are clustered at one end of the molecule, distal from the receptor-binding site. The four carbohydrate side chains contribute approximately 40% of the mass of the hormone.1
Reference
1. Egrie JC, et al. Br J Cancer. 2001;84(suppl 1):3-10.

7. Typical tetra-antennary carbohydrate
Sialic acid
N-linked carbohydrate
Key points
The three N-linked carbohydrate chains present on recombinant human erythropoietin (rHuEPO) can each contain up to four sialic acid residues.
An additional O-linked carbohydrate chain can contain up to two sialic acid residues.
Therefore, the maximum number of sialic acid residues that can be found on one molecule of rHuEPO is 14.
Background
Amgen’s research showed that the number of sialic acid-containing branches attached to carbohydrate side chains varies between rHuEPO isoforms. Some rHuEPO isoforms have two, three or four carbohydrate side chains. These are known respectively as biantennary, triantennary and tetraantennary.
On the more highly sialated isoforms, the greater number of sialic residues is responsible for the observed increase in erythropoietic potency. For example, isoforms with tetraantennary side chains are more potent than those with triantennary side chains. This is because the pharmacokinetic profile of the rHuEPO changes as the sialic acid content increases. Specifically, it remains in the circulation longer.
The number of branches in carbohydrates and therefore the number of sialic acids is variable

8. In vivo activity in mice increases with greater sialic acid content
Longer serum half-life
Higher receptor binding
Greater in vivo activity 30 25 20 15
Increase in Hct at day 30 (points) 10
Key points
Experiments on isolated recombinant human erythropoietin (rHuEPO) isoforms have demonstrated that the carbohydrate moieties of erythropoietin have significant effects on the in vivo biological activity of the hormone; increased sialic acid-containing carbohydrate is associated with greater in vivo activity in mice.1,2
Early on, Amgen researchers realised that the rHuEPO expression system produced not just one form of the molecule, but a mixture of many different isoforms. Each of the isoforms identified in the laboratory was assigned an individual number, and subsequent studies showed that rHuEPO released as product post-purification consisted of a mixture of isoforms 9–14.
Isoforms with a greater sialic acid content have a longer serum half-life but a lower receptor binding affinity. Thus, serum clearance (serum half-life) is the primary determinant of in vivo activity, rather than receptor binding affinity.2
Background
A series of experiments were carried out to explore the relationship between carbohydrate content and in vivo biological activity in rHuEPO.1–3 rHuEPO isoforms 8 through 14 were individually purified, and the efficacy of each isoform in stimulating erythropoiesis in mice was measured. Mice were injected with a vehicle control or an equimolar dose (2.5 µg/kg of each peptide) of each of the individual isoforms, three times per week, for 1 month.
Isoforms having a greater sialic acid-containing carbohydrate content were the most potent in inducing an increase in haematocrit (Hct).3 By day 30, the group mean Hct of isoform 14-treated animals increased by 26.2 ± 2.7 points (to 76.2%), compared with an increase of only 6.3 ± 3.5 points for the isoform 8-treated group.2
References
Egrie J, et al. Glycoconj J. 1993;10:263.
Egrie JC, et al. Br J Cancer. 2001;84(suppl 1):3-10.
Egrie J, et al. Blood. 1997;90:56a. 5 8 9 10 11 12 13 14
Adapted from Egrie JC, et al. Br J Cancer. 2001;84(suppl 1):3-10.
Isoform

9. Darbepoetin alfa development strategy
Introduce N-linked glycosylation consensus sequences (Asn-Xxx-Thr/Ser) into rHuEPO
Identify individual variants that have the desired properties
Test optimal combinations of variants
Key points
The first step in developing the new erythropoietic molecule was to introduce additional carbohydrate. In the recombinant human erythropoietin (rHuEPO) expression system, one particular amino acid sequence triggers glycosylation of the polypeptide backbone.
At one end of this sequence is asparagine (Asn); at the other end is either serine (Ser) or threonine (Thr). Scientists refer to this amino acid sequence as an N-linked consensus glycosylation sequence. To add extra carbohydrate chains to rHuEPO, it was therefore necessary to add extra N-linked glycosylation sequences.
Numerous candidate molecules were generated. With each molecule, it was necessary to determine whether carbohydrate was added and the effect on folding and erythropoietic potency.
Finally, combining certain amino acid alterations might result in a molecule that shows greater activity than any single alteration on its own.

10. Amino acid sequence of rHuEPO
ALA 1
PRO 3 2
ARG 4
LEU 5
CYS 7
ASP 8
SER 9 10
ILE 6
VAL 11 12
GLU 13 14
TYR 15 16 17 18 19
LYS 20 21 22 23
ASN 24 25
THR 26
GLY 28 29 30 27 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47
PHE 48 49 50
TRP 51 52 53
MET 54 55 56 57
GLN 58 59 60 61 62 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 95 96 97 98 99
100
101
102
103
104
105
106
107
108
109
110
111
112
GlY
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
161
160
159
158
157
156
155
154
153
152
151
150
139
140
141
142
143
144
145
146
147
148
149
162
163
164
165
NH2
COOH
HIS 94
N-glycosylation sites
Disulphide linkages
O-glycosylation site
Key points
In order to create a new erythropoietic molecule, Amgen had to start at a very basic level, namely the primary structure of recombinant human eyrthropoietin (rHuEPO).
Amgen scientists recognised that adding new carbohydrate chains to rHuEPO would require alterations to the 165-amino acid sequence of the polypeptide backbone.
In order to alter the amino acid sequence, one needs to know, for example, which amino acid residues are involved in stimulating the erythropoietin receptor, and what critical folding and structural elements must be preserved in order to retain the full bioactivity of the molecule.
Altering specific amino acids involves a technique known as site-directed mutagenesis; an individual mutation is targeted at a particular site on the backbone.
Early observations had already shown that certain amino acids had to remain intact for rHuEPO to retain its full bioactivity
residues 7 and 161, as well as residues 29 and 33, are cross-linked by disulphide bridges. Disrupting these inactivates rHuEPO.
four amino acid residues in rHuEPO are glycosylation sites. Three of the glycosylation sites are N-linked (orange in diagram), and one is O-linked (green in diagram). The carbohydrate is important for activity.

11. rHuEPO has two receptor binding sites
Key point
The active sites indicated in red are crucial for the binding of recombinant human erythropoietin (rHuEPO) to the erythropoietin (EPO) receptor. Alteration of the amino acid residues in this area leads to a loss of activity.
Background
The spatial relationships between the more and the less crucial residues become clearer when we look at crystal structures of rHuEPO.
The crystal structure depicted on the right-hand side of the slide shows how the amino acid side chains are arranged around the four polypeptide helices and associated connecting polypeptide loops of rHuEPO.
The space-filling model of rHuEPO depicted on the left-hand side of the slide shows that the crucial residues, shown in red, are clustered together into two distinct groups. Changing these residues destroys activity.
Amino acid residues critical for activity are grouped together in two separate regions, providing important evidence that rHuEPO binds to not just one, but two EPO receptors. The one rHuEPO molecule: two EPO receptors theory supports evidence suggesting that rHuEPO itself has two functionally indispensable domains – presumably one to bind with each receptor.
The effect of mutations on in vitro activity is indicated:
red <2% active, orange <20% active, yellow <70% active

12. The following needed to be addressed in order to make darbepoetin alfa
Would the glycan addition be efficient?
Would the molecule be properly folded/stable?
Would the ability to stimulate erythropoiesis be retained?
Would in vivo activity be increased?
Key point
In the development of a novel erythropoietic protein certain challenges needed to be addressed. Extensive research was directed at ensuring that the resultant protein would retain a similar conformation to that of endogenous erythropoietin (EPO) and thus retain (and increase) the erythropoietic activity of the native hormone.
Background
To ensure that large enough quantities of the protein could be manufactured, it was important to ensure that the process of glycan addition was efficient.
It was also important to ensure that the protein would fold in a manner comparable to the native hormone, and hence exhibit comparable stability. The three-dimensional conformation of a protein may be exquisitely sensitive to changes in the amino acid sequence, and there were concerns that certain amino acid mutations would produce a protein that would not fold correctly.
Finally, it was important to ensure that not only would the resultant protein retain the erythropoietic activity of endogenous EPO, but that the in vivo activity of the new molecule would be increased.

13. Discovery of new glycosylation sites in rHuEPO
Good bioactivity
Poor bioactivity
Glycosylated
Partial glycosylation
Unglycosylated
NH2 a1 B1 a2 a3 B2 a4
COOH
S126
N24T26
N38T40
N83S85
N-linked carbohydrate consensus sequences were introduced into rHuEPO at positions indicated by vertical lines
Each molecule was tested to see if it had the desired properties
Two positions, Ala30 and Trp88, were selected for further development work
Key points
Since in vivo biological activity increases with increasing carbohydrate, it was hypothesised that increasing the sialic acid-containing carbohydrate of erythropoietin (EPO) (beyond the maximum of 14 sialic acids in recombinant human erythropoietin [rHuEPO]) would create a molecule with enhanced in vivo activity.
To test this hypothesis, recombinant DNA technology was used to introduce consensus glycosylation sequences into the rHuEPO molecule, thereby allowing the addition of N-linked carbohydrate chains.1
Numerous candidate molecules were generated, with one or more new carbohydrate addition sites. With each molecule, it was necessary to determine whether carbohydrate was added and whether the molecule had the correct tertiary structure and retained biological activity.1
Darbepoetin alfa differs to endogenous human EPO at five amino acid positions, with two positions (alanine [Ala30] and tryptophan [Trp88]) allowing for additional carbohydrate attachment.1
Background
N-linked consensus glycosylation sequence triggers glycosylation of the polypeptide backbone. At one end of this sequence is asparagine; at the other end is either serine or threonine.
Reference
1. Egrie JC, et al. Br J Cancer. 2001;84(suppl 1):3-10.

14. Optimisation of glycosylation sites
Key points
The amino acid sequence of darbepoetin alfa differs from that of endogenous erythropoietin (EPO) at five amino acid positions, allowing for the addition of two N-linked carbohydrate chains.
Darbepoetin alfa exhibited comparable immunoreactivity to recombinant human erythropoietin (rHuEPO), suggesting that the folding and stability of darbepoetin alfa would probably be similar to that of rHuEPO.
Background
These tests of protein folding relied upon the fact that a certain monoclonal antibody (9G8A) could accurately distinguish the folding patterns of native from non-native proteins. Thus a rHuEPO variant with a folding pattern markedly different from that of the native hormone would bind more tightly to the antibody. The amount of binding of 9G8A to an altered protein can therefore be used as an index of the extent to which its folding pattern is distorted.
An immunoassay with antibody 9G8A was used to assess the immunoreactivity, and hence the likely degree of folding distortion, of the EPO variants generated. The reactivity of the variants was expressed as a percentage of reactivity of rHuEPO.
The challenge now was to find a combination of mutations that would preserve, or even enhance, the erythropoietic activity of rHuEPO.
Changing the peptide backbone to asparagine (Asn88) and threonine (Thr90) produced a molecule with only three N-linked side chains (the same number as rHuEPO); yet the immunoreactivity was about four times that of rHuEPO.
When residue 87 was changed to serine (Ser), a fourth N-linked carbohydrate chain was added to molecule NM177, but the immunoreactivity was nearly eight times that of rHuEPO. However, when residue 87 was changed to valine (Val), an additional carbohydrate side chain was added and the immunoreactivity was normalised.
After testing combinations of different molecules, it was determined that the molecule now known as darbepoetin alfa, with its five N-linked carbohydrate chains and comparable immunoreactivity to that of rHuEPO, would probably have a suitable folding and stability profile.
A two-fold increase in 9G8A immunoreactivity is suggestive of an altered conformation. Val87 substitutions allow carbohydrate addition at position 88 and normalisation of the conformation
Ala = alanine; Asn = asparagine; His = histidine; Leu = leucine; Pro = proline; Ser = serine; Thr = threonine; Trp = tryptophan; Val = valine

15. Comparison of rHuEPO and darbepoetin alfa
Receptor 1
rHuEPO
Receptor 2
Carbohydrate side chains
New carbohydrate side chains
Receptor 1
Darbepoetin alfa
Receptor 2
Three N-linked carbohydrate chains
Maximum 14 sialic acids
MW ~ 30,400 daltons
40% carbohydrate
Five N-linked carbohydrate chains
Maximum 22 sialic acids
MW ~ 37,100 daltons
51% carbohydrate
Key points
Darbepoetin alfa was rationally designed to contain an increased number of carbohydrate chains and sialic acid molecules through the creation of two additional glycosylation sites.
Darbepoetin alfa contains up to eight sialic acid residues more than recombinant human erythropoietin (rHuEPO) resulting in a three-fold longer serum half-life and greater in vivo biological activity.1
Darbepoetin alfa binds to and activates the same erythropoietin (EPO) receptor as endogenous EPO and rHuEPO.2
Background
Darbepoetin alfa is a biochemically distinct molecule, which has an increased sialic acid content and thus an increased molecular weight (MW) when compared with rHuEPO. Each additional N-linked carbohydrate chain increases the MW of the protein by approximately 3,300 daltons, correlating to four additional sialic acid residues.1
Thus, in comparison with rHuEPO, the two extra carbohydrate chains on darbepoetin alfa increase the MW by 22% and the maximum number of sialic acid residues from 14–22.1
The result of this modification is a three-fold longer serum half-life and increased in vivo biological activity.2
References
1. Egrie JC, et al. Br J Cancer. 2001;84(suppl 1):3-10.
2. Macdougall IC, et al. J Am Soc Nephrol. 1999;10:

16. Darbepoetin alfa has a longer half-life than rHuEPO: single-dose PK of IV administration
100 10 1
0.1
0.01
Darbepoetin alfa (oncology; 0.5 µg/kg, n = 20)*1
Darbepoetin alfa (dialysis; 0.5 µg/kg, n = 11)2
rHuEPO (dialysis; 100 IU/kg, n = 10)2
Mean (SD) baseline-corrected
serum concentration (ng/mL)
t1/2 = 38.8 hours
t1/2 = 25.3 hours
Key points
Darbepoetin alfa has a longer serum half-life than recombinant human erythropoietin (rHuEPO) in dialysis patients with anaemia (25.3 hours vs 8.5 hours).
In a separate study, darbepoetin alfa was administered to patients with nonmyeloid malignancies receiving chemotherapy at a dose of 2.25 µg/kg once weekly. The normalised serum half-life for a dose of 0.5 µg/kg was 38.8 hours.1
Data from a study in patients with cancer who were not receiving chemotherapy reported a serum half-life for darbepoetin alfa of 50 hours following subcutaneous administration.2
Background
Macdougall and colleagues conducted a randomised study to compare the pharmacokinetics (PK) of a single dose of darbepoetin alfa (0.5 µg/kg) and an equimolar dose of rHuEPO (100 IU/kg) after intravenous administration to dialysis patients with anaemia.3
The results of immunoassays showed that darbepoetin alfa has an approximately three-fold greater terminal half-life and decreased clearance compared with rHuEPO.
Increased sialic acid-containing carbohydrate content of darbepoetin alfa resulted in a longer serum half-life.
Reference
Heatherington A, et al. Proc Am Soc Clin Oncol. 2002;21:256b. Abstract 2844.
Smith R, et al. Proc Am Soc Clin Oncol. 2002;21:367a. Abstract 1465.
Macdougall I, et al. J Am Soc Nephrol. 1999;10:2392–2395.
t1/2 = 8.5 hours
Time post-IV injection (hours)
*Oncology patients received 2.25 µg/kg
Data shown are normalised for 0.5 µg/kg
SD = standard deviation; IV = intravenous
1Heatherington A, et al. Proc Am Soc Clin Oncol. 2002;21:256b. Abstract and poster; 2Macdougall I, et al. J Am Soc Nephrol. 1999;10:

17. In vivo activity in mice increases with increasing number of glycans80
59Fe incorporated
(% of maximum) 60 40
rHuEPO (three chains) 20
NM279 (four chains)
Key points
A bioassay of erythropoietic activity in mice was used to compare the potency of three molecules
recombinant human erythropoietin (rHuEPO), with three N-linked carbohydrate side chains
NM279, with four side chains
darbepoetin alfa, with five side chains.
In this assay, the incorporation of radioactive iron (59Fe) into animal red blood cells was measured at various concentrations of test molecule, and the results plotted on a logarithmic scale.
The assay showed that as the number of carbohydrate side chains increased from three to four to five, the erythropoietic response shifted progressively to the left.
As these results were plotted on a logarithmic scale, it became clear that the addition of sialic acid residues to darbepoetin alfa is associated with a dramatically increased erythropoietic activity in vivo.
Darbepoetin alfa (five chains) 10
100
1,000
10,000
100,000
1,000,000
ng/mL sample

18. The anti-EPO monoclonal Ab F12 does not neutralise EPO bioactivity
Effect of antibodies on EPO in vitro bioactivity
120
100
Non-neutralising anti-EPO
monoclonal Ab (F12)
Neutralising anti-EPO
monoclonal Ab (D11)
polyclonal Ab (862) 80 60
In vitro activity (%)* 40
Key points
Although intuition suggested that an antibody that may bind to the altered peptide sequence of darbepoetin alfa would not be neutralising, this concept required testing. An attempt was therefore made to develop experimental assays that would provide quantitative evidence.
For this experiment it was necessary to use recombinant human erythropoietin (rHuEPO) as F12 does not bind to darbepoetin alfa. rHuEPO was incubated with two known neutralising antibodies as well as F12, the antibody directed against the peptide region that was changed to produce darbepoetin alfa. With increasing concentrations of either of the known neutralising antibodies, the bioactivity of rHuEPO decreased sharply. By contrast, even when the concentration of F12 in the incubation medium was increased a thousand-fold, the rHuEPO bioactivity remained intact.
The experiment with F12 shows that, whether shielded by glycans or not, the peptide region at which rHuEPO has been changed to produce darbepoetin alfa does not provoke a neutralising antibody effect. These findings suggest that, even if it were possible to generate an antibody to the polypeptide that was changed to make darbepoetin alfa, the bioactivity of darbepoetin alfa may not be neutralised. 20 *
In vitro activity assay measures formation of erythroid colonies from human bone marrow in soft agar
0.001
0.01
0.1 1 10
100
Amount of Ab added (µg/mL)
EPO = erythropoietin; Ab - antibody

19. Development of darbepoetin alfa
A new erythropoietic protein, biochemically distinct from rHuEPO
Increased sialic acid content, resulting in
a longer circulating half-life (2–3-fold greater than rHuEPO)
less frequent dosing requirements
increased biological activity
Pharmacokinetics offer potential for higher response rates and faster onset of action
Key points
The discovery of darbepoetin alfa resulted from basic research on the structure and function of recombinant human erythropoietin (rHuEPO) and its attached carbohydrate.
Darbepoetin alfa has similar immunoreactivity, folding and stability properties to rHuEPO, although it has been altered at five amino acid positions, and has five N-linked carbohydrate side chains.
The increased serum residence time and in vivo bioactivity of darbepoetin alfa derive from its increased sialic acid content, resulting from two additional carbohydrate chains. These properties have been confirmed in both preclinical and clinical studies.
The greater potency of darbepoetin alfa and its longer duration of action allow it to be dosed less frequently than rHuEPO. The convenience of administering darbepoetin alfa at extended intervals benefits both patients and healthcare professionals.
The above characteristics also have clinical implications in terms of potentially enabling darbepoetin alfa to produce higher and faster responses (increases in haemoglobin) than rHuEPO.

20. Darbepoetin alfa: conclusions
Darbepoetin alfa has a similar conformation to rHuEPO
Darbepoetin alfa binds to and activates the same receptor as rHuEPO
Darbepoetin alfa has increased sialic acid-containing carbohydrate resulting in increased in vivo activity and a prolonged half-life1
Provides the opportunity to dose less frequently: QW, Q2W, Q3W or Q4W2–4
Clinical benefits have been demonstrated (high and rapid haematological responses at convenient dosing schedules)5
Key points
Darbepoetin alfa was rationally designed to have a similar conformation to recombinant human erythropoietin (rHuEPO) and to bind to and activate the same erythropoietin receptor.
Darbepoetin alfa has increased sialic acid-containing carbohydrate content resulting in increased in vivo activity and allowing for less frequent administration.1
Background
The greater potency of darbepoetin alfa and its longer duration of action allow it to be dosed less frequently than rHuEPO.2–5 The convenience of administering darbepoetin alfa at extended intervals benefits both patients and healthcare professionals.
1Egrie JC, et al. Br J Cancer. 2001;84(suppl 1):3-10.
2Glaspy JA, et al. Br J Cancer. 2002;87:
3Kotasek D, et al. Eur J Cancer In press. 4Kotasek D, et al. Proc Am Soc Clin Oncol. 2002;21:356a. Abstract 1421.
5Glaspy JA, et al. Cancer. 2003;97:
QW = once every week; Q2W = once every 2 weeks; Q3W = once every 3 weeks; Q4W = once every 4 weeks