Information Technology Reference
In-Depth Information
neous versus sequential randomization, enrollment discretion, and the pos-
sibility of intent to treat.
Allocation discretion may be available to the investigator, the patient,
both, or neither (dictated allocation). Were investigators permitted to
assign treatment based on patient characteristics? Could patients select
their own treatment from among a given set of choices?
Was actual (not virtual, quasi-, or pseudo-) randomization employed?
Was the allocation sequence predictable? (For example, patients with even
accession numbers or patients with odd accession numbers receive the
active treatment; the others receive the control.)
Was randomization conventional , that is, was the allocation sequence
generated in advance of screening any patients?
Was allocation concealed prior to its being executed? As Vance W.
Berger and Costas A. Christophi relate in a personal communication,
“This is not itself a reportable design feature, so a claim of allocation con-
cealment should be accompanied by specific design features. For example,
one may conceal the allocation sequence; and instead of using envelopes,
patient enrollment may involve calling the baseline information of the
patient to be enrolled in to a central number to receive the allocation.”
Was randomization restricted or unrestricted? Randomization is unre-
stricted if a patient's likelihood of receiving either treatment is independent
of all previous allocations and is restricted otherwise. If both treatment
groups must be assigned equally often, then prior allocations determine
the final ones. Were the proportions also hidden?
Were treatment codes concealed until all patients had been randomized
and the database locked? Were there instances of codes being revealed
accidentally? Senn [1995] warns, “investigators should delude neither
themselves, nor those who read their results, into believing that simply
because some aspects of their trial were double-blind that therefore all the
virtues of such trials apply to all their conclusions.” Masking can rarely, if
ever, be assured; see, also, Day [1998].
Was randomization simultaneous, block simultaneous, or sequential ? A
blocked randomization is block simultaneous if all patients within any given
block are identified and assigned accession numbers prior to any patient in
that block being treated.
And, not least, was intent to treat permitted?
Missing Data 2
Every experiment or survey has its exceptions. You must report the raw
numbers of such exceptions and, in some instances, provide additional
Material in this section is reprinted with permission from Manager's Guide to Design and
Conduct of Clinical Trials , Wiley, 2002.
2
Search WWH ::




Custom Search