edc-randomization

Randomization objects for clinicedc projects

Overview

The Randomizer class emulates the randomization of a clincial trial participant in realtime. This module doesn't actually randomize in realtime. Instead, a CSV file is prepared in advance by the statistician. This CSV file lists the order in which subjects are to be randomized. The Randomizer class initially imports the entire list in order into a model. When a subject is to be randomized, the Randomizer class selects the next available row from the model.

A very basic randomization_list.csv prepared in advance might look like this:

site_name,sid,assignment
temeke,1000,active
temeke,1001,active
temeke,1002,placebo
temeke,1003,active
temeke,1004,placebo
temeke,1005,placebo
...

For large multisite trials this may be thousands of lines ordered using some type of block randomization.

This module will import (only once) all rows from the CSV file into a model. The Randomizer class selects and allocates in order by site_name one row per participant from the model.

randomizer_cls = site_randomizers.get("default")
randomizer_cls.randomize(subject_identifier=subject_identifier, ...)
# or just:
site_randomizers.randomize("default", subject_identifier=subject_identifier, ...)

Usually, the Randomizer class is instantiated in a signal once the subject's eligibility is confirmed and the subject's informed consent is submitted. A signal attached to the subject's informed consent is a good place to do this assuming the sequence of events are 1) pass eligibility criteria, 2) complete informed consent, 3) randomize and issue study identifier 4) start baseline visit.

@receiver(
    post_save,
    weak=False,
    sender=SubjectConsent,
    dispatch_uid="subject_consent_on_post_save",
)
def subject_consent_on_post_save(sender, instance, raw, created, **kwargs):
    if not raw:
        if created:
            ...
            # randomize
            site_randomizers.randomize(
                "default",
                subject_identifier=instance.subject_identifier,
                report_datetime=instance.consent_datetime,
                site=instance.site,
                user=instance.user_created,
            )
            ...

Registering a randomizer

The default Randomizer class is edc_randomization.randomizer.Randomizer. Unless you indicate otherwise, it will be automatically registered with the site controller, site_randomizers with the name default. It is recommended you access the Randomizer class through site_randomizers instead of directly importing.

randomizer_cls = site_randomizers.get("default")

Customizing the default randomizer

Some attributes of the default Randomizer class can be customized using settings attributes:

EDC_RANDOMIZATION_LIST_PATH = 'path/to/csv_file'
EDC_RANDOMIZATION_ASSIGNMENT_MAP = {
    "intervention": 1,
    "control": 2,
    }
EDC_RANDOMIZATION_ASSIGNMENT_DESCRIPTION_MAP = {
    "intervention": "Fluconazole plus flucytosine",
    "control": "Fluconazole"
    }

Creating a custom randomizer

If you need to customize further, create a custom Randomizer class.

In the example below, gender is added for a trial stratified by gender.

Custom Randomizer classes live in randomizers.py in the root of your app. The site_randomizers controller will autodiscover them.

# my_app/randomizers.py

@register()
class MyRandomizer(Randomizer):
    name = "my_randomizer"
    model = "edc_randomization.myrandomizationlist"
    randomization_list_path = tmpdir
    assignment_map = {"Intervention": 1, "Control": 0}
    assignment_description_map = {"Intervention": "Fluconazole plus flucytosine", "Control": "Fluconazole"}
    extra_csv_fieldnames = ["gender"]

    def __init__(self, gender=None, **kwargs):
        self.gender = gender
        super().__init__(**kwargs)

    @property
    def extra_required_instance_attrs(self):
        return dict(gender=self.gender)

    @property
    def extra_model_obj_options(self):
        return dict(gender=self.gender)

    @classmethod
    def get_extra_list_display(cls):
        return [(4, "gender")]

The register() decorator registers the custom class with site_randomizers.

With a custom randomizer, the default Randomizer class is no longer needed, update settings to prevent the default class from registering.

Use the settings attribute:

EDC_RANDOMIZATION_REGISTER_DEFAULT_RANDOMIZER = False

Confirm this by checking the site_randomizers:

>>> randomizer_cls = site_randomizers.get("default")
NotRegistered: A Randomizer class by this name ...

>>> randomizer_cls = site_randomizers.get("my_randomizer")
>>> randomizer_cls.name
"my_randomizer"

Manually Importing from CSV

A Randomizer class will call import_list when it is instantiated for the first time. If you want to load the CSV file manually, import the Randomizer class and call import_list().

>>> randomizer_cls = site_randomizers.get("my_randomizer")
>>> randomizer_cls.import_list()
Import CSV data
  Randomizer:
    -  Name: my_randomizer
    -  Assignments: {'active': 1, 'placebo': 2}
    -  Model: edc_randomization.myrandomizationlist
    -  Path: /home/me/.etc/randomization_list.csv
    -  Imported 5 SIDs for randomizer `my_randomizer` into model `edc_randomization.myrandomizationlist`
       from /home/me/.etc/randomization_list.csv.
    -  Verified OK.

Manually Export to CSV

>>> from edc_randomization.utils import export_randomization_list
>>> export_randomization_list(randomizer_name="default",path="~/", username="erikvw")

If the user does not have permissions to view the randomizationlist table, a RandomizationListExporterError will be raised:

RandomizationListExporterError: User `erikvw` does not have permission to view 'edc_randomization.randomizationlist'