In addition to the gift certificates, each enrolled subject will receive a small baby gift following each visit. These gifts will include rattles, teethers and other small gifts appropriate for young infants. The DCC will ship the gifts to each site on a quarterly basis.
III.13. Randomization Three steps are needed so the Data Coordinating Center (DCC) can perform the randomization procedure described in Section II.2.4.6. "Randomize the Subject, Form 04 (Randomization/Exclusion), continued". The steps are:
• Generate a treatment assignment schedule for each Enrollment Site,
• Write the interactive randomization program used by the registrar during each randomization, and
• Train the registrars.
These steps are described in the following paragraphs.
DCC staff members prepare a treatment assignment schedule for each Enrollment Site, which is used to make assignments for that Site during randomization phone calls. Each Site's randomization schedule is constructed using permuted blocks of lengths 2 and 4, ensuring that the two treatment groups at a Site never differ in number randomized by more than two subjects. The block lengths of 2 and 4 are alternated at random to make it harder to predict future treatment assignments. For each Site, the schedule is twice as long as the anticipated enrollment (2 times 204 = 408) to ensure the Site has enough treatment assignments.
The computer program to check eligibility and assign a subject's treatment leads the registrar through a series of windows containing questions to verify the subject's eligibility and to assign the subject to a treatment group. The program does not permit a given patient ID (PID) to be randomized twice. The program is written in NOMAD, adapted from programs used in previous studies.
The primary registrar will be the Data Entry/Quality Control Operator, backed up by the Statistical Study Manager. The Statistical Study Manager trains registrars, who practice randomizations using a set of dummy databases.
III.14. Data Collection Procedures III.14.1. Forms completion Data recorded on forms are obtained
• At the Enrollment Sites:
– during face-to-face interviews at the baseline visit;
– during periodontal examinations;
– during follow-up visits and oral examinations;
– by chart abstraction (for delivery data); and
• At the Microbiology Lab and the Host Response Lab
– from laboratory assays of plaque and serum.
III.14.1.1. Forms completed at Enrollment Sites. All forms completed at the Enrollment Sites are checked by the Site's Study Coordinator for accuracy, consistency, and completeness. Whenever possible, the Study Coordinator performs these checks before the subject has left the Enrollment Site, so problems can be rectified.
III.14.1.2. Forms completed at the Labs. Laboratory assay forms are completed at the appropriate laboratory. Specifically, each such form is an Excel file containing modest standard formatting consistent with the OPT study's other case report forms. Before each file is transmitted, it is checked for accuracy, consistency, and completeness by the Lab's Director or by a person designated by the Director.
III.14.2. Procedures for transmitting forms and retaining copies (Form 80, "Standard Packing List" and Form 81 "OPT Study Shipment Receipt") III.14.2.1. Forms completed at Enrollment Sites. Each week, the Study Coordinator ships a package to the DCC containing all forms and all Error Correction Requests (ECRs) completed since the last shipment.
• Packages should be shipped via first-class mail or Priority Mail, depending on the package's weight.
• Before sending a completed form or ECR, the Study Coordinator photocopies the originals and stores the copies in the appropriate subject files.
• Each package contains a packing list and a return postcard. The purpose of this procedure is to track every form. It is like an insurance policy: it has a cost up-front and it won't be needed often, but it is indispensable when it is needed.
• The packing list is Form 80, Standard Packing List. This form will also be used as a packing list for packages sent from the DCC to the Sites. The packing list is completed by these steps, illustrated by the example in Figure 3:
– Fill in the identifying information in the upper right-hand corner of each page. This includes a shipment number assigned sequentially by the Site Coordinator; the date the package was shipped; the page number and how many pages are in the packing list in total; the Site from which it originated; the package's destination; and the name of the person completing the form. It would be helpful to keep your photocopy of each packing list in a notebook so you can easily see what the next shipment number will be.
– In the main body of the page, the column on the far right, "rec'd", is used by the recipient. The example in Figure 3 (next page) shows several different ways of arranging the listing of items sent. Sometimes it makes more sense to list each patient ID (PID) number followed by the form numbers, while at other times it may be easier to list a form such as Randomization/Exclusion (Form 04) followed by the PIDs. The person who checks in your forms at the DCC will let you know if you should change the way you order your forms.
– All other parts of the page are for the recipient's use and are not completed by the Study Coordinator.
– The Study Coordinator uses as many pages as necessary to list all forms in the shipment; the same shipment number goes on all pages.
• The return postcard is Form 81 "OPT Study Shipment Receipt". This form will also be used as a return postcard for packages sent from the DCC to the Sites. The return postcard is completed by entering, in the box labeled "To be completed by sender:", the same shipment number and shipping date that were written on the packing list.
• The Study Coordinator photocopies the packing list, files the copy, and includes the original in the shipment.
• When the shipment arrives at the DCC, the Data Entry/Quality Control Operator notes on the return postcard that the shipment arrived and the date it arrived. S/he then compares the packing list to the actual contents of package, form by form, noting any discrepancies on the part of the packing list labeled "Please return postcard". If there are no discrepancies, The Data Entry/Quality Control Operator checks the box on the return postcard labeled "Shipment OK". If there are discrepancies, the Data Entry/Quality Control Operator checks the box on the return postcard labeled "Other" and notes the discrepancies on the card. This completes the return postcard, which is then mailed to the Site.
• When the return postcard arrives at the Site, the Study Coordinator should resolve any discrepancies noted on the postcard and file it with the photocopy of the corresponding packing list.
III.14.2.2. Forms completed at the Labs. Lab data are recorded in Excel files having a modest format consistent with the OPT study's other forms. Usually, many such files will be transmitted to the DCC in a single e-mail.
• Specifically, the Excel files are transmitted as attachments to the e-mail. Also attached to the e-mail is a packing list, a Word document having the same format as the packing list used by the Sites for paper forms.
• The packing list is completed in much the same manner as for paper forms, as described above in Section III.14.2.1, "Forms completed at Enrollment Sites".
– The most important difference is that Forms from the Labs are not specific to subjects, but rather to groups of samples. Thus, an item on the packing list now refers to an individual form, not to a type of forms.
– Each such form is described in the packing list by the date on the form and by the analyte described in the form. For forms sent by the Microbiology Lab, this is a bacterial species.
• The packing list itself serves as the return postcard, so no separate form is needed for shipments from the Labs to the DCC.
When the e-mail arrives at the DCC, the Data Entry/Quality Control Operator detaches each attachment from the e-mail and checks each one against the packing list. Discrepancies are noted in the box on the packing list labeled "Please Return Postcard". If there are no discrepancies, The Data Entry/Quality Control Operator types "Shipment OK" in this box. If there are discrepancies, the Data Entry/Quality Control Operator notes the discrepancies in the box. Then the Data Entry/Quality Control Operator notes, in the box on the lower right, the date the shipment was received, its general condition, and who received it.
Having completed the packing list, the Data Entry/Quality Control Operator prints a copy for DCC files and e-mails the modified packing list back to the Lab. The designated Lab personnel check the returned packing list, resolving any discrepancies, and file the packing list for future reference.
III.15. Data Management and Error Correction The primary data-management system for this study is based on NOMAD, a fourth-generation relational database package, in a UNIX implementation running on the DCC's SUN workstations, which are networked with other workstations at the University of Minnesota's Coordinating Centers for Biometric Research (CCBR). UNIX NOMAD is essentially a user interface that relies on an underlying ORACLE data management system. ORACLE can be used directly on the same NOMAD databases if desired. ORACLE provides the capability of creating database files that can be read directly by SAS and other languages.
All necessary software will be modified or written and will be maintained by the Database Administrator. Security and confidentiality of computer records are protected by
• restricting access to computer files only to staff with authorized passwords;
• using numeric patient IDs (PIDs) as identifiers rather than names or Social Security numbers; and
• not referring to individual subjects by name or other non-study identifiers in any publications or presentations.
Section III.7 "Maintaining Confidentiality" discusses security and confidentiality in more detail.
Paper files in the DCC will be stored in locked file cabinets with access limited to authorized staff.
III.15.1. Forms processing at the Data Coordinating Center (DCC) Based on planned enrollment, the weekly forms load at the DCC should be roughly 100 confirmation forms, which require minimal data entry and have a small chance of error, and roughly 140 forms with substantial data entry and error correction burdens.
The following steps occur when a package of forms and error-correction requests is received from an Enrollment Site.
1. A package is received from an Enrollment Site. (Details on shipping forms from Sites to the DCC are in Section III.14.2 "Mailing, Documenting, and Filing Backup Copies".) The enclosed packing list and package contents are checked for accuracy and completeness. The packing list is stored in a notebook specific to the Site. The receipt postcard is mailed back to the site. The Data Entry/Quality Control Operator stamps each form with a unique sequence number.
2. Forms of a similar type are collected into batches for keying. Batch header and trailer tags, which include counts of the forms, are attached.
3. Forms batches are keyed and re-keyed by the Data Entry/Quality Control Operator, using a PC-based data entry package that includes an initial interactive edit of the data. Errors discovered from the double-keying process are corrected. Data files corresponding to the form batches are created.
4. Data file batches are uploaded to the Division of Biostatistics UNIX network, to the main study data area.
5. Data file batches are subjected to a thorough edit. The edit system, based in part on schemas that incorporate a data dictionary and the database contents and structure, has full capabilities for checking on data types (numeric, dates, special codes, check digits), ranges, logical consistency within and across different databases, and table lookups for special items.
6. An edit report is produced and if errors are detected, error correction requests (ECRs) are printed. ECRs are reviewed manually by the Data Entry/Quality Control Operator and edited to remove redundant messages arising from a single apparent error. The ECRs are then e-mailed to the originating Site. (For details on ECRs, see Section III.15.2 "Error Corrections", below.) ECRs are logged out, and the existence and status of the unresolved error is stored in a forms status database that is updated when the ECR returns from the Site.
7. Daily backup of all new and modified data occurs automatically. Each day, study files that have been altered that day are backed up on the Division of Biostatistics UNIX network. Two copies of the complete backup are made; one is stored offsite.
8. Each form has a corresponding database. When a database is complete -- for example, when randomization is complete -- a formal closeout process is begun. This involves a final edit of all fields in the database with resolution of all outstanding errors. Fields for which errors cannot be resolved are left blank and a permanent error marker is stored in the forms status database.
9. Analysis files are created to protect the main databases and to facilitate reporting. These are fixed-format ASCII sequential files, readable by SAS, S+, and other software.
III.15.2. Error Corrections [NOTE: The database and associated error-correction software are not yet completed. Thus, this section uses examples from previous studies that used the same approach to data management. When the present study's database is fully operational, this section will change but only in minor details.]
Error correction requests (ECRs) are generated by the DCC when information reported on a form is missing, inconsistent, or appears to be incorrect. Site personnel determine the correct values of all data items, because they have access to the subjects and to source documents containing the correct values. Because it is hazardous for the DCC to assume anything about apparent errors, even in cases that seem obvious, the Site must respond to every error correction request, with rare exceptions that will be specified by the DCC.
This section does not cover all possible cases of error correction request. Instead, it gives the general procedure and describes the most common kinds of ECRs.
III.15.2.1. What is an error correction request? Error correction requests (ECRs) are sent by the DCC for the following reasons:
(i) A question was left blank that should have been answered according to the directions on the form.
(ii) A question was answered that should have been left blank according to the directions on the form.
(iii) A calculation appears to be incorrect, for example, on Form 11 (Periodontal measurements), the fraction of sites that bled on probing.
(iv) An answer is high or low given typical values for that measurement, for example, an age of 250, which is probably a typographical error for age 25.
(v) An answer is inconsistent with other information the DCC has for that subject, or with other information on the same form. For example, the date on a form may be given as "01-05-03" (5 January 2003) when the subject actually enrolled in November 2003 and the correct date is "01-05-04" (5 January 2004).
(vi) A subject was enrolled without an ultrasound to confirm gestational age (this will occur only at the Kentucky site).
ECRs are generated by computer and pruned by the DCC's Data Entry/Quality Control Operator before being sent to the Site. This pruning mainly removes multiple and potentially confusing error messages arising from a single apparent error.
ECRs are sent from the DCC to the Site by e-mail. Each e-mail containing ECRs is structured as follows.
• The e-mail has a cover sheet (see Figure 4) containing the date the e-mail was sent by the DCC, a shipment number referring to the group of ECRs in this particular e-mail, and a listing of patient IDs (PIDs) and forms in the shipment. There may be many ECRs in a shipment.
• In the shipment (the e-mail), each apparent error usually gets its own page,
allowing the Study Coordinator ample room to make corrections. For example,
see Figure 5.
Sometimes a single page can refer to more than one apparent error. For example, suppose a "yes" answer to Item 2 requires answers to Items 3 and 4 but a "no" answer to Item 2 requires no answers to Items 3 and 4. If Item 2 is answered "yes" but Items 3 and 4 have been left blank, both Items 3 and 4 will be handled on the same page of the e-mail.
• Each page in the e-mail, referring to a single ECR, has a header section. The
– Identifies the ECR's subject by PID and Site. [In Figure 5, these are 100026-O76
and New York, respectively.
– Gives the date the ECR was sent out by the DCC. [In Figure 5, this is 03-05-
– Identifies the form by form number (e.g., Form 04, the
Randomization/Exclusion form) and subject Visit (if applicable), and by the
sequence number assigned to the form when it arrived at the DCC. [In Figure
5, these are Form OPT04, Visit is not listed, and sequence number 000101,
– Assigns a sequence number to this specific page in the e-mail, so each ECR is
identified by a unique sequence number. [In Figure 5, this sequence number is
• The main body of the ECR identifies the page on the form and the data item on that page where the error occurred. It reproduces the question, gives the value that was actually entered on the form, and describes the inconsistency or apparent error. [For example, in Figure 5, the error is that on page 2 of the form, the answer to Question 5 – the answer "yes" was entered on the form in question – is inconsistent with the answer to Question 5.]
Sometimes the ECR will contain a list of possible responses to the data item in question and invite the Study Coordinator to circle the correct value of the data item. This is illustrated in Figure 5. Also, sometimes there will be text comments further describing the nature of the error and/or the required information. In this example, a text comment was added stating, “This question should not be completed if question D is answered, “Yes.”
The above is not consistent with the following lead-in question:
Page 2 Question 5.
Did the subject fail to meet all eligibility criteria?
Value entered on form: 1
--> Circle correct value: 1 = Yes 2 = No
This question should not be completed if question D. is answered
III.15.2.2. Instructions for processing error correction requests (ECRs). Perform the following steps for each error correction.
(i) Retrieve the original form from the subject's folder/binder.
(ii) Check that the identifiers on the ECR match the identifiers on the form.
For example, the ECR is for Visit #3, and the form retrieved is for Visit #3.
(iii) Read the error correction thoroughly, including any comments.
All questions associated with the error need to be checked for accuracy, not just the first question printed on the ECR. It is possible that the first question printed on the ECR is correct, but only seems incorrect to the editing software because of the order in which the software processes the responses. Any of the associated questions could have an erroneous response. (iv) Understand the error correction.
If you do not understand why the ECR was sent or what the apparent error is, call the DCC with your questions or comments. The DCC tries to send out corrections or queries only for genuine errors, so if you do not understand the error or do not agree that there is an error, call the DCC and state your concerns. If, in fact, there is no error, the DCC will learn from your comments and avoid sending such non-errors in future ECRs. Sending back the ECR with "not in error" written on it does not resolve the problem: the ECR will be sent back and/or the DCC will call the Site asking why the ECR was returned to the DCC without a satisfactory resolution.
(v) Make the necessary corrections. For closed-ended questions (e.g., "yes/no" questions), if the possible answers are listed on the ECR, circle the correct answer.
For open-ended questions (e.g., gestational age at birth), or for closed-ended questions for which the ECR does not list the possible answers, write the correct value in the space provided.
Figure 5 illustrates a closed-ended question for which the ECR does not list the possible answers. To correct this error, the Study Coordinator does one of the following things:
• Cross out the "4" that was entered for Question 6 and write the correct value next to it; if the correct value was to leave the item blank, write "blank".
• Cross out the "5" that was entered for Question 5 and write the correct value next to it; if the correct value was to leave the item blank, write "blank".
• Cross out both the "4" entered for Question 6 and the "5" entered for Question 5 and write a text reply in the blank space.
(vi) Initial the error correction.
Writing initials in the space provided on the ECR ensures that the DCC will contact the appropriate person at the Site if there are additional questions concerning the error correction.
(vii) Return marked-up ECRs to the DCC.
Marked-up ECRs should be sent to the DCC in the next shipment of completed forms.
III.15.2.3. Site-Generated Error Corrections Sometimes Site personnel detect an error or otherwise need to change a value entered on a form, for which the DCC has not sent an error correction request. In such cases, the Site sends the DCC a site-generated error correction using the following steps.
(i) Make the necessary correction(s) and circle them in red ink on the Site's copy of the form.
(ii) Photocopy the page or pages containing the error.
(iii) Affix a label saying "Site-generated error correction" (provided by the DCC for this purpose) to the top center of the photocopied page, or write the words "Site-generated error correction" in the top center of the photocopied page.
(iv) Write in block letters the patient ID (PID), form date, and other identifiers for that form, for example, subject's visit, in the upper right-hand corner of the photocopy. Failing to write in the proper identifiers could result in the DCC mis-identifying the form and changing the wrong form.
(v) Check the photocopy to make sure all necessary information was copied legibly. Re-write any illegible information. Re-circle the correction in red so it will be obvious to DCC staff.
(vi) Send the photocopy to the DCC in the next regular shipment of completed forms.
III.15.2.4. Reminders on overdue error correction requests (ECRs) Periodically, the DCC creates a report of overdue ECRs for each Site. For a given Site, this report includes, for each overdue ECR, the patient ID, form number, visit number (if applicable), the ECR's shipment number, the date it was shipped from the DCC, the original form's sequence number (created at the DCC), and the error-correction sequence number.
Error corrections should not be made on the report of overdue ECRs. If the Site can no longer locate the page containing the original ECR, the Study Coordinator asks the DCC for a replacement of the missing page.
III.16. Data Analysis Requests by Enrollment Sites If any study personnel want study data analyzed or summarized for any purpose, they request the analyses by completing Form 70 (Data Analysis Request), described in detail in Section III.16.2 below, and transmitting it to the Data Coordinating Center (DCC). Use of this form ensures clarity in the request and fairness in determining the order in which analyses are performed by the DCC.
III.16.1. Appropriate Data Analysis Requests Most requests are for analyses for:
• Scientific presentations, abstracts, and papers, consistent with the Publications & Presentations policy (see Section IV.4.5);
• Local IRB consideration; and
• Local quality assurance efforts
Requests for other purposes may also be appropriate. Generally speaking, appropriate requests:
• Conform to Publications & Presentations policy (see Section IV.4.5) and
• Do not break the study's blind.
Requests for analyses for local use only will usually not require approval of the Steering Committee (SC). This is discussed in further detail in the instructions for completing Form 70, Part B (see Section III.16.2 immediately below).
The following requests will be routed to the SC for approval under the Publications & Presentations policy:
• Requests for data for use in abstracts, scientific presentations or papers, or
• Requests for data that is not limited to the requester's Site
III.16.2. Form 70 (Data Analysis Request) P art A identifies the person requesting the analysis.
• Item A.3 should be the phone number at which the requester can be called or left messages.
• Item A.5 indicates how the requester wants the analyses delivered. "Other, specify" includes things like overnight mail.
Section B describes the purpose and scope of the request.
• Items B.1 and B.2: If the data will be used within your Site or for a local presentation, including health fairs, check the "local use only" box in Item B.1 and indicate the type of local use in Item B.2. If your request is for a scientific presentation, either an oral delivery or poster presentation of scientific information (e.g., study design, methods, results, or interpretations) at a major scientific meeting, check the "Presentation to/for" box and indicate the scientific meeting on the line provided. Check the remaining boxes as appropriate, indicating where the paper, abstract, or other presentation will be submitted.
• Item B.3: Indicate the scope of the request by checking the appropriate box. If the request is for data from one or more specific Sites, list each Site.
Part C describes the data and analyses being requested. It will often be helpful to call or send e-mail to the Statistical Study Manager or the DCC's Director to discuss this before filling out Part C.
• Item C.1: If your request is not limited to a particular time interval, check the first box. If you want data only in a certain time interval, indicate the first and last dates that you want included in the data analyzed.
• Items C.2, C.3: Indicate if this analysis request is a modification to an analyses that was done previously. If it is, either attach the first page of that report to the Form 70 or indicate the report name and date from the upper left and right corners of the report. (Examples include the routine reports or special reports prepared for local IRB use or for local quality assurance purposes.) If there are any changes in the request other than the dates to be included, check "yes" and go to Item C.4. If the only change is in the included dates, check "no" and stop.
• Item C.4: If you only want the analysis to include a subset of the subjects at the indicated Sites, describe the subset here. Subsets of subjects can generally be selected based on any data item, e.g., date enrolled, race, whether they had a particular medical condition at baseline, etc.
• Item C.5: List the order in which to sort or group the data or analyses requested. You may specify page breaks as well, for example, "sorted by site and by PID with a new page starting for each site".
• Item C.6: List all data items or calculations to be included. Use specific form and item numbers whenever possible (e.g., Form 60, Item B.3).
• Item C.7: Use this area for any additional notes, comments, or sketches of tables or figures. Attach extra sheets to the Form 70 if necessary.