NCT01847274 — SoA extraction uncertainty report

Prompt 3.8.1 (single pass). Source: NCT01847274_soa.pdf (7 PDF pages = document pages 63, 64, 65, 66, 72, 73, 78 per PAGEMAP.md). No protocol markdown was available, so all text comes from the PDF text layer.

Decisions needed (7)

#	where	call made	alternative	detail
D1	p63, Table 1 row 1 "Cycle" (also Table 3 row 1, p78)	Header row labelled "Cycle" typed as a cycle row	Type it as an epoch row, because it also holds study phases (Screening, Discontinuation, Post Treatment)	2.3
D2	p66, Table 1 row 30 "Bone marrow aspirate and biopsy"	Top-level activity (indent 0); the single leading space in the source cell is treated as a typing artefact	Read the leading space as indentation, making it a child (level 1) of "Survival assessment"	2.3
D3	p72, Table 2 (printed Table 8)	Typed track, label "Food Effect Sub-Study" (separate sub-study population with its own FE Day 1/FE Day 8 visits)	Type as main_soa (independent schedule table, no track label)	3.1
D4	p73, Table 2 header markers 12, 13	Reprinted header de-duplicated; markers 12/13 carry the same text as 1/2 and are not emitted separately	Emit footnotes 12 and 13 as their own notes on the header cells	3.4
D5	p73, Table 2 full-width "Note: Patients in food effect sub-study…" row	Not an activity; one table-wide footnote, synthesised marker `n1`, anchored on header row 1	Bind it to specific activity rows, or leave its scope `unresolved`	3.3
D6	p78, Table 3 (printed Table 9)	Typed track, label "Extended Visit Cycle" (alternative later-phase visit pattern for main-study patients)	Type as main_soa, or as subsidiary of Table 1's "Subsequent Cycles" column	4.1
D7	p78, Table 3 markers 4–15 (rows 5, 7, 10–18)	Footnote texts are not printed in the excerpt; each note states "definition not printed" and adds a clearly-labelled probable equivalent from Table 1 where one exists	State only "definition not printed", no cross-reference	4.3

Recorded, not open (8)

Table numbering: table_number is sequential 1–3 across the protocol's SoA tables (schema definition); printed numbers are Table 7 → 1, Table 8 → 2, Table 9 → 3, recorded in each table_metadata.notes.
Printed Table 8 page 73 "(Continued)" is the same printed table under a reprinted header → extracted as one table spanning pp72–73, not a continuation (taxonomy: page break inside one printed table).
Reprinted header rows on pp64–66 (Table 1) and p73 (Table 2) de-duplicated (§1b).
Abbreviation blocks (Table 1 p63; Table 2 pp72/73) yield zero annotations: no term is printed as a marker in any grid cell, header or label (§6).
Leading label columns: L = 1 for all three tables; first data column = 2 (§5).
Synthesised property_names: Table 2 rows 1 "Period" and 2 "Cycle" (unlabelled header bands; §3), structure_method: inferred_from_layout.
Superscript "20" in the bone-marrow footnote ("WHO criteria²⁰", Table 1 fn 28 / Table 2 fn 15) is a bibliographic reference citation in the note text, not a table marker; kept as text.
Vertical placement of X marks inside the tall FOSI/EQ-5D-5L row (p65: Screening and Post-Treatment marks top-aligned, others centred) is presentation only — one row, all marks kept.

1. Method

Text layer present on all pages (not glyph-spread; no §1c reconstruction needed). No image-based pages.
Rule lines recovered from 200-dpi renders (§1d; used uniformly). Vertical rules gave 10 column bands (Table 1), 9 (Table 2), 6 (Table 3); horizontal row bands read from the label column.
Mechanical mark check (§1b): pdftotext -bbox tokens binned to raster column bands per row band; merged spans taken from missing internal vertical rules within each row band. The bbox matrix agreed with the visual read cell-for-cell on every table; no disagreements.

2. Table 1 — printed Table 7 "Schedule of Events – Main Study" (pp63–66)

2.1 Summary

main_soa. 9 data columns (2–10): Screening; Cycle 1 D1, D8, D15, D21; Cycle 2 D1; Subsequent Cycles (Cycle n, Day 1); Study Treatment Discontinuation; Post Treatment Assessments.
2 schedule properties (Cycle, Day); 28 activities; 90 schedule cells; 28 annotations (all footnotes).
Activity rows per page: p63 = 7, p64 = 11, p65 = 6, p66 = 4. Every page contributed.

2.2 Merged cells

Header row 1 "1" (Cycle 1) merged across columns 3:6.
"Bone marrow aspirate and biopsy" (row 30): one X²⁸ in a cell merged across columns 3:10 (Cycle 1 Day 1 through Post Treatment; Screening excluded). Distributed to 8 cells with source_range "3:10"; marker 28 placed on every covered cell.

2.3 Low-confidence calls

D1: row 1 values mix numbered cycles with phases; kept as cycle after the printed label "Cycle". Same call for Table 3 row 1.
D2: Only "Bone marrow aspirate and biopsy" has a leading space in the cell (p66); the table has no grouping headers and the row carries its own mark, so all rows are level 0 (indentation_method: assumed_flat). Raw leading space kept in cell_text.

2.4 Annotations

Markers on the header: 1 on the "Cycle" row label (schedule_property), 2 on the Subsequent Cycles header cell (col 8), 3 on the C1 "Day 1" header cell (col 3).
Footnote 14 ends "(Section 6.1)" but also explains — kept as footnote.
Footnote 19 has an unbalanced parenthesis "(within 30± 15 minutes Note:" — transcribed as printed.
No containment pairs: fn 10 (Table 1) and fn 14 (Table 2) share the same text ("SAEs recorded…"), and fn 28 / Table 2 fn 15 share the bone-marrow text, but these are in different tables and are source-faithful repeats.

3. Table 2 — printed Table 8 "Schedule of Events – Open-Label Food Effect Sub-Study (Study Completed)" (pp72–73)

3.1 Classification

track, track_label "Food Effect Sub-Study" (D3). Different population (sub-study patients) and different columns (FE Day 1, FE Day 8 ahead of Cycle 1). 8 data columns (2–9): Screening; FE 1; FE 8; C1 D1; C1 D15; C2 D1; Subsequent Cycles; Treatment Discontinuation.
3 schedule properties (Period [synth], Cycle [synth], Day); 17 activities; 79 schedule cells; 14 annotations.
Activity rows per page: p72 = 14, p73 = 3.

3.2 Merged cells

Header: "14-Day Food Effect" 3:4, "Cycle1" 5:7 (row 1); "C1" 5:6 (row 2). Screening, 14-Day Food Effect, Subsequent Cycles and Treatment Discontinuation cells span header rows 1–2 vertically; their row-2 grid cells are recorded as empty.
"Bone marrow aspirate and biopsy" (row 19): X¹⁵ merged across 3:9 (FE 1 through Treatment Discontinuation; Screening excluded), distributed to 7 cells.

3.3 Note row (D5)

The full-width note row at the foot of the p73 grid (rule-bounded cell spanning all columns) is one note, taken whole from its cell. Synthesised marker n1, location schedule_property row 1, method: synthesized; n1 added to row 1's annotation_markers.

3.4 Header reprint markers (D4)

The p73 reprinted header renumbers the Cycle1 / Subsequent Cycles footnotes as 12 / 13 (same text as 1 / 2). Header de-duplicated, so markers 1 and 2 stand; 12 and 13 are not emitted. Marker 1 is placed on each covered cell of the merged "Cycle" header (cols 5–7).

4. Table 3 — printed Table 9 "Schedule of Events – Main Study –Extended Visit Cycle" (p78)

4.1 Classification

track, track_label "Extended Visit Cycle" (D6). Main-study patients in a later phase following a different visit pattern (in-clinic extended cycle / telephone contact / local-clinic or in-home nursing), with its own column structure; it does not give finer timing for Table 1, so it is not subsidiary.
5 data columns (2–6). 3 schedule properties (Cycle, Day, Location = modality); 15 activities (all p78); 40 schedule cells; 15 annotations.

4.2 Merged cells

"Bone marrow aspirate and biopsy" (row 18): X¹⁵ merged across 2:6 (all data columns), distributed to 5 cells.

4.3 Orphan risk — markers without printed definitions (D7)

Markers 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15 are printed in the grid/labels but their footnote text is not on p78 (it ends after footnote 3; the following page is not in the excerpt). Each gets an annotation stating the definition is not printed, with a probable Table 1 equivalent where obvious (5→fn10, 6→fn12, 7/8→fn13/14 (order undeterminable), 9→fn16, 10→fn22, 11→fn24, 12→fn25, 13→fn26, 14→fn27, 15→fn28). Marker 4 (Vital signs, weight at the local-clinic column) has no obvious equivalent. All are placed at their printed locations, so none is an orphan.
Header markers 1, 2, 3 sit on the three "Subsequent Cycles" header cells (cols 2, 3, 4) and are defined on p78.

5. Synthesised items

Property names: Table 2 "Period" (row 1), "Cycle" (row 2).
Annotation markers: Table 2 n1 (note row).

6. Method provenance (non-default)

All activities: indentation_method: assumed_flat (flat tables, no grouping headers).
Table 2 schedule properties rows 1–2: structure_method: inferred_from_layout.
Table 2 n1 location: method: synthesized.
No unresolved locations; no proximity-bound note text; no visual-only cells (all marks confirmed from the text layer + raster rules).