David Orentreich

Profession	Dermatologist. The documents read for this page give his name as “Dr David Orentreich”1 and as “Dr. Orentreich”2 and never with a degree. One sheet gives him a specialty: an office schedule reminder reading “to see dermatologist Dr. Orentreich at 2pm”3
Practice named in the files	Orentreich Medical Group, New York45
Dates carried by the documents read for this page	25 June 2010 to 10 August 2020. Sixteen of the seventeen are single messages, filings or statements dated between 26 November 2013 and 20 June 2016; the seventeenth is a compilation of card statements whose own closing dates run from 25 June 2010 to 10 August 2020167
Charge lines carrying the practice’s descriptor	17 February 2012 to 9 January 2018, on card statements6
Relationship to Epstein	None recorded. He is named in Epstein’s office correspondence as the dermatologist an appointment was booked with, and his practice is a merchant on card statements, not all of which name an account holder. Nothing read for this page records him writing to Epstein, meeting him, or acting at his request1236
Acts by him in the record	None found in any document read for this page

David Orentreich is a New York dermatologist named in files released under the Epstein Files Transparency Act. Seventeen documents reach his surname. Sixteen of them are single messages, court filings or statements, and those sixteen carry dates between 26 November 2013 and 20 June 2016. The seventeenth is a compilation of 9,824 sheets of card statements, and the closing dates printed on its own statements run from 25 June 2010 to 10 August 2020, so the dated material in the documents read for this page spans about ten years.⁷ The charge lines that carry his practice’s merchant descriptor run from 17 February 2012 to 9 January 2018.⁶ In all of these documents he is a name other people write down: a dermatologist recommended in a private message, an appointment Jeffrey Epstein’s office booked and reminded a correspondent about, and a practice that appears as a merchant on card statements, not all of which name an account holder.

A dermatologist who treats a patient has done nothing wrong. Nothing on this page is a finding of wrongdoing, and appearing in the released files is not evidence of it. Many people in these documents did nothing wrong. This page exists because the editor’s standing direction is that every doctor in the files should have an article, and what follows is the whole of what the documents record.

This page reproduces no clinical detail about any individual. Four of the seventeen documents are copies of one private message about one person’s own medical care. They were opened, they were read, and nothing from them is reproduced here or used anywhere on this page. Where a list of procedures appears below it is the practice quoting its own published terms of business, which is a fact about the practice and not about any patient. The identity numbers, addresses, telephone numbers, and medical and family details of every person in these files are outside what this site reproduces, and that applies to patients and to doctors alike.

How the dates on this page are read

Every date and clock time below comes from a stamp, and the stamps are of two kinds. A stamp that carries a zone is directly comparable. A stamp with no zone is not, and this page takes a date from one only after checking that the conversion cannot move it.

Where this page gives a span it names the unit being spanned and the scope it is spanning, because the two are different numbers here: the date a document itself carries is not the date of an entry printed inside it, and one of the seventeen documents is a compilation holding ten years of statements. Both endpoints of every span below were located on a sheet rather than inferred from the ends of a run, which matters because the compilation is not in chronological order: a statement closing in August 2020 sits on its printed page 9455 and a statement closing in November 2019 on its printed page 9823.⁷

The production’s own Sent: headers on these sheets carry no offset. Read as coordinated universal time and converted to the New York clock, each falls on the same calendar date either way, because none of them stands in the window where a conversion can roll a date backwards. New York was on Eastern Standard Time, five hours behind UTC, until Sunday 13 March 2016 and on Eastern Daylight Time, four hours behind, after it. At five hours that window is 00:00 to 04:59 UTC; at four hours it is 00:00 to 03:59. No stamp on this page falls inside either window. One of the sheets cited below fixes that boundary itself: Epstein’s office schedule prints “Reminder: Daylight Saving Time Begins” against Sunday 13 March 2016.³

Each footnote states which kind of stamp it relies on and prints the subtraction.

February and March 2016: an appointment his name was booked for

On 27 February 2016 a correspondent of Epstein’s answered a request for a dermatologist. The message was sent to Epstein and to a second recipient whose name the producing party has covered, and the writer’s own name is covered too. It names two doctors, and the first of them is “Dr David Orentreich”.¹ This message and the office schedule below are the only two places in the documents read for this page where anything is said about him beyond his name, and both give him as a dermatologist. The second doctor named in the same message is not named here.

Five days later the request reached Epstein’s office. On 4 March 2016 a message whose sender the producing party has covered asked his assistant to make an appointment with Dr Orentreich.²

Lesley Groff answered the same day. Her reply is the one document in the file in which Epstein’s office states a term of business with the practice:

“I have made you an appointment w/Dr. Orentreich for Wed. March 9th at 2pm. The appointment will cost $350 which you must pay the day of. If you want to purchase any medicine from them that day, you will also have to pay for it then.”²

The greeting before those words, and the name it carries, are covered on the sheet and are not supplied here.

On the morning of Wednesday 9 March 2016 Groff sent a reminder in almost the same words, and Epstein replied to it. His whole reply is two words: “all ok”.⁴ Later that day a message of thanks went back to Epstein. The sender is covered on that sheet, and the sheet carries no stamp from which a time of day can be taken.⁸

Epstein’s office schedule for that week carries the entry as a reminder line reading “to see dermatologist Dr. Orentreich at 2pm”, with the name of the person it concerns covered on the sheet.³

None of these four documents records the appointment taking place. They record it being asked for, booked, reminded about, approved in two words, and thanked for. A diary entry is evidence of intent, not of attendance, and the message of thanks does not say what it is for. The documents establish an arrangement. Whether it was kept is not in anything read for this page.

April 2016: the practice states its own terms

On 12 April 2016 the practice’s appointments mailbox answered an enquiry. The reply is signed “Orentreich Medical Group” and sets out the practice’s commercial terms in its own words:

“Please note, we do not accept any insurance except Medicare. An initial office consult with the practitioner is $350 and that does not include any treatment (injections, lasers, medications that we compound here in our pharmacy or chemical peels). We do provide you with an itemized receipt for you to give your insurance company.”⁵

The consultation fee the practice states is the same figure Groff had given a month earlier, and the two are independent of each other: one is the practice writing about itself, the other is Epstein’s office writing to a third party.

Of the documents read for this page it is the only one sent by the practice, and it is signed by the practice rather than by any individual. David Orentreich’s name does not appear on it.

2012 to 2018: the practice as a merchant

Four documents in the file are card-account statements, and they are not all the same account. Across every sheet of all four, the practice’s merchant descriptor appears fifty-one times.

The three numbers that could be meant by that figure are different numbers, so this page states all three: fifty-one occurrences, on forty-eight sheets, in four documents. Occurrences exceed sheets because three sheets carry two charge lines each. Fifty of the fifty-one have a posting date printed in the statement’s own date column; the earliest is 17 February 2012 and the latest 9 January 2018. The remaining one sits on a sheet that prints no transaction date at all. An earlier version of this page gave those two figures as forty-nine and two; that was a reading of the corpus text layer, where one of the two sheets splits its columns into separate vertical runs so that its date column stands eleven lines above its merchant column. The sheet itself was then opened at the image and it prints a ruled table with a date on the descriptor’s own row, so the count is fifty. The vocabulary that produced the fifty-one, the scope it was run over, the date forms the sheets print and the matching rule are set out in the footnote, with the per-arm counts and that correction.⁶

What the four documents are, separately, because they are not interchangeable. The largest is a compilation of American Express Centurion statements whose headers print the cardholder as Jeffrey Epstein and which carries more than one account ending, the shape of an account with supplementary cards. A second is an American Express Centurion statement covering 1 January to 20 June 2016 in sixteen sheets, on which no cardholder is named at all and the account is masked. A third is a single sheet, the eighth of ten, from a statement dated 11 February 2014 whose page footer carries Epstein’s name. The fourth is a run of forty-eight sheets of a different card account, not a Centurion one, which names a cardholder on its first sheet.⁶

Three things about the figure matter more than the figure.

First, a charge on an account is not a visit by the account holder. More than one account ending appears, which is what an account with supplementary cards looks like, and many of the lines sit beside merchants that plainly belong to different people’s days. Nothing read for this page establishes who attended the practice on any of those dates, and this page does not say.

Second, the four are not one account and should not be added as though they were. Three of the four name somebody on their own face, and the fourth names nobody at all: the word Epstein appears nowhere on its sixteen sheets and its account is masked.⁶

Third, no amount from any of those lines is reproduced here. The statement columns extract badly: the same sheets render amounts with digits run together, decimal points lost and dollar signs turned into figures, and across these lines the practice’s own name is rendered in at least four different damaged forms. An amount taken from that text layer would be a guess wearing a decimal point. The two fee figures this page does give, both $350, were read at the image on the sheets that carry them.²⁵

A case citation that is not about this

Three documents in the file reach the surname by an entirely different route. Each is a copy of a brief filed on 10 April 2016 in Giuffre v. Maxwell opposing a motion to disqualify counsel, and each carries, in a string of authorities on the same point, the citation Rosefield v. Orentreich, No. 98 Civ. 2721 (TPG), 1998 WL 567750 (S.D.N.Y. Sept. 4, 1998).⁹

That is a reported decision of 1998 cited for a proposition about disqualification motions. It is not about Epstein, it is not about any event in these files, and nothing read for this page establishes which Orentreich it concerns. The underlying case is not in the release. It is recorded here because a reader searching the surname will meet those three documents and is owed an account of what they are.

What the record does not show

The test this site applies to a subject is a named person, an act they took, and the document recording it. On the documents read for this page, David Orentreich does not pass it.

• Nothing read for this page was written by him. He is not the sender of any message in the file, and his name appears in no signature block.
• Nothing read for this page records him speaking, in any document, in quotation or in paraphrase.
• Nothing read for this page records a meeting, a visit, a referral made by him, an opinion given by him, or any instruction he gave or received.
• Nothing read for this page connects him to Epstein at all beyond being the dermatologist an appointment was booked with for somebody else and the name over a practice that took card payments.
• The one document the practice itself sent does not carry his name.

A name appearing in these files is not a subject, and the honest description of what these documents record about him is that it is thin. What they record about him is set out in full above rather than summarised, because there is little enough of it that a reader can check the whole of it. The documents themselves are not thin: one of them runs to 9,824 sheets, of which nine have been opened at an image by anybody working on this page, and the rest of that run has not been read.⁷

Outside the released files

Searches were made outside the record and are reported here as made, not as a survey.

He is living and in practice. His practice’s own site describes him as a partner at Orentreich Medical Group who “has practiced in New York City since 1984”, a graduate of Columbia University’s College of Physicians and Surgeons who served his dermatology residency at Mount Sinai Medical Center, board certified in dermatology in 1984, Co-Director of the Orentreich Foundation for the Advancement of Science, and “Clinique Laboratories’ guiding dermatologist” since 1987. The page gives his own words on his work: “By striving to achieve a healthy, youthful looking-skin and appearance for my patients, I can also improve their sense of well-being and confidence.”¹⁰

His father, Norman Orentreich, founded the practice and the foundation and died on 23 January 2019; he has a sister, Catherine Orentreich.¹¹ Neither name appears anywhere in the released files: measured against the released files, “Norman Orentreich” and “Catherine Orentreich” each return nothing.

No statement by David Orentreich about the released files, or about anything in them, was found. Two web searches were run and seven pages were fetched. Six returned HTTP 200 and were read in full, among them three articles about doctors named in the Justice Department files, and none of the three names him under any spelling. The seventh, a hospital profile page for him, returned HTTP 403 at 403 bytes and was not read; that is a failure to resolve on the attempts made, not a finding that the page does not exist or that it says nothing. That is the extent of the outside search made on 18 October 2026. It is not a survey, and it is not a claim that no such statement exists. No request for comment was made to him, and this page carries no response from him because none was found.

Name forms, and what the counts are counts of

The figures below are counts of documents containing a string. They are a floor on reach, not a measurement of a subject.

The surname “Orentreich” returns 17 documents. Every narrower form lies inside that set, measured rather than assumed: “Dr. Orentreich” returns 9 and adds none, “Orentreich Medical” returns 5 and adds none, “David Orentreich” returns 4 and adds none. Reversing the field order to “Orentreich David” returns 0, and so do “Norman Orentreich”, “Catherine Orentreich”, “Orentreich Foundation”, “David S. Orentreich”, “Orentreich MD” and a run of professional handle forms. A negative control, “Orentreichzzz”, returns 0.

The 17 partition exactly, and the partition sums to 17: four copies of one private message about one person’s treatment, which this page does not use; three copies of one court brief carrying an unrelated 1998 case citation; four card-statement documents; four documents of the March 2016 appointment thread; one office schedule; and one message from the practice. Of those seventeen, the four appointment-thread documents are the only ones in which anybody writes his forename.

The clipped form “orent” is not a name form here. It returns 21 further documents, and every one of them was opened. Not one concerns him. They are the extraction’s damage to ordinary words, by three mechanisms. On credit-application forms an unticked checkbox glyph set hard against the following word is read as a leading capital letter, so the option “rent” in an ownership-status row acquires a spurious initial O; that letter is the empty box itself and not any part of the neighbouring option, and where the box is ticked a different letter appears in its place. In brokerage-statement boilerplate and in securities tables the word “current” is damaged to “orent”. In a privilege log the word “Client”, in the column phrase “Attorney Client/joint defense”, is damaged to “Orent” in several different ways on a single sheet. A union of 38 documents can be formed from the surname arm and this clipped arm together, and 21 of the 38 reach nothing.

Three further strings that occur in the released files are the same merchant descriptor damaged in different ways on the card statements: ORENTREKH on thirteen sheets, ORENTRECH on one, and ORENTRUCH on one. All three are the extraction’s misreadings and not spellings anybody used, and that was settled at the image rather than assumed: one sheet carrying each of the three was rendered at 600 dpi, and all three print ORENTREICH MEDICAL on the paper.¹² All lie inside the single statement compilation already counted, so they add no document. They are recorded here because a reader searching the correct spelling will not find those sheets. They are not given in this page’s alias list, because a list of aliases is a list of names and these are not names.

The union was screened for container rows by size rather than by identifier, because a container row is a class and not a particular document. Every one of its 102 sheets was listed with its character count: the largest carries 12,403, three exceed 10,000, none exceeds 100,000, none carries a non-EFTA identifier, and the ordinary page in the set runs to about four thousand. The screen was proved able to fire on rows of the real container class, which run past a million characters. One document in the union is itself an aggregate of 9,824 sheets, but its individual sheets are ordinary pages.

Footnotes

1. Message of 27 February 2016, forwarded into Epstein’s office a week later. Read at the image at 400 dpi. The writer’s name and address, the second recipient’s name and address, and the salutation are all covered by the producing party’s bars; nothing is supplied for any of them. Below the line “Dr David Orentreich” the sheet carries a further bar of the producing party’s, about two lines deep, and a bar of the same shape stands below the line naming the second doctor; nothing from either is reproduced and nothing is supplied for either. Its inner header reads Date: February 27, 2016 at 15:53:54 MST, which carries a zone: MST is seven hours behind UTC, so 15:53:54 MST is 22:53:54 UTC, and on the New York clock that day (EST, five hours behind UTC) 17:53:54 on 27 February 2016. The date is 27 February on every one of those clocks. The quoted words run continuously on the sheet, broken only by its own line wrap, with no intervening content. https://epstein-data.com/EFTA02058274 p.1 ↩ ↩² ↩³ ↩⁴

2. Thread “Re: Dermatologists”, 4 March 2016, one sheet, carrying three earlier messages below the covering one. Read at the image at 400 dpi. The covering header is From: Lesley Groff, Sent: Fri 3/4/2016 5:11:42 PM, and it carries no zone: read as UTC and converted to the New York clock of 4 March 2016 (EST, five hours behind UTC) it is 17:11:42 − 5:00 = 12:11:42 on the same date, and 4 March either way, so the date is taken and the clock time is not. Three quoted preambles stand on the sheet. Two carry a clock time, “On Mar 4, 2016, at 9:13 AM” and “On Mar 4, 2016, at 12:10 PM”; both are rendered by the sending clients in local time and neither is converted or relied on for any interval. The third carries a date and no clock time. Groff’s words are quoted from the preamble that carries them; the greeting that opens that message is covered by a bar and the quotation here begins after it. The $350 figure was read at the image, and so was the whole of the quoted passage: the sheet prints a double space after “2pm.” and after “day of.” where the corpus text layer of this sheet has a single space in each place, and the quotation above reproduces the sheet. https://epstein-data.com/EFTA02058274 p.1 ↩ ↩² ↩³ ↩⁴ ↩⁵

3. Epstein office schedule, printed by day. The document runs to six sheets and covers 1 to 31 March 2016; printed page 3, which covers 10 to 17 March, is the only sheet read at the image for this page, at 400 dpi. The reminder line is the first content on printed page 3 and stands above the heading for Thursday 10 March 2016, and the heading for Wednesday 9 March 2016 is the last heading on printed page 2, so the line belongs to Wednesday 9 March; printed page 2 was not opened at the image and that reading of it is from the text layer. The Thursday heading is legible at the image and the corpus text layer damages it, rendering the year as 201M-12.1, so a checker will not find the date string on this sheet; the name of the person the reminder concerns is covered by a bar on the sheet. The sheet carries no machine-printed transmission stamp. The clock times on it are the schedule’s own entries, and two of those entries print an explicit zone, “(10:15am EST)” and “(1:30pm EST)”, beside times labelled “STT time”; no time of day is taken from any of them, and the 2pm in the reminder line is the schedule’s own text. The same sheet prints “Reminder: Daylight Saving Time Begins” against its heading for Sunday 13 March 2016, which is the authority for the zone boundary stated above. Other entries on the sheet concern other people and are not reproduced. https://epstein-data.com/EFTA00315398 p.3 ↩ ↩² ↩³ ↩⁴

4. Epstein to Lesley Groff, copy to a recipient whose name is covered, 9 March 2016, whose subject line names the appointment and the person it concerns, that name being covered on the sheet. Read at the image at 400 dpi. His whole message is “all ok”. The produced header Sent: Wed 3/9/2016 12:13:12 PM carries no zone: read as UTC and converted to New York on 9 March 2016 (EST, five hours behind) it is 12:13:12 − 5:00 = 07:13:12, on the same date. That conversion is corroborated on the sheet itself, which quotes the message he is answering at “7:03 AM” in a client-rendered preamble ten minutes earlier; an unconverted reading would put his reply five hours after a message he answers in ten minutes. The reminder stands undamaged on this sheet, and its last two sentences are word for word those of the booking quoted at ²; their opening clauses differ, which is why the page says the reminder was in almost the same words rather than the same ones. Below the quoted reminder the sheet carries the line ”>> Dr David Orentreich” followed by lines covered by the producing party’s bars; nothing from those lines is reproduced. https://epstein-data.com/EFTA02057888 p.1 ↩ ↩²

5. Orentreich Medical Group appointments mailbox, reply of 12 April 2016, subject “Re: Appointments at Orentreich Medical Group”. Read at the image at 400 dpi. The header Date: Tue, 12 Apr 2016 15:15:49 +0000 carries a zone: +0000 is UTC, and New York on 12 April 2016 was on EDT, four hours behind, so 15:15:49 − 4:00 = 11:15:49 on the same date. That conversion is corroborated on the sheet, which quotes the enquiry it answers at “11:05 AM” ten minutes earlier. The sender address, the enquirer’s address, and a contact value in the body are covered by bars. The quoted paragraph is contiguous on the sheet and the double spacing after “peels).” is the sheet’s own. The signature is the practice’s name and no individual’s. https://epstein-data.com/EFTA00534430 p.1 ↩ ↩² ↩³

6. Four card-account statement documents, described here from their own sheets rather than from each other. https://epstein-data.com/EFTA00590022 pp.7, 8, 11 and 16, an American Express Centurion Card statement of 16 sheets, its header period running from 1 January to 20 June 2016; the word Epstein appears on none of its sixteen sheets and its account is masked on the face of the document, so it establishes no cardholder. Printed page 16 of it was read at the image at 400 dpi for this page, and on that sheet the absence of a name is a reading of the paper rather than of the extraction: the field that would carry the cardholder’s name stands under a bar of the producing party’s, the account number is masked on the face of the statement by the card issuer, and every cell of the sheet’s own card-member column is barred. Nothing from any of those is reproduced and nothing is supplied for any of them. For the other fifteen sheets the absence of the name remains a text-layer reading. https://epstein-data.com/EFTA01578205 p.1, a single sheet, the eighth of ten by its own page numbering, carrying a statement date of 11 February 2014 and Epstein’s name in its page footer. https://epstein-data.com/EFTA01586218 p.37, one sheet of a 48-sheet run of a different card account, which carries no Centurion marking anywhere and names a cardholder on its first sheet; no account value, address or contact value from that sheet is reproduced here. https://epstein-data.com/EFTA01671962, a compilation of 9,824 sheets, 1,733 of them carrying the Centurion marking and its statement headers naming Epstein as cardholder under more than one account ending; each of its sheets also carries a second identifier of its own in the trailer. How the fifty-one was counted, so that a reader can reproduce it or refute it. The scope is every sheet of all four documents as the corpus holds them: 16 + 1 + 48 + 9,824 = 9,889 sheets. The unit is OCCURRENCES of the descriptor, not sheets and not documents. The vocabulary is the case-insensitive pattern orent(reich|rekh|rech|ruch) at a word boundary, which is the correct spelling plus the three damaged forms listed further down this page. Three further vocabularies were run over the same scope as a check and each returns the same fifty-one: a loose orentr plus up to four following characters; orent plus any characters plus med plus any characters; and a guard allowing a zero in place of the leading O. A fifth, requiring the descriptor to be followed by MED and a G, returns forty-nine occurrences, because two occurrences are cut short by the column extraction; that forty-nine is a count of the descriptor and is not the forty-nine that appears further down this footnote, which counts dated charge lines. A sixth, the bare five-character prefix orent, returns fifty-two, the extra being a wholly unrelated token in the compilation. The negative control orentreichzzz returns nought. The arms sum as stated: fifty-one occurrences break down 4 + 1 + 1 + 45 across the four documents in the order given above, and the forty-eight sheets become fifty-one occurrences because the compilation’s printed pages 696, 1016 and 7893 each carry two. The matching rule for “dated” is a posting date printed within three lines of the descriptor in the statement’s own transaction columns, excluding statement-period, closing-date and payment-due headings. The date vocabulary is three forms, which are the forms these four documents print: an abbreviated month name with a day and a four-digit year, as in Jun 17 2016; a numeric month-first form with a two-digit year, as in 02/17/12; and a bare numeric month and day with no year at all, as in 01/29. Of the forty-nine the text layer dates, forty-four are dated by the second form, three by the first and two by the third, and those three figures sum to the forty-nine. ISO, day-first and four-digit-year numeric forms were also run over these four documents, and none of the three puts a date within ten lines of any of the fifty-one. That is not the same as saying they occur nowhere: the four-digit-year numeric form occurs 437 times in the compilation, just never beside one of these charge lines, while ISO and the day-first form occur nowhere in any of the four. Run against the corpus text layer the rule gives forty-nine dated and two undated, at a window of one line, three lines and five lines alike. Both of the two were then opened at the image, at 400 dpi, and that is where the count changes. The thirty-seventh sheet of the third document is genuinely undated: it is a merchant-name run on the paper as well as in the extraction, its amounts are printed as a separate block of five against forty-one merchant rows, and the only date-shaped string anywhere on it is in the footer furniture, so position pairs nothing there. The sixteenth sheet of the first document is dated. On the paper it is a ruled four-column table headed Date, Description, Card Member and Amount, with four transaction rows, and the practice’s descriptor stands on a row whose date cell reads Jun 17 2016; all four rows carry that same date, so nothing turns on pairing a row to a column. The extraction splits that table into separate vertical runs, putting the four dates under their column header eleven lines above the descriptor, which is why a line window centred on the descriptor cannot see them. So the answer is fifty dated and one undated, the extra one established at the image and not from the text layer. An earlier version of this page said both sheets carried no transaction date at all; that was an absence drawn from an extraction and it is withdrawn. These posting dates carry no clock time and no zone, so no conversion applies to any of them and none is performed. The correction does not move either end of the range: 17 June 2016 lies inside 17 February 2012 to 9 January 2018, and both of those endpoints were re-derived under the vocabulary above and are unchanged. Nor can the one undated charge line move them, which is worth saying because an undated line reads as an unknown and this one is bounded: it sits in the 48-sheet run, whose own face prints a billing period of 12 January to 11 February 2014, and that period lies inside the range. Because the descriptor is damaged on these sheets in at least four ways, any count of it is a floor: a form damaged in a fifth way would not be reached by any of the vocabularies above, and the compilation was not read sheet by sheet. No amount from any of these lines is reproduced, because the same text layer renders the practice’s name on these sheets in at least four different damaged forms and renders adjacent amounts with digits welded and decimals lost. The 9,824-sheet compilation is 757,294,549 bytes at the public document route and was not rendered whole for this page; the merchant lines in it are text-layer readings and are described as such. ↩ ↩² ↩³ ↩⁴ ↩⁵ ↩⁶ ↩⁷

7. The dates carried by the documents. Sixteen of the seventeen carry one date each, or a short period, and they were taken from the sheets themselves: 26 November 2013 for the four copies of the private message, two of which carry a produced date header that states an explicit offset of zero, so those headers are zoned, are read as UTC, and convert to the New York clock of that day (EST, five hours behind, daylight saving having ended on 3 November 2013) by subtracting five hours; both stamps are afternoon stamps in UTC and neither falls inside the 00:00 to 04:59 window in which that subtraction could move the date, so 26 November stands on either clock. No header, clock time or other field of those four documents is transcribed here. Then 11 February 2014 for https://epstein-data.com/EFTA01578205 , from its own statement-date field; 12 January to 11 February 2014 for the 48-sheet run at https://epstein-data.com/EFTA01586218 , which prints that billing period on its face and repeats the same statement date on five of its sheets, and whose 8 March 2014 is a payment due date and not a date the statement covers; 27 February, 4 March and 9 March 2016 for the appointment thread; 1 January to 20 June 2016 for https://epstein-data.com/EFTA00590022 , from its own header; 1 to 31 March 2016 for the office schedule; 10 April 2016 for the three copies of the court brief; and 12 April 2016 for the practice’s reply. The earliest of those sixteen is 26 November 2013 and the latest 20 June 2016. The seventeenth, https://epstein-data.com/EFTA01671962 , is a compilation of 9,824 sheets holding many statements, and it is not in chronological order, so its span cannot be read off the ends of the run. Every one of the 9,824 sheets was scanned for the words Closing Date, which occur 5,490 times. That total is not a count of fields, and the five classes below are what it is made of. Each label was classified by what stands after it and before the next line break in the corpus text of its sheet, which is the whole rule and the only one used: 4,476 are followed by a clean MM/DD/YY or MM/DD/YYYY value; 4 by a value written with a month name; 447 by a run of four or more digits with no letters in it that no date form parses, which is the extraction damaging the value’s interior; 6 by nothing at all; and 557 by ordinary prose, these last being the statement’s own printed conditions, where Closing Date is a defined term and labels no value. Those five classes sum to the 5,490. Cutting across them, 551 of the 5,490 are Next Closing Date rather than Closing Date, and 170 of those 551 fall in the first class. Of the 447 damaged values none can be read, but the trailing two digits recover a year between 2010 and 2019 in 424 cases, recover an impossible value in 18 and recover nothing in 5, and the four that recover 2010 all recover a month later than June, so no damaged value falls outside the range given here. An earlier version of this page said that 4,548 of the labels had “a date beside them”. That figure does not reproduce under any reading of “beside” this page could state. Ten readings of the word, by character distance and by line, give figures from 4,318 to 4,837 and none gives 4,548, while an eleventh, requiring the value to stand alone on the line after the label, gives nought. Since the word named no rule a reader could apply, the figure is withdrawn and replaced by the classification above. The earliest and latest values located are 25 June 2010 and 10 August 2020. Both were then opened at the image at 400 dpi and read: printed page 33 prints Closing Date 06/25/10 and printed page 9459 prints Closing Date 08/10/20, the latter being the fifth sheet of a five-sheet statement that prints the same closing date on three of its five sheets. A single 05/17/10 was discarded, and that discard was settled at the image rather than argued from the text layer: printed page 6153 was opened at 600 dpi and the paper prints that sheet’s closing date as 05/17/19, with Next Closing Date 06/16/19 beside it and a second 2019 date standing under an early-pay heading below it, so the text layer’s 10 is a misreading of a 19 and no 2010 date earlier than 25 June stands behind it. Eight date forms were run over this document, and each is reported on its own rather than as a union, because a union hides a dead arm and hides a noisy one. Over all 9,824 sheets, as occurrences and as sheets: an abbreviated month name with a four-digit year, 84 on 56; a full month name with a four-digit year, 31 on 24; a month name with a day and a four-digit year, 816 on 411; a numeric month-first form with a four-digit year, 437 on 307; ISO, 0 on 0; a day-first form with a month name, 0 on 0; a numeric month-first form with a two-digit year, 105,099 on 8,224; and a bare numeric month and day with no year, 14,749 on 7,845. A negative control of the same shape as the first form, with letters appended, returns nought. The seventh form is the one that matters and the earlier version of this page did not name it: both of the endpoints given above are printed in it, and it outruns every other form on this document by two orders of magnitude because it is the form in which the statements print their own transaction dates. Its hits were read rather than counted, by classifying every one of them and then reading a seeded sample of thirty in full: 98,670 of the 105,099 are extracted lines consisting of nothing but the date, which is a transaction-date column cell; 4,474 stand on a line carrying a closing-date label; 91 on a payment-due line; 2 on a statement-period line; and 1,862 sit inside ordinary sentences, most of them the statement’s own autopay boilerplate. That 4,474 is a count of date strings on such lines and the 4,476 above is a count of labels carrying a value, which are two different questions over the same sheets and are not expected to agree exactly. The eighth form is almost entirely not a date on this document, and that is the reason to report it rather than to leave it out: 6,181 of its 14,749 hits are page-of-total furniture of the form p. 7/20, which is the same digit shape as a month and a day; 4,911 are a travel field giving a date of departure, which is a different field from a closing date and is not used here; 291 stand alone on a line; and the rest sit inside longer lines. It reaches no undamaged closing-date value at all; where it appears to reach one it has matched the surviving fragment of one of the 447 damaged values described above, which it does 137 times. A date probe set is never known to be complete, so a nought from one of these forms is a statement about the form and not about the document. Two of the eight return nought here and those are measured noughts over this document, not failures to look; the 447 damaged values are a named hole, and a closing date damaged in some further way would not have been found by any of the eight. What was not done, named and sized. Nine of the 9,824 sheets have been opened at the image by anybody working on this page: printed pages 33, 261, 9459 and 9823; printed pages 312, 605 and 655, recorded at the footnote above; and printed pages 6153 and 9455, opened later. The other 9,815 sheets have not been read by anyone here, and apart from the nine the figures in this paragraph are text-layer readings of a document whose text layer is demonstrably damaged. A closing date damaged beyond the eight forms above, or damaged in its interior as the 447 are, would not have been found, so 25 June 2010 and 10 August 2020 are the earliest and latest located, not the earliest and latest that exist. Its closing dates reach August 2020 while the charge lines carrying the practice’s descriptor stop in January 2018, and nothing in the years after that has been read for this page. ↩ ↩² ↩³ ↩⁴

8. Message to Epstein, 9 March 2016, the same subject line, sender covered by a bar. Read at the image at 400 dpi. Above the quoted chain its whole text is a single line of thanks. The produced header Sent: Wednesday, March 9, 2016 9:50 PM carries no zone: read as UTC and converted to New York on 9 March 2016 (EST, five hours behind) it is 21:50 − 5:00 = 16:50 on the same date; no time of day is taken from it and none is given in the body, because the stamp carries no zone. 21:50 does not fall in the 00:00 to 04:59 window in which a five-hour conversion could move the date, so 9 March stands on either clock. This sheet carries quoted-printable damage printed on the face of the document, w=ote, 2p= and =ave among others, which is why the appointment wording is quoted at ² from an undamaged copy rather than from here. https://epstein-data.com/EFTA01797358 p.1 ↩

9. Giuffre v. Maxwell, No. 1:15-cv-07433-LAP, Document 89, filed 10 April 2016, memorandum opposing disqualification of counsel, 10 pages. Page 5 read at the image at 300 dpi, where the ECF header and the printed folio agree. The citation stands in a string of authorities, between Norman Reitman Co. v. IRB-Brasil Resseguros S.A. and Laker Airways Ltd. v. Pan Am. World Airways, and is given with the parenthetical “(denying motion to disqualify where it was not clear that case would proceed to trial or that counsel would be called to testify; “too much remains unknown”)”. The three copies are https://epstein-data.com/EFTA02781211 p.5, https://epstein-data.com/EFTA00600136 p.5 and https://epstein-data.com/EFTA01182947 p.5; the first was read at the image and the other two at the text layer, where the same string stands. https://epstein-data.com/EFTA02781211 p.5 ↩

10. Orentreich Medical Group, “Dr David Orentreich”. Fetched 18 October 2026, HTTP 200, 196,545 bytes. Each quotation here was matched character for character against the fetched bytes, with and without its closing punctuation, and a reworded control was run and missed. The page is the practice’s own and is not an independent source about him. Contact values printed on it are not reproduced here. https://orentreich.com/about-us/dr-david-orentreich/ ↩

11. “Norman Orentreich”, Wikipedia. Fetched 18 October 2026, HTTP 200, 145,319 bytes, and re-fetched 27 October 2026, HTTP 200, 145,305 bytes. It gives his dates as 26 December 1922 to 23 January 2019, states that “He is the founder of the Orentreich Medical Group and the Orentreich Foundation for the Advancement of Science”, records that he founded that foundation in 1961, and names two of his children, David Orentreich and Catherine Orentreich. The quoted sentence was matched character for character against the fetched bytes and a reworded control was run and missed. This is a tertiary source and is cited only for the family relationship, the founding and the date of death, none of which bears on anything in the released files. https://en.wikipedia.org/wiki/Norman_Orentreich ↩

12. Three sheets of https://epstein-data.com/EFTA01671962, each read at the image at 600 dpi: printed pp. 312, 605 and 655, whose text layers render the merchant descriptor ORENTREKH, ORENTRECH and ORENTRUCH respectively. On all three the paper prints the same descriptor, beginning ORENTREICH MEDICAL. The damaged forms are therefore the extractor’s and not the document’s, which is the opposite of a spelling a person chose and is why they are recorded here as artefacts and kept out of this page’s alias list. The run was fetched whole for this purpose: 757,294,549 bytes, HTTP 200, and pdfinfo reports 9,824 pages, the same number of sheets the released text carries for it. The thirteen sheets carry fourteen occurrences of ORENTREKH, one sheet carrying it twice. No amount, account value or contact value from any of these three sheets is reproduced here. ↩