I am the lead biostatistician on a Phase III oncology trial, and when I compared the sponsor’s proposed endpoint harmonization against version one of the Statistical Analysis Plan and the EDC audit logs, I realized our chief medical officer was rewriting a non-significant trial into a positive one.

 

I am the lead biostatistician on a Phase III oncology trial, and when I compared the sponsor’s proposed endpoint harmonization against version one of the Statistical Analysis Plan and the EDC audit logs, I realized our chief medical officer was rewriting a non-significant trial into a positive one.

Three months before that morning, I sat at a workstation in the CRO’s biostatistics suite on the ninth floor and walked a junior statistician named Dev through reading a version-control diff.

“Pull the original SAP first,” I said.

“Then pull the amendment.”

He had both documents side by side on his monitor.

“Now tell me what moved,” I said.

He scrolled.

He stopped.

“The primary endpoint definition,” he said.

“Paragraph 4.2.”

I nodded.

“Read the original definition aloud,” I said.

He read it.

The endpoint was progression-free survival, defined as the interval from randomization to radiographic progression or death from any cause, whichever came first.

“Now read the amendment’s version,” I said.

He read the amended paragraph aloud.

The amendment added a clause: “excluding events adjudicated as non-treatment-related by the independent review committee.”

“What does the exclusion clause do to the hazard ratio?” I asked.

Dev paused.

He opened SAS on the adjacent screen and ran the log-rank test both ways.

“It moves the p-value from 0.062 to 0.041,” he said.

“From non-significant to significant,” I said.

“On the same data.”

I picked up a pencil from the desk tray and underlined the clause on the printout.

Dev said nothing.

I clipped the printout to the version-control log and put both in the locked drawer of my filing cabinet.

That same week I ran the blinded interim Kaplan-Meier curves for the Data Monitoring Committee.

The work required a closed door.

I sat at my desk with the laptop shut and the monitor tilted so no one passing in the corridor could see the survival curves.

I pulled the blinded dataset from the trial’s electronic data capture system, filtered for the intent-to-treat population, and specified the sensitivity analysis I had pre-registered in the SAP.

The curves diverged slightly at the twelve-month mark.

Not enough.

Not yet.

I saved the output as a PDF and encrypted it.

I printed one copy for the DMC sealed envelope.

I wrote the date on the envelope flap in blue ink: March 14, 2026.

I locked the envelope in the second drawer.

A biostatistician’s filing cabinet is her laboratory bench.

Every drawer holds a version of the truth the sponsor cannot edit after the fact.

The SAP version one was already on file with FDA.

The sponsor’s emailed edits were date-stamped in the CRO’s document management system.

The EDC audit log recorded every status change on every subject in the trial.

Two weeks after the interim analysis, the sponsor hosted a leadership dinner at a restaurant in Research Triangle Park.

Garret Lyle sat across from me.

He was the sponsor’s Chief Medical Officer, tall, silver-haired, with a voice that stayed one degree softer than every other voice at the table.

He refilled my water glass himself.

“Lupe,” he said, “you are the steady hand on this trial.”

He smiled.

I thanked him.

The CRO’s senior vice president raised a toast to the collaboration, and Garret set his napkin on the table.

He told the table he had been a physician for twenty-three years before moving to the industry side.

He said he still thought about patients first.

He said he trusted the CRO’s statistical team because they were careful.

He looked at me when he said it.

I set my water glass down on the white tablecloth and did not speak.

Two days after the dinner, I logged into the EDC to review adverse-event submissions for the weekly query cycle.

One report caught me.

Subject 10047.

A Grade 3 neutropenic fever.

It had been classified as “treatment-related” in the previous query cycle.

It was now classified as “unrelated.”

The reclassification had not triggered the DMC review flag.

I checked the audit trail.

The status change had been entered at 16:58 by a user ID I did not recognize.

The change reason field read: “Operational alignment per sponsor medical review.”

I printed the audit trail page.

I folded it once and placed it in the filing cabinet next to the SAP diff and the sealed DMC envelope.

The wall clock in the CRO operations room read 17:30.

The regulatory operations team was logging off for the day.

Someone’s printer hummed a last cycle and went quiet.

The fluorescent lights above the cubicle row buzzed at their usual pitch.

Seventeen-thirty was the daily submission cutoff.

After that hour, the regulatory queue closed until the next morning.

It was the most ordinary number on the clock.

My name is Lupe Salazar.

I am a biostatistician.

Garret Lyle thought a protocol amendment could harmonize an endpoint into a positive trial, but he forgot the SAP version one was already on file.

The SAP version one was the trial’s original contract with the truth.

I pulled it from the CRO’s document management system on a Tuesday morning, two days after the adverse-event reclassification.

I sat at my desk with the door closed, the monitor angled toward the wall, and a cup of black coffee I had not touched.

The document management system tracked every version, every author, every timestamp.

I opened the original SAP, dated fourteen months earlier, and placed it beside the proposed amendment on my screen.

Paragraph 4.2 had not merely been clarified.

The amendment inserted an exclusion clause that redefined which progression events counted toward the primary endpoint.

Under the original SAP, every radiographic progression event in the intent-to-treat population counted.

Under the amendment, events adjudicated as “non-treatment-related” by the independent review committee would be excluded.

I opened the sponsor’s email chain from the CRO’s regulatory inbox.

Garret Lyle had sent the first draft of the amendment language on January 9.

The CRO’s regulatory operations lead had replied with two edits on January 11.

Garret had accepted both edits and added a third clause — the exclusion language — on January 12.

The email was sent at 21:47.

It was addressed to the CRO’s regulatory lead and the project manager.

It was not addressed to me.

It was not copied to the DMC.

I saved the email chain as a PDF with the SMTP headers visible.

I printed one copy and placed it in the filing cabinet beside the SAP diff.

The coffee was still warm.

I did not drink it.

The EDC audit log was the third source.

I closed the email PDF and opened the electronic data capture portal in a new browser tab.

I accessed it through my trial-team credentials and filtered for Subject 10047.

The log showed six status changes on the adverse-event report in four weeks.

The first three were routine — query opened, investigator response, query resolved.

The fourth was the reclassification from “treatment-related” to “unrelated.”

The fifth was a retroactive update to the change reason field.

The sixth was a system-generated flag: “DMC notification suppressed — override by sponsor medical monitor.”

I read the sixth entry again.

Override by sponsor medical monitor.

The override had been entered by the same unrecognized user ID I had seen two days earlier.

I cross-referenced the user ID against the CRO’s access log.

It belonged to a contract data manager named R. Pham, hired three months ago on a temporary requisition signed by Garret’s office.

R. Pham’s access profile included edit rights to adverse-event status fields — a permission level that contract data managers on this trial did not normally receive.

I printed the audit log filter.

I printed the access log cross-reference.

I placed both pages in the cabinet drawer and turned the key.

The drawer lock clicked once.

That evening I drove home and sat at my kitchen table with the overhead light on and the rest of the house dark.

My husband was already asleep.

I opened FDA’s Guidance for Industry on Bioresearch Monitoring on my personal laptop.

I read the section on data integrity inspections under 21 CFR Part 312, Subpart D.

The guidance described the circumstances under which the Office of Scientific Investigations could initiate a for-cause BIMO inspection: “evidence of data fabrication, falsification, or selective reporting.”

I opened the Office for Human Research Protections website in a second tab.

I read the reporting requirements for unanticipated problems involving risks to subjects — 45 CFR 46.103(b)(5).

The regulation specified that a responsible institutional official must report to the IRB, to the sponsoring institution, and to OHRP any unanticipated problem that may have placed subjects at increased risk.

An adverse-event reclassification that suppressed a DMC notification was a problem that affected how the committee assessed whether the trial should continue.

Subjects on the trial extension protocol were receiving the drug based on a benefit-risk profile the DMC had not been allowed to recalculate.

I marked three passages with yellow sticky notes and closed the laptop at 11:40.

The kitchen faucet dripped once.

I did not wipe the counter.

The next morning, I requested a videoconference with the DMC chair.

Dr. Eleanor Tsao appeared on screen from her home office in Philadelphia.

Behind her was a bookshelf with the spine of Piantadosi’s Clinical Trials visible on the second shelf.

She wore reading glasses pushed halfway down her nose.

“I have a pattern I need to walk you through,” I said.

I shared my screen and showed her the audit log for Subject 10047.

I showed her the suppressed DMC notification and the override flag.

I showed her the contract data manager’s access profile and the requisition signature.

Dr. Tsao did not speak for twelve seconds.

She picked up a pen from her desk and wrote something on a yellow legal pad.

“Show me the SAP diff,” she said.

I showed her paragraph 4.2, the original and the amendment, and the p-value shift.

“And the sponsor’s email chain?” she said.

I showed her Garret’s 21:47 email with the exclusion clause added.

Dr. Tsao set her pen down on the legal pad and removed her reading glasses.

“Lupe,” she said, “the committee needs to see this formally.”

“I know,” I said.

“I am preparing a written notification to you, to the CRO’s Quality Assurance lead, and a parallel report to FDA OSI.”

She nodded once and ended the call.

Garret Lyle believed operational refinements at the endpoint level were routine in late-stage pharmaceutical development.

He had shepherded two drugs through NDA submissions before this trial.

Both times, protocol amendments had adjusted secondary endpoints without incident.

He was confident that a CRO-side biostatistician — even the lead — would not move faster than her CRO’s commercial relationship with the sponsor allowed.

The CRO billed the sponsor fourteen million dollars a year for trial operations.

A biostatistician who threatened that contract would be managed through her project manager, not confronted directly.

That was Garret’s arithmetic.

He had not calculated that the SAP was already on file, that the emails were date-stamped, and that the EDC audit log could not be edited retroactively by a contract data manager whose access trail was already printed and locked in a drawer.

The NDA submission was scheduled for the following Tuesday.

The regulatory operations team had begun assembling the submission package.

At 17:30 on Thursday, the team queued the final dataset lock for Monday.

Once the NDA was filed, the harmonized endpoint analyses would become the formal FDA review record.

The hour stopped being a cutoff ritual.

It became the moment the trial’s statistical truth would be sealed inside a package the agency would read first under the manipulated definition — before the original SAP could be argued back into focus.

Seventeen-thirty was no longer the end of the workday.

It was the last clean hour the trial had left.

I sat at my desk after the operations team left.

I closed the EDC browser tab.

I placed the SAP version-control diff, the email chain printout, and the audit log pages in a sealed manila envelope.

I wrote the date on the seal in blue ink.

I picked up my desk phone and dialed FDA’s Office of Scientific Investigations BIMO reporting line.

The hold tone was a single repeating pitch.

I waited.

I did not hang up.

On Friday morning, a regulatory affairs email arrived in the CRO’s operations inbox.

The sponsor was accelerating the NDA submission by twenty-four hours.

The package would be transmitted Monday afternoon instead of Tuesday morning.

The email cited “regulatory-calendar advantage” and a Friday cutoff at the FDA Document Room that would give the submission a four-day head start on review assignment.

The email was signed by the sponsor’s VP of Regulatory Affairs.

It was copied to Garret Lyle.

I read it twice.

The acceleration compressed my timeline from four working days to two.

I walked to the regulatory operations suite and stood in the doorway.

The submission coordinator was reorganizing the electronic Common Technical Document modules on her monitor.

She looked up.

“Monday at 14:00,” she said.

“Dataset lock is being pulled forward to this afternoon.”

I nodded and walked back to my office.

I had already drafted the FDA OSI BIMO complaint, the OHRP unanticipated-problem notice, and the written notification to the DMC chair with the CRO’s Quality Assurance lead copied.

The BIMO complaint referenced the SAP version-control diff, the email chain, and the EDC audit log.

The OHRP notice cited the suppressed DMC notification as a potential unanticipated problem under 45 CFR 46.103(b)(5).

The documents sat on my desk in three separate folders — one blue, one white, one manila.

All three were ready.

The question was whether they would arrive before the NDA package left the building.

That Saturday evening, Garret Lyle appeared at an industry investor event streamed from a hotel conference room in Manhattan.

I watched the stream on my laptop from the kitchen table.

He stood at the podium in a dark blue suit with no tie.

He held a wireless clicker in his right hand and advanced slides showing the trial’s enrollment curves and the subgroup analyses.

The slides did not show the original SAP endpoint definition.

They showed only the amended definition.

“The result is robust,” he told the room.

“Repeatable across subgroups.”

He paused for timing.

“Biostatisticians do the math,” he said.

“We make the medicine.”

The audience laughed.

Garret clicked to a slide showing the projected NDA timeline.

He told the investors the submission was on track for Monday.

He said the FDA review division had already signaled receptivity to the indication.

He said the drug would be the first-in-class therapy for this tumor type.

He set the clicker down on the podium and picked up a glass of water.

He drank without hurrying.

A questioner from the second row asked about the safety profile.

Garret said the adverse-event data was “clean and consistent with the known mechanism of action.”

He did not mention the reclassified neutropenic fever.

He did not mention the suppressed DMC notification.

He did not mention the endpoint amendment, the SAP version-control diff, or the contract data manager who had been given edit access to adverse-event status fields.

The stream ended with applause.

I closed the laptop.

I had seen the signs for eleven weeks.

The first was the SAP amendment language appearing without my initials on the version-control signature page.

I noticed it on the second day and assumed a routing error.

The second was the regulatory operations lead forwarding Garret’s emails to the project manager but not to me.

I noticed it on the ninth day and asked the PM, who said it was an oversight.

The third was the contract data manager’s user ID appearing on the adverse-event query log with permissions I had not approved.

I noticed it on the thirty-seventh day and submitted an IT access review request.

The request was closed without action two weeks later.

The closure note said: “Per sponsor medical office — access approved.”

For eleven weeks, I chose to interpret each sign as administrative friction.

I chose to believe that a sponsor CMO who called me “the steady hand on this trial” would not rewrite the endpoint behind my back.

I chose wrong three times.

On Sunday evening I sat at the kitchen table with the three folders arranged in a row.

The house was quiet.

My husband was reading in the other room.

I opened the secure reporting portal for FDA’s Office of Scientific Investigations and uploaded the BIMO complaint.

I sent the OHRP unanticipated-problem notice via certified email to the address listed in the 45 CFR 46.103 guidance.

I sent the written notification to Dr. Tsao with the CRO’s QA lead copied on every attachment.

Then I picked up my phone and called the CRO’s QA lead directly.

“Margot,” I said, “I have filed with OSI and OHRP and notified the DMC chair. The submission is scheduled for tomorrow at 14:00. I am requesting that the CRO delay until the DMC has reviewed the materials.”

Margot was quiet for four seconds.

“Send me the DMC notification,” she said.

“I will be in the office at seven.”

I hung up.

On Monday morning I drove to the CRO campus and walked through the lobby at 6:50.

The security guard nodded from behind his desk.

The elevator was empty.

I carried the three folders — blue, white, manila — in my briefcase.

At 9:00, an email from the CRO’s EVP of Clinical Operations announced a pre-NDA submission alignment meeting at 14:00 in Conference Room B.

The invite list included the sponsor’s senior leadership, the CRO’s senior leadership, the DMC chair by phone bridge, and — one name I had not expected — a representative from the FDA Office of Scientific Investigations, attending under OSI protocol as an observer.

The observer had been invited because my complaint was already in the system.

I put the briefcase under my desk.

The three folders were inside, in order.

I walked to the conference room to check the phone bridge connection.

Conference Room B had a long oval table with twelve chairs, a speakerphone in the center, and a projection screen at the far wall.

I arrived at 13:45.

The room was empty.

The phone bridge was active — a green light on the conference unit pulsed steadily.

I placed my briefcase on the floor beside the second chair from the end and sat down.

At 13:52, the CRO’s EVP of Clinical Operations entered and sat at the head of the table.

Margot Reeves, the QA lead, followed and took the chair directly across from mine.

She set a binder on the table and did not open it.

At 13:56, three members of the sponsor’s leadership team entered: the VP of Regulatory Affairs, the VP of Clinical Development, and a man in a gray suit I had not met.

They sat together on the far side of the table.

At 13:58, a woman in a navy blazer entered and took the chair nearest the door.

She placed a small notepad and a government-issue pen on the table.

She did not introduce herself.

Her badge lanyard read “FDA” above a photo ID I could not read from my seat.

At 14:00, Garret Lyle walked in.

He wore a charcoal suit and carried a leather portfolio.

He sat at the center of the sponsor side and placed the portfolio flat on the table.

He looked around the room and smiled briefly at the CRO’s EVP.

“Thank you all for making time on short notice,” he said.

The EVP nodded.

“Dr. Tsao,” the EVP said, “are you on the line?”

“I am,” Dr. Tsao said through the speakerphone.

Her voice was level and clear.

The EVP said the meeting had been called to discuss an internal quality matter raised by the CRO’s biostatistics lead regarding the trial’s primary-endpoint definition and adverse-event reporting.

Garret’s expression did not change.

“I’m happy to address any questions,” he said.

“Endpoint refinements are operational decisions consistent with standard FDA practice.”

I opened the blue folder.

“The Statistical Analysis Plan version one,” I said, “defines the primary endpoint as progression-free survival from randomization to radiographic progression or death from any cause.”

I placed the printout on the table.

“The proposed protocol amendment inserts an exclusion clause that removes events adjudicated as non-treatment-related by the independent review committee.”

I placed the amendment beside it.

“The exclusion clause moves the p-value from 0.062 to 0.041 on the same intent-to-treat dataset.”

I placed the SAS output on the table between the two documents.

Garret looked at the three pages.

“Those are internal working documents,” he said.

“The amendment follows standard practice for endpoint clarification.”

“FDA practice does not refine an endpoint mid-trial in a way that flips a hazard ratio from non-significant to significant without Data Monitoring Committee review,” I said.

I opened the white folder.

“The EDC audit log for Subject 10047 shows a Grade 3 neutropenic fever reclassified from treatment-related to unrelated by a contract data manager whose access was requisitioned by your office.”

I placed the audit log printout on the table.

“The reclassification suppressed the DMC notification flag.”

I placed the access log cross-reference beside it.

“The contract data manager, R. Pham, had edit rights to adverse-event status fields that contract-level staff on this trial do not normally receive.”

Garret straightened in his chair.

“You are reaching, Lupe,” he said.

I opened the manila folder.

“An NDA package is a story a sponsor tells the agency,” I said.

“The SAP version one and the EDC audit log are two other stories. The DMC is the third. The submission cannot go on the harmonized endpoint.”

The room was quiet for three seconds.

Dr. Tsao’s voice came through the speakerphone.

“The Data Monitoring Committee is invoking its independent authority to pause the trial under Article 7.3 of the DMC charter,” she said.

Her tone did not rise.

“The committee will conduct a formal review of the primary-endpoint analysis under the original SAP and of the adverse-event reporting irregularities described in Ms. Salazar’s notification.”

The sponsor’s VP of Regulatory Affairs had been writing notes on a yellow legal pad throughout my presentation.

When Dr. Tsao said “pause,” his pen stopped mid-word.

He set it down parallel to the pad’s edge, aligned it with the corner, and did not pick it up again.

His folder was still open.

He closed it slowly with both hands and placed it flat on the table.

Margot Reeves had been sitting with her binder closed and her hands folded since the meeting began.

When the DMC chair finished speaking, Margot opened the binder to a tabbed page.

“The CRO’s Quality Assurance division is requesting that the regulatory operations team hold the NDA submission pending the DMC’s review and the outcome of the FDA OSI inquiry,” she said.

She looked at the EVP.

The EVP placed both palms on the table and nodded.

“The submission is held,” the EVP said.

The man in the gray suit — sponsor’s general counsel, I realized now — leaned toward Garret and said something low enough that I caught only the word “advisable.”

Garret shook his head once.

The FDA OSI representative in the navy blazer had been sitting near the door with her notepad open and her pen resting on the page.

When I placed the EDC audit log on the table, she had picked up the pen.

When Dr. Tsao invoked the DMC pause, she tapped the pen twice on the notepad, wrote one phrase, and underlined it with a single ruled line.

She did not speak.

She did not look at Garret.

Garret looked at the documents spread on the table in front of me — the SAP diff, the amendment, the SAS output, the EDC audit log, the access log.

He looked at the speakerphone where Dr. Tsao’s voice had just delivered the pause.

He looked at the FDA representative near the door, who was still writing.

His jaw moved once, as if he were about to speak.

He did not.

“I will refer further questions to sponsor’s outside counsel,” he said.

He picked up the leather portfolio, pushed his chair back, and walked out through the side door.

The door closed with a soft click.

No one at the table moved for four seconds.

The sponsor’s VP of Regulatory Affairs remained seated.

He spoke quietly, his eyes on the closed folder in front of him.

The sponsor’s audit committee, he said, would freeze the protocol amendment effective immediately, pending the DMC review and the outcome of the OSI inspection.

Outside counsel had been retained as of Sunday evening.

He did not look at anyone when he said it.

The EVP asked if there were additional items.

There were not.

Dr. Tsao’s line disconnected with a single tone.

Margot closed her binder.

The FDA representative put her notepad in an inside pocket of her blazer and left the room without speaking.

I gathered the three folders from the table — blue, white, manila — and placed them back in my briefcase in the same order I had packed them that morning.

The conference room emptied.

The green light on the speakerphone pulsed twice more and went dark.

Six weeks later, I started at a different CRO.

The office was smaller — a single room on the third floor of a concrete building near the interstate, with a window that faced a parking lot and a row of pine trees.

The fluorescent tube in the corridor hummed at a slightly different pitch than the one at my old building.

A whiteboard on the wall still had equations from the previous occupant written in blue marker.

The marker smell was faint but present when I opened the door each morning.

I had a new trial.

A different sponsor.

A different compound.

The disease area was hematology, not oncology.

The protocol was clean and the SAP had been finalized before enrollment began.

I ran blinded interim analyses on Tuesdays and Thursdays.

I locked datasets on Fridays.

I filed my reports with the DMC through the standard notification channel, and the notification flags worked.

The trial I had left was paused.

FDA’s Office of Scientific Investigations had opened a for-cause BIMO inspection of the sponsor’s clinical data management practices.

The Data Monitoring Committee had ordered a complete re-analysis of the primary endpoint under the original SAP.

The sponsor’s audit committee had frozen the protocol amendment.

The NDA submission was suspended.

The re-analysis and a follow-up confirmatory trial would take twenty-eight months.

During that time, seven patients remained on the trial’s extension protocol.

They had been receiving the experimental drug under the interim benefit-risk assessment the DMC had approved before the pause.

When the trial paused, the extension protocol was suspended pending the re-analysis.

One of those patients was a fifty-three-year-old high-school art teacher named Mr. Villanueva.

He taught ceramics at a public school in Albuquerque.

His students called him Mr. V.

He had been responding to the drug — his CT scans at months six and nine showed stable disease, and his oncologist had noted improved functional status in the clinical notes.

When the extension protocol suspended, Mr. Villanueva lost access to the drug.

His disease progressed over the following four months.

He died eleven months into the twenty-eight-month pause, at home, with his sister present.

His sister sent me a letter.

It arrived at the CRO’s general mailbox and was forwarded to my new office.

The envelope was a standard white business envelope with a handwritten return address.

The letter was one page, handwritten in blue ink.

She thanked me for stopping the trial manipulation.

She said her brother had understood why the pause was necessary.

She said he had told her, before the last scan, that the trial needed to be honest even if the drug took longer to reach the people who needed it.

I read the letter twice at my desk with the door closed.

I folded it along the original creases and put it in the top drawer of my new desk, beneath the blank analysis log forms.

The letter did not undo his absence.

The twenty-eight months would pass, and the drug would eventually be approved on the corrected analysis, and Mr. Villanueva would not be among the patients who received it.

At 17:30 on a Thursday, the regulatory operations team at my new CRO transmitted the interim analysis package for the hematology trial.

The package was clean.

The dataset lock had been completed at 16:00 under the original SAP with no amendments pending.

The DMC had reviewed the interim results through the standard notification channel.

No flags had been suppressed.

No audit log entries had been overridden.

I watched the operations coordinator confirm the transmission on her monitor.

The confirmation timestamp read 17:31.

Seventeen-thirty was still the submission cutoff.

It was still the moment the regulatory queue closed for the day.

But at this CRO, on this trial, the hour meant what it had always been supposed to mean — the end of a day’s work done correctly.

The coordinator closed the submission portal and logged off.

I stood by her workstation for another ten seconds and watched the screen go dark.

There was no triumph in watching a clean submission go through.

There was no relief.

There was only the difference between an hour that had been bent against the truth and an hour that closed on its own terms.

Garret Lyle thought endpoint harmonization was an operational refinement.

He forgot the SAP version one was already on file.

I walked back to my office.

The whiteboard equations were still on the wall.

The old marker smell had faded but not disappeared.

A printout of a Kaplan-Meier curve from the hematology trial was clipped to the corkboard above my desk.

The corridor was quiet.

I sat down and opened the analysis log.

I wrote the date, the trial code, and my initials in pencil.

I closed the log and turned off the desk lamp.