What the Trial Did Not Ask

Over the following twelve days, Nora did not describe what she was doing as an investigation.

The distinction mattered. Investigation implied conclusion in search of proof. What she was doing, as she arranged it in her own mind, was review. She was reading documents she had access to, tracing a change backward through its approvals, and checking whether the logic of one methodological decision remained consistent with the logic of the system built on top of it.

At 1:00, on the Tuesday of the quarterly review, she sat across from Rena Moss in an office with a door and two chairs that matched the rest of the executive floor's preference for rounded edges and soft fabric that suggested calm without inviting rest.

Rena listened while Nora described the denominator shift and the threshold revision. Nora kept to the surface of it. She did not mention the Phase III footnote yet. She said only that tightening bereavement-cohort inclusion in the current quarter altered interpretive continuity with prior surveillance periods and might deserve a clearer internal rationale than the one attached to Version 12.5.

Rena folded her hands on the desk.

“The rationale is cohort specificity,” she said. “You saw that.”

“I saw the sentence,” Nora said. “I’m not sure the sentence and the effect are identical.”

Rena's expression changed by less than a degree. “The effect is fewer ambiguous intakes. That improves reliability.”

“For the recovery rate.”

“For the dataset.”

Nora nodded once. “Who requested the revision?”

“Data Governance raised the issue after site-level variability reviews.”

“Was Clinical involved?”

“Clinical is always involved.”

Rena let the answer sit there, complete enough to close a normal conversation. Then she added, in a tone that was not defensive and not open, “If you want the committee notes, I can see what's shareable.”

“Thank you,” Nora said.

Rena studied her for a moment. “Is this about the quarter's table,” she asked, “or is this about something larger?”

Nora could have answered incorrectly in several ways. She could have said it was routine curiosity. She could have said she did not know. She could have raised the protocol footnote and converted the meeting into something with shape too early.

Instead she said, “I think I need to understand the architecture before I know what the question is.”

Rena sat back. “All right,” she said. “Understand it.”

That evening Nora stayed at the office until after 8:00, which was late enough that the sound profile of the floor changed. Conversations disappeared first. Then the doors stopped opening. Then the printers, one by one. By 8:17 the only steady noise was ventilation and the intermittent soft chime of elevators arriving on levels where no one got off.

She reopened the original protocol and read it front to back.

Not the summary. Not the endpoint tables alone. All of it. Inclusion criteria, exclusion criteria, stratification logic, handling of missing data, adverse event categorization, the appendix on actigraph calibration, the statistical analysis plan. She took no notes for the first hour. She was trying to let the design reveal its own priorities without forcing a theory onto it.

The priorities were not hidden.

Scalability. Generalizability. payer relevance. occupational function. sleep restoration. distress attenuation. social reintegration. The language changed slightly across sections, but the pressure beneath it did not. The trial had been designed to answer the question all institutions wanted answered: when do people become functional again.

The thing it did not ask remained exactly where she had first found it, in a footnote small enough to be procedural.

Neuroimaging endpoints were considered and excluded due to cost constraints and logistical complexity in a multi-site trial design.

Nora opened a second window and searched the 2026 literature on grief-specific neuroimaging methodology. Twenty-three results in journals she recognized. Several multisite harmonization papers. Two NIH consortium studies. At least three validated approaches for measuring grief-cue processing and adaptive connectivity change over time. Expensive, yes. Logistically difficult, yes. Impossible, no.

She looked at the timestamp in the corner of her screen. 9:04.

Then she opened the trial's endpoint justification memo.

Functional recovery measures were prioritized because they mapped directly to real-world patient benefit and regulatory utility.

Regulatory utility.

Nora sat with the phrase until it separated cleanly into its parts. Real-world patient benefit was one category. Regulatory utility was another. The memo placed them side by side as if they were naturally aligned. Most of the time, in most trials, they were close enough to let the sentence pass without damage.

Here, she was beginning to suspect, they were not.

At home, the apartment was dark except for the kitchen light she had left on in the morning. She set her bag down, filled a glass with water, and stood at the counter without drinking it.

The photograph of her mother was where it always was now. Ming had placed it beside the fruit bowl in March and Nora had left it there because moving it would have required making a decision she did not have language for. The frame was thin wood. One corner of the backing paper had started to lift.

Nora looked at Liling's face and attempted, with the seriousness of a person repeating an experiment under controlled conditions, to retrieve the funeral.

Not the date. She knew the date.

Not the attendees. She had the program in a drawer.

The experience.

The room. The sound of chairs. The quality of the air. What her own hands had felt like. Whether her father had looked at her when the service ended. Whether she had cried in the car. Whether the flowers had been lilies or white chrysanthemums. She could reconstruct the sequence because sequence was data. What she could not retrieve was ownership. The memory existed in the third person. A record of a day on which certain things had happened near her.

She set the water glass down without drinking from it.

Then she opened her laptop again, at the kitchen table this time, and began searching for independent studies.

Pallinex + neuroimaging.

Pallinex + grief cue fMRI.

Pallinex + attachment network modulation.

Most of the first-page results were review articles, all citing the same broad functional literature and none mentioning direct grief-processing markers. On the third page she found a title from the University of Wisconsin that she did not remember seeing before:

Neural Processing Trajectories in Bereavement Patients Treated With Pallinex: A Preliminary fMRI Study.

The author was Amira Osei.

The sample size was small. Thirty-four participants. Sixteen-week observation period. Enough, if the design was careful, to show pattern. Not enough to alter practice on its own.

Nora downloaded the paper and read the abstract standing up.

By the second sentence she sat down.

The methods section was competent, unembellished, and precise in the way that suggested a researcher who had already argued with reviewers before publication. Standard grief-cue exposure paradigms. Longitudinal scanning. Control group matched on age, loss type, and time since bereavement. Functional metrics collected in parallel.

The results section contained one paragraph that changed the shape of the room.

Pallinex-treated subjects demonstrated marked suppression of amygdala-prefrontal connectivity during grief-cue exposure, with no evidence of the progressive connectivity normalization observed in the control group over the 16-week observation period. Functional recovery metrics were equivalent between groups by week 8.

Nora read the paragraph again. Then a third time.

Equivalent by week 8.

From the outside, the groups were the same. At work. Sleeping. Reporting reduced distress. Moving through the questionnaires as expected.

Inside the brain, one group was doing the work of mourning and the other was not moving at all.

Nora went to the citation index.

Fourteen citations in three years.

She clicked through them. Methodology papers. One review of longitudinal fMRI harmonization. A paper on affective attenuation pathways. Nothing in clinical psychiatry. Nothing in prescribing guidance. Nothing in the FDA post-market synthesis she had read the previous winter.

The study had not been argued with. It had not been disproven. It had been left to exist where it could not matter.

The next morning, a Thursday, Vantara's Employee Health portal reminded her of her quarterly wellness screening. The message had arrived two weeks earlier; she had deferred it twice because the available appointment windows conflicted with review meetings. At 10:40 she walked downstairs to the Adaptive Recovery Suite, checked in at a pale wood desk, and sat for four minutes under a wall display showing moving water and the words REGULATION SUPPORTS RESILIENCE.

The clinician who called her name was young enough that Nora had to suppress the reflexive arithmetic of career stage. Late twenties, perhaps. Friendly. Efficient. She wore a badge identifying her as a behavioral health coordinator and asked the questions from a tablet with the practiced calm of someone trained to sound non-directive while keeping the interview on time.

Sleep quality.

Appetite.

Concentration.

Work-related stress.

Question 14 appeared on the tablet when the coordinator turned it toward her for self-entry.

In the past four weeks, have you experienced intrusive thoughts about a past loss?

Below it were the response options:

Not at all.

Several days.

More than half the days.

Nearly every day.

Nora looked at the screen.

She had not experienced intrusive thoughts about her mother. She had experienced something that the questionnaire did not contain a category for: the organized absence of thought. The clean space around the fact of loss. Not avoidance. Not distress. Not forgetting. A structural blank.

The coordinator, noticing the pause, said gently, “Take your time.”

Nora marked Not at all.

It was the most accurate answer available. It was also, she thought as she returned the tablet, exactly how a system failed without ever becoming technically incorrect.

At lunch she ate at her desk and reread the Osei paper with the trial protocol open beside it. On one screen, exclusion due to cost constraints and logistical complexity. On the other, direct evidence that the excluded measure captured the one distinction functional metrics could not see.

She drafted an email to the corresponding author.

Dear Dr. Osei,

My name is Nora Chen. I am a senior biostatistician working in pharmaceutical outcomes research, with prior postdoctoral training in neuroimaging methodology. I recently read your 2028 paper on neural processing trajectories in Pallinex-treated bereavement patients. I am writing because I believe the study identifies a methodological gap with broader clinical implications than the current literature reflects.

She stopped there and deleted pharmaceutical outcomes research. It was true and too vague. She replaced it with:

I work in clinical trial analysis and have been reviewing outcome-measure selection in grief-related therapeutic frameworks.

Also true. Less searchable.

She finished the email, attached no documents, and read it twice before sending. The moment after Send was always slightly physical. A transfer of weight from private to external record. Small, but measurable.

That afternoon, Rena forwarded the Data Governance committee notes she had promised. The memo chain was ordinary in the way that real institutional decisions usually were. Site variability concerns. Intake ambiguity. Recommendation to improve specificity. Approval granted. No one in the chain had written anything that would survive as evidence of bad faith. No one had needed to.

What remained visible was preference. The preference for clean denominators. For stable rates. For data that traveled smoothly upward through decks and reviews and board summaries without carrying too much of the world inside it.

At 6:12 PM, Amira Osei replied.

Dr. Chen,

Thank you for your note. You are not the first person to contact me with this concern.

If you would like to speak, I am available tomorrow between 4 and 6 Central. I think a phone call would be more useful than email.

Best, Amira Osei

Nora read the sentence again.

You are not the first person to contact me with this concern.

She did not know, until that moment, how much of the last twelve days had depended on the possibility that she was wrong alone. Not wrong in the data; wrong in shape. Wrong in seeing a structural problem where there was only methodological untidiness. The sentence did not make her right. It made her not singular.

Friday moved badly. An oncology modeling meeting at 9:30. A compliance documentation pass at 11:00. An email from Commercial asking for a clean explanatory note on denominator variability in non-specialist language. She answered that one at 11:46 with three sentences that said nothing false and omitted everything that mattered.

At 5:08 Eastern she closed her office calendar, took her phone to the parking structure, and sat in her car for the call with Amira Osei.

The neuroscientist's voice was lower than Nora expected and more tired. Not defeated. Used.

“You found the paper,” Amira said, after introductions. “Usually people find the op-eds about grief treatment first and then work backward. You're going in the correct direction.”

Nora looked through the windshield at the concrete wall opposite her parking space. “The paper has fourteen citations.”

“Yes.”

“That seems low.”

A brief sound from the other end that might have been a laugh. “It is low.”

Amira explained what had happened to the study after publication. It had landed where it was sent. NeuroImage had reviewed it seriously. Methodologists had read it. Neuroimagers had cited it. Clinicians had not. The proposal for a larger replication study had gone through three grant rounds and failed at each one. Reviewers asked what clinical question the imaging would answer that had not already been answered by functional outcomes.

“The question,” Amira said, “was whether the functional outcomes were answering the wrong thing. But that's not a persuasive sentence to a panel built around functional outcomes.”

Nora said nothing.

“The literature is circular,” Amira continued. “Not maliciously. Structurally. If the approval framework is based on function, then the funded studies are based on function, then the review panels are populated by people trained to think in function, then imaging looks supplementary instead of diagnostic. The loop closes.”

Nora watched a man three rows over load groceries into the back of an SUV. “Do you still have the raw imaging data?”

A pause. Longer than the previous ones. “Yes,” Amira said. “Under data-sharing conditions. Why?”

“Because I think the original trial was designed in a way that made this effect unmeasurable by construction.”

Amira was silent for a beat. Then: “Yes.”

The word landed without emphasis. Not surprise. Confirmation.

Nora tightened her grip on the phone by less than an inch. “I've been looking at the endpoint selection memo.”

“And?”

“And I don't think this is a side effect of an otherwise valid design. I think the design selected for metrics that would classify suppression as recovery.”

Another brief silence. Then Amira said, “That is also my view.”

By the time the call ended, they had agreed to exchange formal materials the following week under an academic data-sharing framework that would give Nora access to anonymized imaging outputs and methodology documentation. Legal, standard, narrow. Enough to work with.

She remained in the car after hanging up. The parking garage lights had come on. The concrete surfaces held that particular evening brightness that made everything look more finished than it was. Her phone screen reflected her face back at her for a moment before going dark.

When she finally drove home, she did not turn on the radio.

The apartment was quiet in the familiar way of places occupied by one person for too long without interruption. She set her bag down, took off her shoes, and went to the kitchen.

The photograph was where it had been the night before. Her mother's hand still halfway lifted toward the camera. The edge of the frame catching the light.

Nora stood in front of it and understood, not emotionally and not yet all the way, that the clean space inside her was beginning to acquire contour. Not feeling. Shape. A missing thing described now by evidence precise enough to enter the body before the body agreed.

The trial had measured return to work, normalized sleep, reduced distress.

The trial had not asked whether grief had been completed or merely locked outside access.

Nora put both hands on the kitchen counter and remained there until the muscles in her shoulders began to ache.

Then she turned on her laptop, opened a blank document, and typed a title.

Methodological Concerns Regarding Outcome Selection in Pallinex Efficacy Assessment.

She looked at the words for a moment, then added the next line.

Primary issue: current functional recovery metrics cannot distinguish grief resolution from grief suppression.

She saved the file.

Outside, somewhere beyond the dark window above the sink, a siren moved past and faded. Inside, the kitchen light stayed steady over the counter, the photograph, and the first sentence of a record that did not yet exist but now would.