Contacting authors for missing data is the structured practice of emailing study authors to request statistics, raw numbers, or clarifications that a published report leaves out, so that an eligible study is not dropped from your synthesis simply because a number is absent. Done well, it is a polite, specific request with a clear deadline, logged so the whole effort is reproducible and reportable.
Why missing data is worth chasing
Studies often omit exactly the value you need: a standard deviation, a subgroup count, an event total, or the result for one of several outcomes. If you quietly exclude every study with a gap, your review tilts toward the better-reported papers and the result is biased. Recovering the missing numbers keeps your data extraction complete and lets more studies enter the meta-analysis rather than being stranded in a footnote. It also strengthens the credibility of your risk of bias assessment, since unreported results are themselves a bias signal worth resolving.
Before you email: exhaust the alternatives
An author request is slower than self-service, so try to recover the value yourself first. Check supplementary files, trial registry entries, and linked reports of the same study. If the number lives only inside a chart, try extracting data from figures before writing. And if a related statistic is reported, you can often derive what you need: reconstruct a standard deviation from an interval with a confidence interval calculator, or translate metrics with an effect size converter. Contacting authors is for what genuinely cannot be recovered.
How to write a request that gets answered
Be specific and make it easy
Vague requests get ignored. Name the study, the exact outcome, the time point, and the precise statistic you need, ideally referencing the table or figure number. The easier you make it to answer, the more likely the author replies. Where it helps, attach a short table for them to fill in rather than asking them to compose a response from scratch.
Reach the right person and set a deadline
Email the corresponding author first, and keep a co-author as a fallback if the primary address bounces or goes quiet. State a reasonable deadline, a few weeks is typical, so the request does not drift indefinitely and your timeline stays predictable, the same discipline we describe in how long a systematic review takes. For an older study where the email has lapsed, look for the author on an institutional page or an academic network before treating the contact as unreachable.
What a workable request looks like
The strongest emails share a predictable shape. Open by identifying yourself and the review, ideally citing its registration so the author can see the work is legitimate. State precisely what you need, anchored to the specific table or figure. Explain why, in one line. Offer an easy reply format. Close with a soft deadline. A request built on these parts reads as:
- Who you are: “I am conducting a systematic review (registered on PROSPERO) on post-operative pain after knee arthroplasty.”
- The exact ask: “For your 2021 trial, could you share the mean and standard deviation of the pain score at 12 weeks for each arm? Figure 2 plots it but the values are not tabulated.”
- Why: “We would like to include your trial in a quantitative synthesis rather than describe it narratively.”
- Easy reply: a small attached table with empty cells for mean, standard deviation, and n per arm.
- Deadline: “If you are able to reply within three weeks it would let us include your data in the analysis.”
Keep the message under a screen of text. Authors are far more likely to fill a short table than to draft prose, and a specific figure reference signals that you have read the paper rather than sent a mass mailing.
What to request, by data type
Asking for the right object saves a second round of emails. Tailor the ask to the synthesis you intend to run:
- For a continuous outcome, request the mean, standard deviation, and sample size per arm at the relevant time point, not just the difference between groups.
- For a binary outcome, request the number of events and the total analysed in each arm, which lets you compute an odds ratio or risk ratio yourself rather than relying on the reported effect.
- For a subgroup or a missing time point, name the exact stratum, since a vague “any further data” tends to be ignored.
- For unclear methods that affect appraisal, ask a precise clarification, for example how allocation was concealed, which directly informs your bias appraisal judgement.
Following up and recording the effort
Send a single, courteous follow-up
Authors are busy and emails get buried, so one polite follow-up after a couple of weeks is reasonable. Beyond that, repeated chasing rarely helps and risks souring the contact. Log the dates of each message so your effort is documented.
Record every request and outcome
Keep a simple table of which authors you contacted, what you asked for, and whether they responded. Reporting this is part of a transparent systematic review process and against the PRISMA 2020 guideline, a reader should be able to see exactly which data was sought and from whom.
Timing the outreach and tracking the replies
Author contact is one of the slowest steps in a review because it depends on someone else replying, so the timing matters as much as the wording. Send the requests as a single batch once extraction has confirmed exactly what is missing, rather than dribbling them out, so every author has the same window to respond and your follow-up dates stay manageable. Build a realistic buffer into the schedule: a three-week deadline plus a two-week follow-up means the outreach alone can add a month to the project, which is worth pricing into the protocol timeline.
A practical tracking sheet carries one row per request with the study, the corresponding author, the email date, the exact statistic sought, the deadline, the follow-up date, and the final status, which can be received, declined, no reply, or bounced. That table is what you summarise in the methods section, and it lets a second team member pick up the outreach without losing the thread. When a number does arrive, treat it like any other extracted field: enter it into your structured extraction form, record that it came from author correspondence rather than the published report, and have a second reviewer confirm it before it enters the analysis.
Be alert to one subtle bias. Authors of studies with favourable results may reply more readily than authors of studies that found nothing, so a high response rate is not automatically reassuring. Note who did and did not respond, and consider whether the pattern itself could skew the synthesis before you draw conclusions from the recovered data.
Planning for the data you never get
Many requests go unanswered, so decide in advance how you will handle a study whose missing data never arrives. Options include excluding the outcome, using a conservative assumption, or testing the impact in a sensitivity analysis that compares results with and without the affected studies. Setting this rule before you see the replies keeps the decision principled rather than outcome-driven, and aligns with the eligibility logic in your inclusion and exclusion criteria.