The Relationship Between Caffeine Consumption and Depression, Anxiety, Stress Level and Sleep Quality in Medical Students

This was an observational study, not an experiment. The researchers tested whether self-reported caffeine consumption (from any source: tea, coffee, solid foods containing caffeine, energy drinks, etc.) was associated with scores on three mental health measures (depression, anxiety, stress) and one sleep quality measure. There was no intervention — no one was asked to change their caffeine intake. The researchers simply surveyed a large group of medical students and looked for statistical relationships between how much caffeine they said they consumed and how they scored on validated questionnaires.

The primary outcomes were:

Depression score (from the Depression Anxiety Stress Scale-21, or DASS-21)

Anxiety score (from DASS-21)

Stress score (from DASS-21)

Sleep quality (from the Pittsburgh Sleep Quality Index, or PSQI)

The comparator was essentially "low caffeine consumers" versus "high caffeine consumers," though the study treated caffeine intake as a continuous variable in most analyses.

The study included 700 medical students at a single university in Turkey. The exact demographic breakdown is not fully detailed in the abstract, but the authors report that gender, age, class level (year of medical school), smoking status, family income, and place of residence were all recorded and analysed. The sample was drawn from a single institution, and all participants were medical students — meaning they are a relatively homogeneous group in terms of age (typically 18–30), education level, and likely socioeconomic background. No exclusion criteria are mentioned in the abstract, but the study was conducted during a specific six-week window (March 1 to April 15, 2022).

The researchers used two well-validated, standardised questionnaires:

1. **Depression Anxiety Stress Scale-21 (DASS-21):** This is a 21-item self-report questionnaire that measures three negative emotional states: depression, anxiety, and stress. Each subscale has 7 items scored from 0 ("did not apply to me at all") to 3 ("applied to me very much, or most of the time"). Scores for each subscale are summed and then doubled to match the full 42-item version. Higher scores indicate worse symptoms. The DASS-21 is not a diagnostic tool — it screens for symptom severity. Cut-off scores exist (e.g., depression: >9 mild, >13 moderate, >20 severe; anxiety: >7 mild, >9 moderate, >14 severe; stress: >14 mild, >18 moderate, >25 severe), but the study appears to have used continuous scores.

2. **Pittsburgh Sleep Quality Index (PSQI):** This is a 19-item self-report questionnaire that assesses sleep quality over the past month. It yields a global score from 0 to 21, with higher scores indicating worse sleep quality. A global score >5 is considered "poor sleep quality." The PSQI covers seven components: subjective sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, use of sleep medication, and daytime dysfunction.

Caffeine consumption was assessed via a face-to-face questionnaire that asked about consumption of tea, coffee, solid foods containing caffeine (e.g., chocolate, energy bars), and other caffeinated products over the past month. The exact questions and how they quantified "caffeine consumption" (e.g., cups per day, mg per day, or a categorical scale) are not detailed in the abstract.

**Study design:** This is a cross-sectional observational study. The researchers collected all data at a single point in time — they did not follow participants over time, nor did they assign anyone to a treatment or control group.

**Why that design matters:** Cross-sectional studies are excellent for generating hypotheses and identifying associations between variables. They are quick, cheap, and can include large samples. However, they cannot establish causality. The classic problem is the "chicken-and-egg" question: does caffeine consumption cause poor sleep and high anxiety, or do students who are already anxious and sleeping poorly turn to caffeine to cope? A cross-sectional study cannot distinguish between these possibilities. It only tells you that two things are correlated.

**Randomisation:** None. This is not an experiment, so there was no random assignment to groups.

**Blinding:** None. Participants knew they were being surveyed about caffeine and mental health, which could introduce social desirability bias (underreporting caffeine or overreporting distress) or recall bias.

**Duration:** The study was conducted over a six-week period (March 1 to April 15, 2022), but each participant only provided data at one time point. The PSQI asks about the past month, and the caffeine questionnaire asked about the past month, so the "window" of data is roughly one month per participant.

**Statistical approach:** The abstract mentions that factors affecting caffeine use were analysed, and that comparisons were made between groups (e.g., by gender, smoking status). The authors likely used t-tests or ANOVA for continuous outcomes and chi-square tests for categorical outcomes. They also likely used correlation analyses (e.g., Pearson or Spearman) to examine relationships between caffeine intake and DASS-21/PSQI scores. The abstract does not report whether they controlled for confounders like age, gender, smoking, or income in a multivariable model — this is a significant omission.

**What this design can prove:**

That a statistical association exists between caffeine consumption and mental health/sleep outcomes in this population.

The direction and strength of that association (e.g., "higher caffeine intake is correlated with higher anxiety scores").

**What this design cannot prove:**

That caffeine causes poor sleep or anxiety.

That reducing caffeine would improve sleep or mental health.

The temporal sequence (which came first: caffeine or the symptoms?).

Whether the relationship is causal, reverse-causal, or driven by a third variable (e.g., exam stress causing both increased caffeine and poor sleep).

**Major methodological weaknesses:**

No objective measure of caffeine intake (no blood or saliva levels, no dietary diaries — just self-report).

No objective measure of sleep (no actigraphy or polysomnography — just self-report).

Single university, single country — limits generalisability.

No control for important confounders like academic workload, exam period, or pre-existing mental health conditions.

The abstract does not report response rate (how many students were approached vs. how many agreed to participate), which could indicate selection bias.

The study was conducted during a specific six-week period — if this coincided with exam season, results might not reflect typical patterns.

The abstract reports the following key results:

**Caffeine consumption prevalence:** More than 80% of medical students reported consuming caffeine in the past month, primarily from tea, solid foods containing caffeine, and coffee.

**Factors associated with caffeine use:** Female gender, increasing age, class level (later years of medical school), smoking, higher family income, and place of residence were all associated with higher caffeine consumption. (No specific effect sizes or p-values are reported in the abstract for these associations.)

**Caffeine and anxiety/stress:** Caffeine consumption was higher among students with high anxiety and high stress scores. (Again, no specific correlation coefficients, p-values, or effect sizes are given in the abstract.)

**Sleep quality:** 73.1% of medical students had poor sleep quality (PSQI global score >5). As sleep quality worsened, caffeine consumption increased. (No specific numbers on the strength of this relationship.)

**Depression:** The abstract does not explicitly mention a relationship between caffeine and depression scores — only anxiety and stress are highlighted. This may mean the depression finding was non-significant, or it was simply not reported in the abstract.

**Critical note:** The abstract is notably sparse on actual numbers. We are told that "caffeine consumption increased in those with high anxiety and stress scores" and that "as sleep quality worsened, students' caffeine use increased," but we are not given:

The mean caffeine intake (e.g., mg/day or cups/day) for high vs. low anxiety groups.

The correlation coefficient (r) between caffeine and PSQI scores.

The p-values for any of these associations.

The mean DASS-21 scores for the sample.

The mean PSQI score for the sample.

This is a significant limitation of the abstract — and likely of the full paper — because it prevents the reader from assessing the clinical or practical significance of the findings.

Without specific numbers from the abstract, we cannot calculate precise effect magnitudes. However, we can make some reasonable inferences:

**Poor sleep prevalence:** 73.1% of students had PSQI >5. This is a very high rate, but it is consistent with other studies of medical students worldwide (typically 50–70% report poor sleep quality). This tells us that poor sleep is the norm, not the exception, in this population.

**Caffeine-sleep relationship:** If the correlation is typical of what other studies find (r ≈ 0.2–0.4), then caffeine might explain 4–16% of the variance in sleep quality. That is a small-to-moderate effect. In practical terms, a student who drinks 3 cups of coffee per day might have a PSQI score 1–3 points higher (worse) than a student who drinks none — but this is a guess.

**Caffeine-anxiety relationship:** Similar small-to-moderate effects are typical. A high-caffeine consumer might score 2–5 points higher on the DASS-21 anxiety subscale compared to a low consumer.

**What this means in plain English:** The relationships are likely real but modest. Caffeine is probably not the main driver of poor sleep or anxiety in medical students — academic stress, irregular schedules, and long study hours are likely much bigger factors. However, caffeine may be making an existing problem slightly worse.

**What the authors likely acknowledge (based on standard practice):**

Cross-sectional design prevents causal inference.

Self-report measures are subject to recall bias and social desirability bias.

Single-centre study limits generalisability.

The DASS-21 and PSQI are screening tools, not diagnostic instruments.

Caffeine consumption was not quantified in mg/day, making it hard to compare with other studies.

**What a critical reader would add:**

**No objective measures:** Without actigraphy or caffeine metabolites, we cannot be sure that self-reported caffeine and sleep are accurate. People often underestimate their caffeine intake (especially from tea and chocolate) and overestimate their sleep duration.

**No control for academic stress:** Medical students' caffeine and sleep patterns are heavily influenced by exam schedules, clinical rotations, and workload. The study was conducted over a six-week period — if this included exams, results could be very different from a non-exam period.

**No adjustment for pre-existing conditions:** Students with diagnosed anxiety, depression, or sleep disorders were not excluded or analysed separately. These conditions could independently drive both caffeine use and poor sleep.

**No dose-response analysis:** The abstract does not report whether there was a linear relationship (more caffeine = worse outcomes) or a threshold effect (e.g., only very high doses matter).

**No reporting of non-significant findings:** The abstract mentions anxiety and stress but not depression. If depression was non-significant, that is important information — it suggests caffeine might be specifically linked to arousal-related states (anxiety, stress) rather than mood.

**No effect sizes or confidence intervals:** This is a major flaw. Without these, we cannot assess the precision or practical importance of the findings.

**Potential for reverse causation:** Students with poor sleep may drink more caffeine to stay awake during the day, and students with high anxiety may use caffeine to self-medicate (or caffeine may trigger anxiety in susceptible individuals). The study cannot distinguish these.

**No information on timing of caffeine:** When students consume caffeine (morning vs. evening) matters enormously for sleep. A student who drinks coffee only in the morning will have a very different sleep profile from one who drinks it at 8 PM. The study likely did not capture this.

For someone running their own n=1 experiment:

**What to test:**

Test the effect of reducing or eliminating caffeine on your sleep quality and daytime anxiety/stress. A reasonable intervention is to cut caffeine entirely for a period, or to restrict it to before noon only. If you are a heavy consumer (3+ cups of coffee or equivalent per day), try tapering down over 3–5 days to avoid withdrawal headaches.

**Minimum meaningful duration:**

Run the experiment for at least 14 days per condition (e.g., 14 days of normal caffeine, then 14 days of reduced/no caffeine). Caffeine has a half-life of 3–7 hours, but withdrawal symptoms can last 2–9 days, and sleep patterns may take 1–2 weeks to stabilise after a change. A crossover design (A-B-A or A-B-B-A) is ideal: baseline (normal caffeine) → reduction → return to baseline → reduction again. This helps control for time-related confounds like exam stress.

**What to measure (specific metrics):**

**Sleep quality:** Use the PSQI (takes 5 minutes to complete) weekly. Also track: bedtime, wake time, sleep onset latency (minutes to fall asleep), number of night awakenings, and total sleep time. Use a sleep diary or a wearable (e.g., Fitbit, Oura Ring) for objective estimates.

**Anxiety and stress:** Use the DASS-21 weekly. Alternatively, use a daily 0–10 rating of "anxiety level" and "stress level" at the same time each evening.

**Caffeine intake:** Log every source of caffeine (coffee, tea, soda, energy drinks, chocolate, caffeine pills) in mg. Use standard conversions: 1 cup of brewed coffee = ~95 mg, 1 cup of black tea = ~47 mg, 1 can of cola = ~34 mg, 1 square of dark chocolate = ~12 mg. Be honest — even small amounts matter.

**Confounders:** Track your daily stress level (e.g., "How stressful was today? 0–10"), hours of study/work, exercise (minutes), alcohol intake (drinks), and screen time before bed (minutes). These are major confounders that could explain changes in sleep and anxiety.

**Key confounds to control for:**

**Academic/work stress:** If you reduce caffeine during a low-stress period and then compare to a high-stress period, any changes could be due to stress, not caffeine. Try to run the experiment during a consistent period (e.g., not during exams).

**Withdrawal:** If you quit caffeine abruptly, you will likely experience headaches, fatigue, and irritability for 2–9 days. This will ruin your data. Taper down gradually (e.g., reduce by 50 mg every 2–3 days) or include a 1-week washout period before collecting data.

**Expectation effects:** If you believe caffeine is bad for your sleep, you might unconsciously report worse sleep during the high-caffeine phase. Consider blinding yourself — have a friend prepare your coffee (caffeinated vs. decaf) without telling you which is which. This is hard to do perfectly but worth attempting.

**Time of day:** Caffeine consumed after 2 PM has a much stronger effect on sleep than morning caffeine. If you are testing "caffeine vs. no caffeine," be consistent about when you consume it. Better yet, test "caffeine only before noon" vs. "caffeine any time."

**Other stimulants:** Nicotine, alcohol, and certain medications (e.g., ADHD stimulants) interact with caffeine. If you smoke or take stimulant medication, these will confound your results.

**What a positive result would look like:**

**Sleep:** A reduction of 2+ points on the PSQI global score (e.g., from 8 to 6, moving from "poor" to "borderline" sleep quality). A reduction in sleep onset latency of 10+ minutes (e.g., from 30 minutes to 20 minutes). An increase in total sleep